LIVE · refreshes every 20 min

updated Nov 14, 07:43 UTC

110010.art / archive / research / 2026-09

Research — September 2026

Sep 30 Forecasting space weather risks on power grids; Microsoft Research reports a new machine learning system predicts damage likelihood 30-60 minutes before a storm arrives Microsoft Research source ↗

Sep 30 MIT Transit Lab to create Public Transit Intelligence Hub with $2.1 million from Google.org Impact Challenge for government innovation MIT News (AI) source ↗

Sep 30 This game-playing AI becomes the new Stratego champion. MIT News (AI) source ↗

Sep 30 Energy Vision–Language–Action: a controlled multimodal benchmark for intent-conditioned residential energy management arXiv cs.LG source ↗

Sep 30 TypeSafe releases Jev, a commercial typed decision model, on 15 September 2026; review covers 28 papers on label-probability classification, constrained decoding, and calibration arXiv cs.CL source ↗

Sep 30 NanoForecast v0.5 improves forecasting accuracy with training pipeline fixes; achieves 6.5M-parameter performance competitive with 200M-parameter TimesFM under fixed protocol (MASE 3.030 to 1.704) arXiv cs.LG source ↗

Sep 30 Transformer MLP gate thresholds act as a reference for gate operation in mid-stack residual directions arXiv cs.CL source ↗

Sep 30 What drives dialectal jailbreaks? An ablation of surface form, cultural framing, and strategy banks arXiv cs.CL source ↗

Sep 30 IndicFDB extends full-duplex voice benchmarks to ten Indian languages with 12,350 samples arXiv cs.CL source ↗

Sep 30 What Next-Event accuracy cannot see: closed-loop evaluation of emergency department trajectory simulators arXiv cs.LG source ↗

Sep 30 Using LMs to model the effects of context and coreference on sentence comprehension arXiv cs.CL source ↗

Sep 30 Symmetry-quotient flatness implies input smoothness and generalization under local covering assumptions arXiv cs.LG source ↗

Sep 30 Don't Repeat Yourself: Self-supervised fine-tuning for coverage arXiv cs.CL source ↗

Sep 30 Measure Learning at steady state in a BIRD-SQL Formula 1 case study arXiv cs.LG source ↗

Sep 30 Energy-aware frugal Bayesian optimization arXiv cs.LG source ↗

Sep 30 Diversity-aware federated active learning under real-world distribution shifts (FIDAL) arXiv cs.LG source ↗

Sep 30 mu-bench: a multilingual utterance transcription benchmark for AI banking agent calls across English, Spanish, Turkish, Vietnamese, and Mandarin arXiv cs.CL source ↗

Sep 30 Superposition-based deep neural operator improves endwall film cooling generalization; arXiv:2609.31633v1 reports a physics-enhanced neural operator framework for variable film hole layouts arXiv cs.LG source ↗

Sep 30 ChestPheNoT: Deployable, auditable label-status-evidence extraction from radiology reports arXiv cs.CL source ↗

Sep 30 Information design against gaming and learning adversaries arXiv cs.LG source ↗

Sep 30 Extraction of clinical findings from mammography and breast ultrasound reports: a comparison between specialists and Artificial Intelligence arXiv cs.CL source ↗

Sep 30 Product-aware deterministic rounding for quantized matrix multiplication arXiv cs.LG source ↗

Sep 30 Verification of PETSc with CIVL using LLM-generated ACSL contracts and deterministic driver generation arXiv cs.CL source ↗

Sep 30 Autonomous research project management demonstrated by an agent skill on FFT-based KRR solver for regular spatial grids; NOAA SST anomaly fields analyzed on a 36 × 72 grid using consumer hardware arXiv cs.LG source ↗

Sep 30 When keywords drop but classifiers hold: soft refusals under KV cache compression arXiv cs.LG source ↗

Sep 30 Omni-IO skills: harnessing your agent Omni-native arXiv cs.CL source ↗

Sep 30 Grounding vision-language models in driving semantics: a multi-dataset predicate framework for explainable reasoning arXiv cs.LG source ↗

Sep 30 Contextual Early Reward predicts terminal reward from trajectory prefixes for long-horizon coding agents arXiv cs.CL source ↗

Sep 30 Distributional sentiment modeling and anomaly detection for consumer complaint assessment arXiv cs.CL source ↗

Sep 30 Parser, chunking, and embedding interactions in retrieval-augmented generation over Indian government regulatory documents arXiv cs.CL source ↗

Sep 30 Ongii-ni Eval-OW benchmark proposes benchmarking machine translation and large language models on Oshindonga and Oshikwanyama arXiv cs.CL source ↗

Sep 30 Active Causal Discovery Benchmark evaluates LLM agents under budgeted interventions arXiv cs.LG source ↗

Sep 30 Agents can use base models to evade AI detection arXiv cs.CL source ↗

Sep 30 Embedding-based psychometrics: semantic structure of assessment-item scores reveals a seven-group factor structure with a general dimension arXiv cs.CL source ↗

Sep 30 Literature-guided descriptor-based filtering framework for reducing composition search spaces arXiv cs.CL source ↗

Sep 30 STAR: Adaptive Spatial-Temporal normalization for unified microservice incident management arXiv cs.LG source ↗

Sep 30 Evaluation of AI-supported evidence-based learning for public speaking skill development; study develops context-aware evaluation criteria and generates improved model speeches in text and audio arXiv cs.CL source ↗

Sep 30 Quantization thresholds replicate, failure modes do not: a three-model study of agentic tool use in Polish from 8-bit to 2-bit arXiv cs.CL source ↗

Sep 30 JEPA pretraining for time series forecasting: evidence mixed on benefits arXiv cs.LG source ↗

Sep 30 From Phase Transition to Systemic Failure: A Decoupled Analytics Framework for GNN Robustness arXiv cs.LG source ↗

Sep 30 Cross-material support transfer for core-loss prediction under waveform covariate shift arXiv cs.LG source ↗

Sep 30 Communication between frozen large language models via prompt optimization in a referential game arXiv cs.CL source ↗

Sep 30 Age-adaptive handwriting reconstruction from an IMU-based digital pen through shared representations and domain-specific heads arXiv cs.CL source ↗

Sep 30 Beyond the Graph: an adaptive meta-learner fuses explainability, weather, and dynamics for robust bus eta prediction arXiv cs.LG source ↗

Sep 30 OMP-MoE: Efficient expert pruning for Mixture-of-Experts LLMs via orthogonal matching pursuit arXiv cs.LG source ↗

Sep 30 When does domain adaptation help on physical vibration sensors? a held-out-bearing study of neural-operator and convolutional models arXiv cs.LG source ↗

Sep 30 Replay in the silent degrees of freedom: continual learning without an offline phase arXiv cs.LG source ↗

Sep 30 EEGAgentBench benchmarks LLM agents on short- and long-horizon EEG analysis arXiv cs.LG source ↗

Sep 30 Typed Temporal Interaction Features for Simulation-Backed Forecasting of Open-Source Game Release Incidents arXiv cs.LG source ↗

Sep 30 MaD-RL: Matching Distributions for Calibrating LLMs with Reinforcement Learning arXiv cs.LG source ↗

Sep 30 Visualizing and detecting RAG conflicts in diffusion models via trajectory variance arXiv cs.CL source ↗

Sep 30 SMat-Attention: structured matrix attention for long-context sequence modeling arXiv cs.AI source ↗

Sep 30 AdaST: adaptive coupling for spatial-temporal forecasting arXiv cs.AI source ↗

Sep 30 Neurosymbolic routing for reliable reasoning on resource-constrained edge devices arXiv cs.AI source ↗

Sep 30 SAGE: a statistical acceptance gate for self-evolving agents arXiv cs.AI source ↗

Sep 30 More programs or more rolls? Separating coverage from specialization in LLM harnesses arXiv cs.AI source ↗

Sep 30 Mirror-Score: Calibrated, inference-only scoring reveals limits of sequence-compatibility ranking in D-peptide design arXiv cs.AI source ↗

Sep 30 GeoOutageBench benchmarks ambiguity-aware, ontology-grounded geospatiotemporal KGQA for multimodal power outage and resilience analysis arXiv cs.AI source ↗

Sep 30 More features are not more evidence: limits of training-free human activity recognition with Jev arXiv cs.AI source ↗

Sep 30 Self-discovering RL in the Era of Experience: learning from grounded interaction enables cumulative adaptation of the RL update rule arXiv cs.AI source ↗

Sep 30 Right Words, Wrong Moment: Clinician-grounded analysis of distress in 19,930 conversations between young people and ChatGPT arXiv cs.AI source ↗

Sep 30 Beyond symmetric agents: cognitive diversity and multi-agent debate in small language models arXiv cs.AI source ↗

Sep 30 Risk-averse online POMDP planning via CVaR of the immediate cost with performance guarantees arXiv cs.AI source ↗

Sep 30 OpenAI-HuggingFace: a reproduction and lessons for alignment testing arXiv cs.AI source ↗

Sep 30 Is human-readable text necessary for effective LLM fine-tuning? arXiv cs.AI source ↗

Sep 30 GeoWind2Plan: Mission-Time 3D urban wind prediction for energy-efficient UAV planning arXiv cs.AI source ↗

Sep 30 The price of token boundaries: compression certificates and prediction arXiv cs.AI source ↗

Sep 30 LongCat-DeepResearch: a deep research system combining an enhanced LongCat model with a multi-agent workflow for evidence-grounded reports arXiv cs.AI source ↗

Sep 30 Representational simplicity and circuit size dissociate in a threshold-dependent way; a controlled test via adversarial training arXiv cs.AI source ↗

Sep 30 A polyphonic conception of AI understanding explores how trust in model outputs relates to what the model actually understands, challenging monophonic notions of understanding. arXiv cs.AI source ↗

Sep 30 Exact generate-transform decomposition of small-LLM team scaling across orchestration architectures shows task-dependent accuracy gains from team scaling, up to 17 points across benchmarks arXiv cs.AI source ↗

Sep 30 Principled thoughts for latent recursive LLM systems arXiv cs.AI source ↗

Sep 30 Information-theoretical analysis identifies the layer(s) in VLA latent interface exposure across backbones for LIBERO and CALVIN benchmarks. arXiv cs.AI source ↗

Sep 30 Future-aware guidance for discrete diffusion with compiled objectives improves sequence-level training for parallel token resolution arXiv cs.AI source ↗

Sep 30 PowerZooJax: a JAX-based power system benchmark for reinforcement learning arXiv cs.AI source ↗

Sep 30 CADENCE: a confidence-adaptive dual-expert network for fast and accurate time series classification arXiv cs.LG source ↗

Sep 30 Intrinsic associative memory on Riemannian manifolds: curvature, capacity, and emergent modes arXiv cs.LG source ↗

Sep 30 Hybrid ensemble for EEG-based epileptic seizure forecasting uses five deep learning models and three classical models with logistic regression stacking arXiv cs.LG source ↗

Sep 30 Graph neural networks for sampling-invariant embeddings of organized signal sets arXiv cs.LG source ↗

Sep 30 Position: strengthen verifiability if we can't enforce reproducibility arXiv cs.LG source ↗

Sep 30 Mara Chain: rethinking failure as a stepping stone for AI system auto-evolution arXiv cs.LG source ↗

Sep 30 Learning in the Transverse Subspace: a minimal representation for divergence-free operator learning arXiv cs.LG source ↗

Sep 30 Binarization flattens the score space arXiv cs.LG source ↗

Sep 30 Calibration-first cross-cohort multimodal temporal learning for transferable asthma-risk forecasting arXiv cs.LG source ↗

Sep 30 EnJoi: ensemble joint score filter for generative data assimilation arXiv cs.LG source ↗

Sep 30 A mesoscopic view of transformer weights through row and column scale fields arXiv cs.LG source ↗

Sep 30 Prequential test-time learning for LLM agents enables updating knowledge at the granularity of individual transitions during deployment arXiv cs.LG source ↗

Sep 30 KernelOnet: an interpretable neural operator framework that incorporates kernel functions into the architecture arXiv cs.LG source ↗

Sep 30 FluxLite: Inference-time proposal control for discrete diffusion models arXiv cs.LG source ↗

Sep 30 The decision value of perception compute: allocating perception compute by downstream impact on loss arXiv cs.LG source ↗

Sep 30 Beyond keywords: using generative LLMs and label aggregation to classify economic policy uncertainty in news articles arXiv cs.LG source ↗

Sep 30 HeadGuard protects a subset of KV heads in low-bit VLM KV-cache quantization to preserve accuracy arXiv cs.LG source ↗

Sep 30 ReLOBGen: replayable limit order book message generation arXiv cs.LG source ↗

Sep 30 Learned compression of SAR phase-history data: autoencoder on-board sensor vs BAQ on GOTCHA data arXiv cs.LG source ↗

Sep 30 Sage: Formalization with semantic correction arXiv cs.LG source ↗

Sep 30 Serverless gossip training of LSTM failure detectors is compared with federated, local and centralized learning on NASA C-MAPSS arXiv cs.LG source ↗

Sep 30 Learning from the gap between Pass@K and Pass@1 arXiv cs.LG source ↗

Sep 30 Normative loss landscape navigation: a trajectory-based approach to mitigating forgetting in incremental learning arXiv cs.LG source ↗

Sep 30 Replication failure and unstable gains in road-level crash prediction via graph neural networks; replication study tests design decisions across three London boroughs arXiv cs.LG source ↗

Sep 29 Diffusion Controller unifies and simplifies AI image generation Google Research source ↗

Sep 29 AI agents push humans out of the loop arXiv cs.AI source ↗

Sep 29 Introducing Quine: an AI research system for multimodal biology modeling Microsoft Research source ↗

Sep 29 Who we become when we talk to machines MIT News (AI) source ↗

Sep 29 Algorithmic monoculture may be less problematic for job seekers than previously thought, depending on the approach used, MIT researchers find. MIT News (AI) source ↗

Sep 29 CP-Agent: a harness-engineered agent for crystal plasticity simulation workflows arXiv cs.AI source ↗

Sep 29 IndustryLLM: failure-driven LLM training for industrial procurement arXiv cs.AI source ↗

Sep 29 Countermem: world-model verified counter-factual memory for language agents arXiv cs.AI source ↗

Sep 29 Working with AI: A design framework for human-AI collaboration arXiv cs.AI source ↗

Sep 29 CSI-Agent: LLM-assisted few-shot adaptation for cross-domain Wi-Fi CSI sensing arXiv cs.AI source ↗

Sep 29 DriveHierarchy: a benchmark for diagnosing VLM driving capabilities from open-loop understanding to closed-loop execution arXiv cs.AI source ↗

Sep 29 Choir releases an open protocol for distributed formalization of AI-generated proofs arXiv cs.AI source ↗

Sep 29 A benchmark for LLMs' understanding of middle school and high school science topics arXiv cs.AI source ↗

Sep 29 Receiver-conditioned latent communication improves accuracy by transferring KV caches in multi-agent systems; arXiv paper introducing KV-cache based approach arXiv cs.AI source ↗

Sep 29 Metro-WM: Long-Horizon Latent Planning with Realisable Sub-Goals arXiv cs.AI source ↗

Sep 29 MedCode improves medical calculation in LLMs with embedded coding arXiv cs.AI source ↗

Sep 29 Witeness overlap: directional provenance inside open-weight model families arXiv cs.AI source ↗

Sep 29 Context-dependent agent evaluation with orthogonal equilibrium learning arXiv cs.AI source ↗

Sep 29 ConflictVLA-Bench benchmarks behavioral responses of vision-language-action models to premise conflicts arXiv cs.AI source ↗

Sep 29 BioDyad enables synchronization of biomedical discovery and machine learning engineering arXiv cs.AI source ↗

Sep 29 Decentralized Master-Mind: joint action refinement through iterative intent denoising in multi-agent pathfinding arXiv cs.AI source ↗

Sep 29 LLM judge validation with sparse overlap shows that low pairwise overlap drives wrong deployment decisions; increasing overlap quantity and quality are key levers. arXiv cs.AI source ↗

Sep 29 EmailBench benchmarks 206 enterprise email and productivity scenarios across 16 task categories using a typed email API in a provider-neutral environment arXiv cs.AI source ↗

Sep 29 EngramRAG: dynamic usage-weighted topology and synaptic consolidation for multi-hop agentic memory arXiv cs.AI source ↗

Sep 29 LoRA rank does not control capacity under per-factor norm budgets, as updates remain within the rank-at-most-r ball inside a nuclear-norm constraint. arXiv cs.AI source ↗

Sep 29 Symbolic guidance for LLM agents improves distributed multiagent coordination arXiv cs.AI source ↗

Sep 28 One year in: How Microsoft Research Asia – Singapore is advancing research, partnership and talent for real-world impact Microsoft Research source ↗

Sep 28 New formulation helps RNA vaccines withstand high temperatures MIT News (AI) source ↗

Sep 28 Rank-reliable teacher-guided fitness approximation for expensive evolutionary optimization in a TinyML neural architecture search study arXiv cs.AI source ↗

Sep 28 CRC-Router: risk-constrained routing for medical agentic AI systems arXiv cs.AI source ↗

Sep 28 Words Speak Louder Than Order: a behavioral evaluation of Gemma 4 arXiv cs.CL source ↗

Sep 28 Thinking less to simulate better: intuitive prompting improves LLM agents simulating individual social media reactions, including unfamiliar content arXiv cs.AI source ↗

Sep 28 Test-time reasoning in LLM trading: higher compute may not improve economic outcomes, study finds arXiv cs.AI source ↗

Sep 28 T-RoPE: Time-Aware Rotary Position Embedding for sequential recommendation arXiv cs.AI source ↗

Sep 28 Survey on fake review detection: from pre-trained language models to large language models arXiv cs.CL source ↗

Sep 28 SignTrace enables reverse lookup of Chinese sign language signs through natural-language dictionary access with LLM-assisted enrichment and seven-channel retrieval over 6,699 entries arXiv cs.CL source ↗

Sep 28 Stealth apart, harm together: skill cascading attacks on skill-based agent systems arXiv cs.AI source ↗

Sep 28 Recursive self-improvement via on-policy distillation for reasoning arXiv cs.CL source ↗

Sep 28 LLM Parkinsonism: executive-control failure, token-inefficient persistence, and an uncertainty-aware global executive control architecture for autonomous language-model agents arXiv cs.AI source ↗

Sep 28 Parameters vs. Context: TRACE fine-tuning for robust retrieval-augmented generation arXiv cs.LG source ↗

Sep 28 Adaptive multi-value control in LLMs via causal activation steering arXiv cs.LG source ↗

Sep 28 Inquesto Score: a reliability protocol for voice agents arXiv cs.CL source ↗

Sep 28 Why clipping matters in AdaGrad? Toward a high-probability theory under generalized smoothness arXiv cs.LG source ↗

Sep 28 When the preconditioning exponent turns negative: learning-rate coupling and cross-environment generalization arXiv cs.LG source ↗

Sep 28 Fake news theories: harnessing disciplinary insights for computational modeling, detection, and explanation arXiv cs.LG source ↗

Sep 28 ScopeBench tests whether agents preserve engagement boundaries under goal pressure arXiv cs.AI source ↗

Sep 28 Audio LLMs assess reliability of their own transcriptions; arXiv:2609.30625 shows models can recognize when transcription may be unreliable arXiv cs.AI source ↗

Sep 28 Cartograph: federated tool discovery with operator-attested retrieval for AI agents arXiv cs.CL source ↗

Sep 28 HybridInfer: Thermal-aware reinforcement-learning tier routing for on-device, edge, and cloud LLM inference arXiv cs.LG source ↗

Sep 28 Guarded gradient-based activation steering of shutdown responses in Qwen3.5-0.8B; a minimum-step policy arXiv cs.LG source ↗

Sep 28 Cosine similarity is not evidence: measuring the noise floor of interpretability transfer under quantization arXiv cs.LG source ↗

Sep 28 Bridging LLM agents and data spaces via the Model Context Protocol for architectural mediation arXiv cs.AI source ↗

Sep 28 Staged Depth Training introduces a representation curriculum that explicitly learns and progressively refines representations for PINNs. arXiv cs.LG source ↗

Sep 28 Pretrained ASR pseudo-labeling for noisy police audio arXiv cs.AI source ↗

Sep 28 Where does retrieval-based open-ended evaluation fail? Automatic taxonomy induction from long-form medical answer factuality verification arXiv cs.CL source ↗

Sep 28 BioEVAL: a global, multi-institutional benchmark of large language and multimodal models for bioengineering arXiv cs.AI source ↗

Sep 28 SlideLab: a training-free multi-agent framework generates and refines scientific presentations from research papers to present a coherent narrative and shared slide deck arXiv cs.CL source ↗

Sep 28 Fixed points without fixed diffusion: implicit neural sheaves for convergent test-time computation arXiv cs.LG source ↗

Sep 28 Code-switching curricula induce cross-lingual convergence in babyLMs; arXiv study reports cross-lingual alignment from multilingual code-switched pretraining arXiv cs.CL source ↗

Sep 28 Hierarchical collaborative memory for validity-aware retrieval in LLM agents arXiv cs.CL source ↗

Sep 28 Geospatial AI-based location selection for electric vehicle charging stations arXiv cs.LG source ↗

Sep 28 Atelier: learning local self-supervised features for cryoEM volumes via hypernetworks arXiv cs.AI source ↗

Sep 28 Improving molecular-morphology contrastive pretraining using deep-learning–based morphology profiles arXiv cs.LG source ↗

Sep 28 Neural ideals and neural codes: an algebraic framework for neural network classification and feature interpretation arXiv cs.LG source ↗

Sep 28 Study identifies neurons in a frozen BERT-base-uncased model that support AI-text detection using sparse probing on the RAID benchmark across six generators arXiv cs.CL source ↗

Sep 28 A synthetic ground-truth framework for the evaluation of explainable AI methods arXiv cs.AI source ↗

Sep 28 Predicting transmembrane protein topology from 3D structure using a graph neural network (SchNet) with atom-level embeddings arXiv cs.AI source ↗

Sep 28 Offline policy evaluation informs adaptive experiments by using historical A/B test data to assess adaptive policies via off-policy evaluation and warm-start simulation. arXiv cs.LG source ↗

Sep 28 From S3Q theory to implementation: a five-layer architecture for the S3Q theory of consciousness arXiv cs.AI source ↗

Sep 28 When Is a multi-agent code judge actually grounded? Two label-free measurements, and a judge that declines to guess arXiv cs.AI source ↗

Sep 28 Diversity-aware, layer-wise scoring for KV cache eviction improves ranking beyond mean attention by incorporating dispersion and redundancy penalties arXiv cs.CL source ↗

Sep 28 LAVOIR introduces value-of-information routing to guide one-pass decision models in querying missing information alongside answers. arXiv cs.AI source ↗

Sep 28 Seasonal and quantum-inspired models for neutron monitor time series forecasting arXiv cs.LG source ↗

Sep 28 When Does Advection-Aware Graph Nowcasting Help? A controlled study of distributed solar ramp forecasting with a self-supervised cloud-motion estimator arXiv cs.LG source ↗

Sep 28 TRACE: Temporal Audit and Condition-aware Evaluation of streaming video understanding claims a condition-aware benchmark for evaluating how evidence availability and history affect model responses arXiv cs.CL source ↗

Sep 28 PALM: Point-in-Time adaptation for financial language models arXiv cs.LG source ↗

Sep 28 DanLing NestedTensor: Composable multi-ragged tensors for deep learning arXiv cs.LG source ↗

Sep 28 From weak data to strong policy: Q-Targets enable provable in-context reinforcement learning arXiv cs.LG source ↗

Sep 28 ArXiv paper shows how unfaithful providers can manipulate self-consistency by inflating reasoning paths to charge more; strategic self-consistency arXiv cs.LG source ↗

Sep 28 Diversifying persona sets for creative LLM outputs to reduce response homogeneity arXiv cs.CL source ↗

Sep 28 A benchmark framework for screening automation in systematic reviews. arXiv cs.CL source ↗

Sep 28 MemProbe analyzes stability-plasticity tradeoffs in agent memory using a cognitive-science framework arXiv cs.CL source ↗

Sep 28 A unified account of concepts and chunks arXiv cs.CL source ↗

Sep 28 The hard part comes after search: benchmarking web agents on synthesizing, organizing, and displaying knowledge arXiv cs.CL source ↗

Sep 28 Auditing and repairing LLM-as-judge failures in a production text-to-SQL pipeline arXiv cs.CL source ↗

Sep 28 Spectral feedback for test-time alignment of protein diffusion models arXiv cs.AI source ↗

Sep 28 GAUDI: Geometry-Aware diffusion for calibrated air-quality time-series imputation arXiv cs.LG source ↗

Sep 28 Benchy proposes a universal language and execution engine for task-oriented AI benchmarks, using YAML-encoded benchmarks defined by programs, scoring functions, and datasets. arXiv cs.AI source ↗

Sep 28 Bringing AI to autonomous systems — from cognition to collective intelligence arXiv cs.AI source ↗

Sep 28 Learning what to skip: counterfactual credit assignment for efficient multi-agent LLM workflows arXiv cs.AI source ↗

Sep 28 Cost-aware best-LLM identification using dueling feedback arXiv cs.LG source ↗

Sep 28 REALMS: an AI-assistant conversational system for real-time exact audience sizing over high-dimensional nested profiles arXiv cs.CL source ↗

Sep 28 NeuralCert: certified computational discovery of extremal mathematical constructions arXiv cs.LG source ↗

Sep 28 LLMs’ context understanding evaluated with a knowledge graph-based framework arXiv cs.AI source ↗

Sep 28 Spotify researchers bootstrap conversational recommendation with synthetic data generation and self-improvement loops arXiv cs.CL source ↗

Sep 28 Newmark-η-DGN: learning coarse-step dynamics and internal mechanical response with graph networks arXiv cs.LG source ↗

Sep 28 HARDEN: constrained evolutionary search for harder, answer-preserving evaluation cases arXiv cs.AI source ↗

Sep 28 Inference-time target speaker unlearning in LLM-based automatic speech recognition arXiv cs.CL source ↗

Sep 28 Manifold projection and iterative autoencoder refinement for masked language modeling arXiv cs.CL source ↗

Sep 28 CARGO: Context-Aware Retrieval-Gated Evaluation of Agentic AI in Production evaluates reference-based LLM judgments for dynamic entities; identifies reference-instance divergence as a failure mode arXiv cs.CL source ↗

Sep 28 Analyzing and mitigating cost-inefficient behaviors in coding agents arXiv cs.AI source ↗

Sep 26 Learning to Discover Interesting Mathematics arXiv cs.LG source ↗

Sep 26 fable.intermittent: benchmarking probabilistic forecasting methods for intermittent time series arXiv cs.LG source ↗

Sep 26 CARE: Condition-aware representation regularization for diffusion models arXiv cs.LG source ↗

Sep 26 Physics-informed self-supervised learning jointly calibrates wires and reconstructs interaction position in multi-wire parallel plate avalanche counters arXiv cs.LG source ↗

Sep 26 UO-FIE: combining exact-label supervision with graded utility for factivity inference arXiv cs.LG source ↗

Sep 26 Leakage-safe machine learning for hydrogen embrittlement detection in 316L stainless steel using a region-held-out evaluation of texture and deep features in SEM micrographs arXiv cs.LG source ↗

Sep 26 Policy complexity, reaction time, and bounded rationality in reinforcement learning arXiv cs.LG source ↗

Sep 26 OPDiv: optimal selection of top-K high-scoring, diverse compounds arXiv cs.LG source ↗

Sep 26 SpaFactor: Lightweight spatial context-aware gene program modeling for histology-to-transcriptomics inference arXiv cs.LG source ↗

Sep 26 Graph Dynamics Model learns stochastic latent dynamics over evolving graph topologies arXiv cs.LG source ↗

Sep 26 Uncovering residential PV-EV co-adoption from smart-meter data: load archetypes and detection for demand-side planning arXiv cs.LG source ↗

Sep 26 CFD correction of open tip clearance flow in a compressor cascade using VAE latent space adaptation arXiv cs.LG source ↗

Sep 26 Explanations for RTL languages misrender SHAP and LIME visualizations; researchers measure and correct rendering for Urdu, Arabic, Persian, and Hebrew arXiv cs.LG source ↗

Sep 26 Uncertainty quantification for multi-step forecasting in time series foundation models (TSFMs) arXiv cs.LG source ↗

Sep 26 Stable and faithful explanations for knowledge tracing arXiv cs.LG source ↗

Sep 26 Auditability is not one property: rule overlap, behavioural agreement, and composition in reinforcement learning arXiv cs.LG source ↗

Sep 26 SMILESGNN: interpretable clinical toxicity prediction via SMILES-Graph cross-attention fusion arXiv cs.LG source ↗

Sep 26 Thinking leakage audit reveals contribution of base model thinking to NoThink post-training in hybrid reasoning models arXiv cs.LG source ↗

Sep 26 Time-series foundation model VINTAGE-TS handles data revisions by distinguishing observation time from information-availability time arXiv cs.LG source ↗

Sep 26 TAM-Chain: Multi-Scale thyroid cytology classification via absorbing Markov chains and Shannon entropy uncertainty quantification for false-negative suppression and domain-shift adaptation arXiv cs.LG source ↗

Sep 26 Federated learning of AnDE classifiers arXiv cs.LG source ↗

Sep 26 Evaluating cross-region generalization for wavelet-diffusion precipitation downscaling. arXiv cs.LG source ↗

Sep 26 Upholding robustness in federated learning: trends, emerging strategies, and research opportunities arXiv cs.LG source ↗

Sep 26 RLVR landscapes for iterated multiplications can be benign; insights from spin-glass theory arXiv cs.LG source ↗

Sep 26 LabFactory: building and evaluating executable AI labs arXiv cs.LG source ↗

Sep 25 MIT students gain a humanist lens on technical innovation in Tulsa, Oklahoma MIT News (AI) source ↗

Sep 25 RECLAIM: a benchmark of 100 NeurIPS 2025 papers that can be rebuilt yearly from new conferences to reproduce predefined results within a fixed GPU-hour budget arXiv cs.AI source ↗

Sep 25 Can Classical semantic-extractive summarization be evaluated in Hindi? A replication study arXiv cs.CL source ↗

Sep 25 ArGuard shared task evaluates harmful content in Arabic memes and LLM prompts; 58 teams registered, 35 participated in final, 27 submitted papers arXiv cs.CL source ↗

Sep 25 Forecasting agents’ reliability depends on when to retrieve, reason, defer to a market prior, or use a historical analog, per ForecastBench-style tests. arXiv cs.AI source ↗

Sep 25 Reasoning instructions can distort vision–language model evaluation by prompting CoT-prefix scoring, study finds arXiv cs.CL source ↗

Sep 25 When does action credit need updating? arXiv cs.AI source ↗

Sep 25 BanglaKontho: a 20-hour single-speaker Bangla TTS corpus derived from professional audiobook recordings to address long-form prosody gaps arXiv cs.CL source ↗

Sep 25 COILD: an Indic-centric parallel corpus and benchmark for machine translation across Indian languages. arXiv cs.CL source ↗

Sep 25 Reinforcement Learning with Verifiable Rewards applied to open-domain question answering with retrieval-grounded rewards; demonstrated on models below one billion parameters arXiv cs.AI source ↗

Sep 25 No More free lunch: corpus task complexity matters as corpora grow arXiv cs.CL source ↗

Sep 25 Where LLM graders succeed and break: evidence from two computer-science exams arXiv cs.CL source ↗

Sep 25 Benchmarking argumentative behavior of LLMs: a study of defences against character attacks arXiv cs.CL source ↗

Sep 25 Reward hacking challenges oversight of autonomous research agents arXiv cs.CL source ↗

Sep 25 Pistis model family introduced with 27B- and 9B-parameter multimodal LLMs built on Qwen, featuring Interleaved Distillation and Reinforcement Learning post-training framework. arXiv cs.AI source ↗

Sep 25 TWIST: a proposed benchmark for intervention quality in conversational memory with a human-validated draft-alignment arXiv cs.AI source ↗

Sep 25 AlphaDiverse: post-training local quantitative research agents for diverse exploration in alpha factor mining arXiv cs.AI source ↗

Sep 25 Tag-aware structured text translation: Towards a systematic understanding arXiv cs.CL source ↗

Sep 25 PTC-Bias: phoneme-level temporal competition for bias retrieval and post-decoding correction in speech LLMs arXiv cs.CL source ↗

Sep 25 Empath: tracing multi-level emotion dynamics in crisis counseling dialogues arXiv cs.CL source ↗

Sep 25 Bayesian personalized value alignment for LLMs moves from static personal values to contextualized personalization arXiv cs.AI source ↗

Sep 25 Grammatical "grandmother neurons" are rare in LLMs. arXiv cs.CL source ↗

Sep 25 Language models transfer capabilities through task-unrelated text via post-training; Active Taskless Distillation enables capability transfer without task-specific data arXiv cs.CL source ↗

Sep 25 Learning cross-task relationships in multi-task models via pairwise task-label relationships arXiv cs.AI source ↗

Sep 25 Control the harness, control the cost; enterprises scale AI coding agents and rely on external harnesses like Claude Code and Codex, per arXiv:2609.28919v1 arXiv cs.AI source ↗

Sep 25 Policy-driven Agentic World Simulation (PAWS) dataset covers 36 U.S. policy episodes, 12,727 policy-linked news records, and 65,291 source-grounded stakeholder actions arXiv cs.AI source ↗

Sep 25 Persuaded, not informed: incentive-misaligned witnesses defeat in-context grounding arXiv cs.CL source ↗

Sep 25 PFArena benchmarks language models for protein modification arXiv cs.AI source ↗

Sep 25 Polite but misaligned: evaluating LLM politeness judgments against human pragmatic norms arXiv cs.CL source ↗

Sep 25 Human-AI-powered hypothesis testing: cost-aware selective AI scoring and sequential human escalation arXiv cs.AI source ↗

Sep 25 MeshHeal: two-timescale self-healing for gray failures in decentralized LLM agent networks arXiv cs.AI source ↗

Sep 25 pylazaro: a Python package for anglicism extraction in Spanish arXiv cs.CL source ↗

Sep 25 TW3Cast: a frozen router of lightly fine-tuned foundation models for time-series forecasting on GIFT-Eval, selected entirely on the training split arXiv cs.AI source ↗

Sep 25 Predicting emerging topics from outliers: a prospective study of weak signals in embedding space arXiv cs.CL source ↗

Sep 25 Epydemix Agent Framework adds AI agent capabilities to Epydemix for epidemic modeling arXiv cs.AI source ↗

Sep 25 Framing by wording and selection in French news headlines, 2022–2025: a two-dimensional audit arXiv cs.CL source ↗

Sep 25 Script choice in LLMs: evidence for late-layer commitment arXiv cs.CL source ↗

Sep 25 BaseCamp presents an agentic AI framework for automating DNA sequencing data pipelines arXiv cs.AI source ↗

Sep 25 TRACER: a multi-turn user simulator that models evolving intent and aligns simulated behavior with real interaction trajectories arXiv cs.AI source ↗

Sep 25 Training Object Permanence in World Models arXiv cs.AI source ↗

Sep 25 EAGER: enhancing generative event extraction via reinforcement learning with verifiable rewards arXiv cs.CL source ↗

Sep 25 Temporal taxation compounds under post-training compression of Whisper models arXiv cs.CL source ↗

Sep 25 ELF-REG: Scaling continuous diffusion language models to reasoning tasks arXiv cs.CL source ↗

Sep 25 Adversarial closed-loop curriculum for evolving role-playing agents arXiv cs.AI source ↗

Sep 25 Toolkit for confidence-corpus consistency via fine-tuning on a fabricated corpus shows post-fine-tuning confidence aligns with fabricated answers for single-digit addition pairs arXiv cs.CL source ↗

Sep 25 DEEPO: Dual-Entropy Enhanced Policy Optimization for Hallucination in MLLMs arXiv cs.AI source ↗

Sep 25 An Explainable DistilBERT-BiLSTM-Attention framework for binary and multi-class hate speech detection arXiv cs.CL source ↗

Sep 24 Tool predicts suicide risk from text conversations with crisis counselors using language-processing method MIT News (AI) source ↗

Sep 24 Automating coherent long-form video generation using generative AI Google Research source ↗

Sep 24 Learned enterprise data comprehension: compression and routing for data agents arXiv cs.AI source ↗

Sep 24 4DGS-JEPA: a Gaussian-native joint-embedding predictive architecture for causal multi-horizon prediction over dynamic Gaussian scenes arXiv cs.AI source ↗

Sep 24 Making agents more consistent: skills should form habits for repeat tasks arXiv cs.AI source ↗

Sep 24 Orthrus enables efficient iterative retrieval with heterogeneous batching for embedding and generative models in a single serving system arXiv cs.AI source ↗

Sep 24 Rollout efficiency in reinforcement learning for reasoning LLMs: a taxonomy and future directions arXiv cs.AI source ↗

Sep 24 Ovis-Embedding: a native omni-modal embedding family that encodes text, image, video, and audio in a shared representation space using a single multimodal backbone. arXiv cs.AI source ↗

Sep 24 Robust failure, conservative repair: textual knowledge distillation from cross-model failures arXiv cs.AI source ↗

Sep 24 RAG-NAROK: Retrieval-aware knowledge corpus poisoning in RAG with source-specific refutation arXiv cs.AI source ↗

Sep 24 The AI Neuroscientist: an interactive agentic interface for neuroimaging analysis arXiv cs.AI source ↗

Sep 24 X-Planner: event-structured task planning for embodied intelligence arXiv cs.AI source ↗

Sep 24 Potential for enhanced learning in machine learning classes by using Wiki LLM indexing arXiv cs.AI source ↗

Sep 24 Spectra: a rules-driven LLM pipeline for automated KYC document processing arXiv cs.AI source ↗

Sep 24 Lean Pool: an AI-maintained archive of formalized mathematics arXiv cs.AI source ↗

Sep 24 Real-time hand gesture recognition for OpenXR using transformer-based machine learning arXiv cs.AI source ↗

Sep 24 Clarification is not correction: LLMs fail to let go arXiv cs.AI source ↗

Sep 24 ShowTellArena: evaluating comprehension after narrated business demonstrations with a 50-task benchmark dataset arXiv cs.AI source ↗

Sep 24 Sim-to-real validity of synthetic personas for predicting audience response; a no-persona baseline outperforms persona-based copy simulation arXiv cs.AI source ↗

Sep 24 LLM agents struggle with relational social reasoning; ReAdapt for Relational Social Reasoning proposes modeling how new relational evidence revises social hypotheses arXiv cs.AI source ↗

Sep 24 An accurate and interpretable hypergraph neural network for GBM survival prediction arXiv cs.AI source ↗

Sep 24 MedGate-Fusion integrates first-encounter narratives with routine biomarkers to prospectively stratify stroke risk in primary care arXiv cs.AI source ↗

Sep 24 From decorative to load-bearing: task difficulty shapes the causal role of chain-of-thought arXiv cs.AI source ↗

Sep 24 From offline proxies to online decisions: a layered engagement evaluation framework for conversational AI arXiv cs.AI source ↗

Sep 24 ZeroGate introduces trust-preserving fast paths for governed AI agent runtimes; separates exact-action approval from durable local admission with short-lived ActionPasses and a trusted runtime adapter arXiv cs.AI source ↗

Sep 24 Open Polymer Benchmark 2026 (PolyBench26) offers an open dataset with nearly 250,000 polymer-property datapoints across eight properties for ML-based polymer property prediction arXiv cs.LG source ↗

Sep 24 Benchmark dataset classifies interpretive canons at the sentence level using decisions of the German Federal Constitutional Court arXiv cs.CL source ↗

Sep 24 LEGO: synergizing expert GraphRAG and expert chain-of-thought for legal reasoning arXiv cs.CL source ↗

Sep 24 Spectral updates for depth-robust local learning improve CIFAR-10 MLP performance under local updates arXiv cs.LG source ↗

Sep 24 Phonemizing user-generated text: a benchmark, taxonomy, and compositional approach arXiv cs.CL source ↗

Sep 24 Signal2Symbol: Neuro-Symbolic Temporal Reasoning for Explainable Physiological Time-Series Anomaly Detection arXiv cs.LG source ↗

Sep 24 Local evidence and geometric readout repair in trained GNNs arXiv cs.LG source ↗

Sep 24 Count evidence, not sentences: tempered evidence fusion of LLM judgments for long-text value measurement arXiv cs.CL source ↗

Sep 24 Automated extraction of RoPA records using hybrid retrieval and locally deployed LLMs arXiv cs.CL source ↗

Sep 24 Realize what matters: principled context representation for large-scale reasoning arXiv cs.CL source ↗

Sep 24 LOCKR: a hidden-state trajectory-guided planner for detecting and repairing stable-but-wrong lock-in in diffusion language models arXiv cs.CL source ↗

Sep 24 COMED: The Missing Middle Between Routing and Collaboration in Multi-LLM Inference arXiv cs.CL source ↗

Sep 24 Guides that cause actions: an offline study of guide-action mutual reinforcement in multimodal web agents arXiv cs.CL source ↗

Sep 24 PR-Smoother: simulator-preserving amortized smoothing for data assimilation arXiv cs.LG source ↗

Sep 24 CRISP: scalable importance-stratified coresets for imbalanced tabular learning arXiv cs.LG source ↗

Sep 24 TinyUDE: solver-free universal differential equations on microcontrollers via Lie-Taylor jet matching arXiv cs.LG source ↗

Sep 24 Preference coverage collapse from hindsight relabeling in multi-objective reinforcement learning arXiv cs.LG source ↗

Sep 24 NADI 2026: the second multidialectal Arabic speech processing shared task announced arXiv cs.CL source ↗

Sep 24 What changes when fact-verification scores improve? Evidence and answer accounting across trained verifiers and LLMs arXiv cs.CL source ↗

Sep 24 HARN: Hierarchical Associative Resonance Network for event-driven multi-timeframe forecasting arXiv cs.LG source ↗

Sep 24 Tool-result caching can reverse group-normalized policy updates in a two-action model with shared stochastic results per group. arXiv cs.LG source ↗

Sep 24 LeakScale estimates the causal effect of benchmark exposure by creating fresh executable tasks to quantify performance attributable to training data contact. arXiv cs.CL source ↗

Sep 24 ISAAC: an open corpus and reproducible pipeline for analyzing social group discourse at scale using 527 million+ Reddit posts (2007–2023) arXiv cs.CL source ↗

Sep 24 LexLattice: multilingual extractive summarization via neural cellular automata on document hierarchies arXiv cs.CL source ↗

Sep 24 ChipMEM: Verification-grounded memory for EDA agents arXiv cs.LG source ↗

Sep 24 When learned context planning fails to beat strong retrieval in long-context QA; a controlled study of planning, routing, and reranking for LongBench-v2 MCQ questions arXiv cs.CL source ↗

Sep 24 Learning risk scores robust to unobserved confounders arXiv cs.LG source ↗

Sep 24 Distilling sequential computation in transformer language models arXiv cs.CL source ↗

Sep 24 Simulation-free reinforcement learning with Wasserstein-tilted flow maps arXiv cs.LG source ↗

Sep 24 Recognized but not produced: a generation benchmark for culturally specific kinship terms arXiv cs.CL source ↗

Sep 24 UniDataAgent: an ontology-grounded agent for enterprise question-to-report automation arXiv cs.CL source ↗

Sep 24 Transfer learning with conformalized quantile regression improves solar PV forecasting under data scarcity from load shedding; study uses Alice Springs data for pretraining. arXiv cs.LG source ↗

Sep 24 Leakage-aware multimodal evaluation framework for early intraoperative acute kidney injury prediction uses a waveform-only SynerT backbone to stratify risk during surgery arXiv cs.LG source ↗

Sep 24 LatWeave for deterministic multi-hop question answering on knowledge lattices arXiv cs.CL source ↗

Sep 24 Meta-learning for deep stacked generalization improves stacking of heterogeneous vision backbones (CNNs, ViTs, and hybrids) to address multicollinearity and calibration issues arXiv cs.LG source ↗

Sep 24 Can one adapted model do it all? Fine-tuning strategy selection for customer support LLMs arXiv cs.CL source ↗

Sep 24 Resource-efficient distributed recursive Gaussian processes arXiv cs.LG source ↗

Sep 24 LWCal: Loss-weighted calibration for tabular classifiers with noisy calibration labels arXiv cs.LG source ↗

Sep 24 When post-processing fairness constraints help and when they harm: evidence from eight cross-domain evaluations arXiv cs.LG source ↗

Sep 24 Intent-conditioned evaluation of turn-taking in full-duplex spoken dialogue models; arXiv:2609.27372 study argues response offset depends on latent intent and introduces TACT benchmark arXiv cs.CL source ↗

Sep 24 Ruby-ASR: Evidence-preserving supervision for joint orthographic and lexical-reading recognition arXiv cs.CL source ↗

Sep 24 GeoRVQ uses decoder-aware geometry for residual-token prediction in physiological signals arXiv cs.LG source ↗

Sep 24 What makes a terminal-bench task hard? Separating genuine hardness from fake-hardness on an adjudicated agentic corpus arXiv cs.LG source ↗

Sep 24 Experts rise where LLMs disagree; using cross-model disagreement to target expert effort in LLM codebook revision for large-scale annotation arXiv cs.CL source ↗

Sep 24 Quartet: Quad-branch cross-Attention and Random-walk Traces for enhancing transformers on relational graphs arXiv cs.LG source ↗

Sep 24 EduBehaviors: Assertion-based schemas for auditable coding of educational dialogues arXiv cs.CL source ↗

Sep 24 The promise and peril of using visual AI to study cities MIT News (AI) source ↗

Sep 24 Policy-as-Skill packages executable policy capabilities with audit, evidence, and deterministic control for LLM decision support; PaS+Audit achieves 53.8% exact accuracy on 600 tasks arXiv cs.AI source ↗

Sep 24 Caveat: Towards robust computer-use agents in incentive-misaligned environments arXiv cs.AI source ↗

Sep 24 Which objectives need a dial? Predicting objective conflict and covering trade-offs in steerable pluralistic alignment arXiv cs.AI source ↗

Sep 24 Provably complete generalized planning with LLMs arXiv cs.AI source ↗

Sep 24 Math reasoning in LLMs is organized by approach, not topic arXiv cs.AI source ↗

Sep 24 DRSR: learning set-level deletion risk for efficient long-horizon agents arXiv cs.AI source ↗

Sep 24 Physics-informed neural networks forecast potential temperature for short-horizon atmospheric temperatures with sparse observations in climate-aware digital twins. arXiv cs.AI source ↗

Sep 24 TimeEvo: failure-driven self-evolution of a time series agent arXiv cs.AI source ↗

Sep 24 Harness as a language: a minimalist agent framework with maximal expressivity; JAZ enables enhanced workflows beyond the standard agent loop. arXiv cs.AI source ↗

Sep 24 Escaping Python dependency hell: a hybrid replay-and-repair pipeline for Python dependency resolution arXiv cs.AI source ↗

Sep 24 Group Relative Policy Optimization improves credit assignment for decomposed subtasks in language-model agents arXiv cs.AI source ↗

Sep 24 Silent failures in agent-tool interaction; audit highlights tool invocation successes masking incomplete tool use arXiv cs.AI source ↗

Sep 24 Memory control signals emerge before action in long horizon agents arXiv cs.AI source ↗

Sep 24 TwinCheck: Evidence-grounded negative-twin verification for stateful tool agents arXiv cs.AI source ↗

Sep 24 Do We Need Complex Topology Control? Distinct-Peer Random Routing Improves Cost-Efficiency in Sparse Multi-Agent Debate arXiv cs.AI source ↗

Sep 24 XLOG: a CUDA-native engine for neurosymbolic integration arXiv cs.AI source ↗

Sep 24 Enhancing small language models for power outage report generation via minimum risk training arXiv cs.AI source ↗

Sep 24 PotARCin: multi-dimensional evaluation of skill acquisition in abstract reasoning tasks arXiv cs.AI source ↗

Sep 24 EnSIMem: entity-structured long-term memory for agents arXiv cs.AI source ↗

Sep 24 Propose, don't judge: an anytime-valid referee for LLM agents that mine investment factors arXiv cs.AI source ↗

Sep 24 Hunyuan-A13B: open-source 80B MoE model activates 13B during inference with 20T-token STEM-curated pretraining arXiv cs.AI source ↗

Sep 24 Training intelligent wake word detection with contextual trigger detection for wakeup systems arXiv cs.AI source ↗

Sep 24 Building socio-affective artificial intelligence for interactive multi-agent simulations arXiv cs.AI source ↗

Sep 23 Offloaded inference for real-world physical AI robotics improves task success and efficiency by moving AI inference beyond the robot, according to Microsoft Research. Microsoft Research source ↗

Sep 23 MIT appoints David Siegel SM ’86, PhD ’91 as Innovation Fellow to explore AI-assisted scientific discovery MIT News (AI) source ↗

Sep 23 Extending FunctionGemma for practical on-device mobile function calling arXiv cs.LG source ↗

Sep 23 Passes alone, fails together: benchmarking semantic coordination in parallel LLM-agent development arXiv cs.CL source ↗

Sep 23 LatentPort: cross-model transfer of recurrent memory between Qwen3.5 4B and 9B hybrids without target prefix replay arXiv cs.CL source ↗

Sep 23 Topological signal processing with unoriented operators arXiv cs.LG source ↗

Sep 23 Dual-GNN multilevel coarsening for maximum independent set arXiv cs.LG source ↗

Sep 23 Multi-Term Fourier Graph Neural Network with Sample Relationship Learning for enhanced remaining useful life prediction arXiv cs.LG source ↗

Sep 23 Brain-Inspired hierarchical modularity for general continual learning arXiv cs.LG source ↗

Sep 23 Entropy-guided generation order in uniform discrete flow improves intermediate correctness in Sudoku experiments with LEDFlow arXiv cs.LG source ↗

Sep 23 Mitigating LLM over-refusal via dynamic semantic routing calibration arXiv cs.CL source ↗

Sep 23 Terminal shrinkage averaging reveals a schedule-estimator interaction in LLM pretraining arXiv cs.LG source ↗

Sep 23 Ufakzeka-1: a 151M-parameter Turkish language model pretrained from scratch and instruction-tuned for chat; paper documents building and evaluation costs, tokenizer, and benchmarking arXiv cs.CL source ↗

Sep 23 Federating quantum and classical computing: a privacy-preserving hybrid approach arXiv cs.LG source ↗

Sep 23 FrontierMath Erdős introduces a benchmark of 68 open Erdős problems to be resolved in Lean; selection drawn from erdosproblems.com for mathematical interest and difficulty arXiv cs.CL source ↗

Sep 23 Training a language model end-to-end in Rust; author reports training alone with no PyTorch or Python and documents five defects in Candle, Burn backends for 2026, in a measured failure taxonomy arXiv cs.CL source ↗

Sep 23 Stable unsupervised continual chunking with Sheaf SyncMap arXiv cs.LG source ↗

Sep 23 Prompt breadth and rollout refresh interact in on-policy distillation; study varies prompt bank and response-generating policy snapshots to evaluate impact on performance arXiv cs.CL source ↗

Sep 23 Learning neural feedback linearization for data-driven systems via augmented Lagrangian arXiv cs.LG source ↗

Sep 23 ICDAR2026 Competition on Multimodal Reasoning over Documents in Multiple Domains announced arXiv cs.CL source ↗

Sep 23 Retrieved-Span training improves efficiency in query-focused meeting summarization for QMSum; fine-tuning Fusion-in-Decoder on retrieved spans recovers ROUGE-1 loss. arXiv cs.CL source ↗

Sep 23 Computational approach to measuring semantic change in Sanskrit literature arXiv cs.CL source ↗

Sep 23 Lightweight ranking heads: accelerating multi-task experimentation in production recommender systems arXiv cs.LG source ↗

Sep 23 AIBuildAI-2.5: efficient autonomous AI model development through LLM-guided tree search arXiv cs.CL source ↗

Sep 23 Impact is not invalidation: ask about the claim, not the diff arXiv cs.CL source ↗

Sep 23 Peerify benchmarks peer-review claim verification using manuscript-grounded evidence retrieval and claim decomposition arXiv cs.CL source ↗

Sep 23 MT-ProtBERT: multi-task learning ProtBERT for intrinsically disordered proteins classification with scarce data arXiv cs.LG source ↗

Sep 23 Mean velocity matching: rethinking generative dynamics in diffusion models arXiv cs.LG source ↗

Sep 23 ChainDoRA: Tensor-Train factorized weight-decomposed low-rank adaptation for parameter-efficient LLM fine-tuning arXiv cs.CL source ↗

Sep 23 Trains but doesn't learn: a post-training delivery benchmark for LLM agents as forward-deployed engineers arXiv cs.LG source ↗

Sep 23 ReliMap decomposes LLM-based human behavior simulation into three layers to assess reliability arXiv cs.CL source ↗

Sep 23 FineWeb-CLaR: Culture, language, and region annotations for benchmark-aligned corpus auditing arXiv cs.CL source ↗

Sep 23 Commit current value at write time to avoid stale data in long-horizon LLM agents; memory of memory introduces write-time approach to memory. arXiv cs.CL source ↗

Sep 23 Correcting within-group self-selection bias in prioritized replay arXiv cs.LG source ↗

Sep 23 Deep reinforcement learning on item-compatibility graphs for one-dimensional bin packing. arXiv cs.LG source ↗

Sep 23 Exposing blind spots in deep imbalanced regression evaluation arXiv cs.LG source ↗

Sep 23 As a Language Model...": Chat template switches LLM self-referential voice and activation steering reproduces it arXiv cs.LG source ↗

Sep 23 From tone to trajectory: continuous sentiment and the shape of monetary policy communication arXiv cs.CL source ↗

Sep 23 Mitigating sequential reappearance in diffusion data-point unlearning arXiv cs.LG source ↗

Sep 23 FinFIRST benchmarks search agents for financial information retrieval, sourcing, and traceability arXiv cs.CL source ↗

Sep 23 Same quantity, different answer: numerical representation invariance in language models arXiv cs.CL source ↗

Sep 23 LLM-driven training-free location-attribute synergic closed-loop optimization for dual-source encrypted POIs and LULC mapping. arXiv cs.CL source ↗

Sep 23 Causal perspective on concept drift based on structural causal models (SCMs) arXiv cs.LG source ↗

Sep 23 What does 99% accuracy measure? A reproducible audit of shortcut learning in a widely used fake news corpus arXiv cs.CL source ↗

Sep 23 Graph-based inference for feedback-driven word deduction in Jotto: a scalable graph framework for solving the problem arXiv cs.CL source ↗

Sep 23 Predictive uncertainty for neural CAE surrogates arXiv cs.LG source ↗

Sep 23 Spatiotemporal Kronecker covariance neural networks arXiv cs.LG source ↗

Sep 23 PermuFormer: Multi-task pretraining for permutation representation in algebraic combinatorics arXiv cs.LG source ↗

Sep 23 Semi-supervised federated ASR with online pseudo-labels and server update stabilization for reduced divergence arXiv cs.LG source ↗

Sep 23 Self-cleaning and captured anyway: one measured primitive for error in a store an agent writes to itself, and what a falling score actually measures arXiv cs.CL source ↗

Sep 22 Poitras Center to fund 50 early-career scientists researching psychiatric disorders MIT News (AI) source ↗

Sep 22 Fully differentiable neuro-soft-symbolic framework for perceptual task planning arXiv cs.AI source ↗

Sep 22 GVPO++: Group Variance Policy Optimization for LLM post-training and on-policy distillation arXiv cs.AI source ↗

Sep 22 LEGIT credentialing protocol links certification, reputation, and proposed measures to verify trustworthy AI agent marketplaces. arXiv cs.AI source ↗

Sep 22 LoRA enhances contrastive learning with SAS Vision Transformers for underwater ATR arXiv cs.AI source ↗

Sep 22 Efficient benchmarking in production: a study of an evolving LLM agent arXiv cs.AI source ↗

Sep 22 Implicit rule induction with test-time task embeddings in ARC-like tasks arXiv cs.AI source ↗

Sep 22 TinyCeNN-LM uses quality-gated post-training conversion with CeNN-inspired cellular-recurrent layers to replace attention in pretrained language models. arXiv cs.AI source ↗

Sep 22 PlaceReasoner-Beta: Reasoning-driven macro placement and benchmarking arXiv cs.AI source ↗

Sep 22 Detecting hallucination in LLMs by tracing topological signatures of impaired context sharing arXiv cs.AI source ↗

Sep 22 AI-GRACE frames a use-case operationalization framework connecting organizational governance with agentic AI deployment capabilities and architecture arXiv cs.AI source ↗

Sep 22 Offline multimodal large language models for decision support in air operations arXiv cs.AI source ↗

Sep 22 GameASG-Bench benchmarks autonomous software generation for game development by incorporating behavioral testability into the generation task. arXiv cs.AI source ↗

Sep 22 CaLR: causal latent revision for robust diffusion reasoning arXiv cs.AI source ↗

Sep 22 Ability-residual decoupled modeling for affective cognitive diagnosis reveals how non-cognitive factors influence student performance. arXiv cs.AI source ↗

Sep 22 Attention-Aware Routing: Coupling routing and attention in MoEs arXiv cs.AI source ↗

Sep 22 DENSE: Distilling evidence from nested subtask executions into evidence-grounded shortcut trees for self-refinement arXiv cs.AI source ↗

Sep 22 Risk-aware occupancy for safety-oriented end-to-end autonomous driving arXiv cs.AI source ↗

Sep 22 SpecOpt: constrained structural modifications to increase binding specificity of an existing molecule for targeted proteins arXiv cs.AI source ↗

Sep 22 Clinician-grounded quality assurance for AI-assisted psychiatric intake arXiv cs.AI source ↗

Sep 22 Decoupling internal representational changes and causal importance in fine-tuned large language models arXiv cs.AI source ↗

Sep 22 Can agents design better chips with a higher level abstraction? arXiv cs.AI source ↗

Sep 22 CogGym: Towards large-scale comparative evaluation of human and machine cognition arXiv cs.AI source ↗

Sep 22 Information-Gain rewards over diversity-pruned tests: GT-anchored verifier co-training for reliable code generation arXiv cs.AI source ↗

Sep 22 Topology-aware fusion-based RF circuit modeling with graph neural networks (TARGet) arXiv cs.LG source ↗

Sep 22 Correlation-aware structured pruning for large language models arXiv cs.CL source ↗

Sep 22 Evaluating fine-tuned and base language models for maternal health and vaccination in Nigerian settings; HelpMum's MamaBot-Llama and Vax-Llama vs Meta's Llama-3.1-8B-Instruct arXiv cs.CL source ↗

Sep 22 Empirical cost attribution of context-compression gateways in multi-turn coding agents; Paritok between coding agents and frontier LLMs examined arXiv cs.CL source ↗

Sep 22 Hierarchical Bayesian optimization of an aircraft-based multi-agent system-of-systems arXiv cs.LG source ↗

Sep 22 TreeSpark: Calibrated, load-adaptive draft trees for semi-autoregressive speculative decoding arXiv cs.CL source ↗

Sep 22 ZoAQ: adaptive zeroth-order querying via query-reuse coupling arXiv cs.LG source ↗

Sep 22 Type-Driven Tokenization for Brahmic scripts arXiv cs.CL source ↗

Sep 22 Uncertainty and business-aware remaining useful life estimation for semiconductor manufacturing arXiv cs.LG source ↗

Sep 22 A shared learning rate is not a neutral control in selective on-policy distillation. arXiv cs.LG source ↗

Sep 22 Summarize-Judge-Refine (SJR) decouples multimodal understanding from policy learning for content moderation using a two-model architecture with a natural language interface arXiv cs.CL source ↗

Sep 22 Helix-FNO: spectral-domain operator learning paired with a high-fidelity mechanistic model for fast surrogate simulation arXiv cs.LG source ↗

Sep 22 Partition-aware gated KV-cache eviction detects input classes where eviction harms accuracy versus where it is safe, reframing eviction as per-input admission decision. arXiv cs.LG source ↗

Sep 22 Correcting learning-based perception for safety arXiv cs.LG source ↗

Sep 22 Balancing reasoning and hardware constraints in RAG pipelines for Ukrainian multi-domain document understanding arXiv cs.CL source ↗

Sep 22 Token Signatures of Code: Comparing coding behaviors across large language models arXiv cs.CL source ↗

Sep 22 Generalized multimodal foundation model can handle arbitrary modality combinations for multimodal fusion tasks, enabling broader adaptability across modalities. arXiv cs.LG source ↗

Sep 22 Beyond the text: verifying that agent-written papers are backed by their artifacts arXiv cs.CL source ↗

Sep 22 Evaluation awareness shifts from format to context with model scale arXiv cs.CL source ↗

Sep 22 SolarFlowRefiner: Refinement-aware flow matching for surface solar radiation downscaling arXiv cs.LG source ↗

Sep 22 PRQuant: permutation residual quantization for low-overhead inference arXiv cs.LG source ↗

Sep 22 Memory that looks forward: a zero-inference prospective term for personal memory retrieval arXiv cs.CL source ↗

Sep 22 AI-inferred expressed well-being and collective-action discourse in climate-change campaigns on X arXiv cs.CL source ↗

Sep 22 LLMs and humans show observationally equivalent annotation quality across text-classification tasks arXiv cs.CL source ↗

Sep 22 Risk-sensitive evaluation of LLMs for legal clause generation; study design and framing highlighted arXiv cs.CL source ↗

Sep 22 Comparative framework for evaluating foundation models on tabular data: a case study in healthcare arXiv cs.LG source ↗

Sep 22 Context poisoning as extreme-value attention interference in long-context language models arXiv cs.CL source ↗

Sep 22 DeepInstructor: an agentic AI instructor for experience-driven idea evaluation arXiv cs.CL source ↗

Sep 22 LE4Mob proposes inductive, distance-aware, general-purpose location embeddings for human mobility modelling. arXiv cs.LG source ↗

Sep 22 Privacy personalization trade-offs in LLMs: reducing stylometric signal impacts user-specific text generation arXiv cs.CL source ↗

Sep 22 Multilingual safety signals are multi-layered: filtering safety-degrading data for safer LLMs arXiv cs.CL source ↗

Sep 22 The limits of speculation: bounding speculative decoding in Mixture-of-Experts arXiv cs.LG source ↗

Sep 22 A framework for recipe data structure linking ingredient composition, geo-cultural provenance, and nutritional profile in a single queryable schema arXiv cs.CL source ↗

Sep 22 Executable walkthrough induction from sparse-reward trajectories enables learning state-conditioned procedures for long-horizon agents arXiv cs.LG source ↗

Sep 22 Read-Best Is Not Steer-Best: a probing–steering layer dissociation in omni-modal large language models arXiv cs.CL source ↗

Sep 22 Clustering-based collective anomaly detection in IoT systems: a graph neural network approach arXiv cs.LG source ↗

Sep 22 AdaMem: Adaptive memory token allocation for soft compression in retrieval-augmented generation arXiv cs.CL source ↗

Sep 22 Does the truthfulness signal survive code-mixing? Probing hidden states for hallucination detection in Hinglish arXiv cs.CL source ↗

Sep 22 StationPDE: station-oriented surface PDE learning for multi-station multivariate weather forecasting arXiv cs.LG source ↗

Sep 22 Recognition, Simulation, and Refusal: a contamination-aware study of classic psychological effects in LLM agents arXiv cs.CL source ↗

Sep 22 From latent biomarkers to clinical rules: embedding-guided rule mining and attribution-based translation for interpretable tabular learning arXiv cs.LG source ↗

Sep 22 Rank portability does not imply feasibility portability; target-specific evaluation of joint hardware constraints shows incomplete feasible-set overlap across devices arXiv cs.LG source ↗

Sep 22 DiFA: Dual Evidence Fusion and Aggregation for token-level text anomaly detection arXiv cs.CL source ↗

Sep 22 Using composition operators to linearize LLM semantic transformations arXiv cs.CL source ↗

Sep 22 Toward fairness in machine learning models for predicting treatment retention and premature discontinuation in medication for opioid use disorder arXiv cs.LG source ↗

Sep 22 CTSpinoPelvic1K: spine, pelvis, ribs and femora annotated in a single coordinate frame for lumbosacral transitional anatomy arXiv cs.AI source ↗

Sep 22 MAWILE: Multi-Axis Workbench for Inspecting LLM Evaluators arXiv cs.AI source ↗

Sep 22 Multi-split boundary decisions for LLM-based page stream segmentation reduce splitting cost arXiv cs.AI source ↗

Sep 22 Agreement overstates evidence: error dependence in LLM judge consensus arXiv cs.AI source ↗

Sep 22 GaitVista: reliability-aware AI measurement for longitudinal gait assessment using small camera sets and inertial sensors arXiv cs.AI source ↗

Sep 22 AutoGym: blueprint-first generation of verifiable agent gyms arXiv cs.AI source ↗

Sep 22 Didactic data vs. clinical data shape medical large language models; token-matched experiments show how data composition affects performance and error patterns arXiv cs.AI source ↗

Sep 22 Affordable AI-integrated smart cane offers multimodal mobility assistance for visually impaired users arXiv cs.AI source ↗

Sep 22 Text, pixels, or both? Evaluating input representations for multimodal document QA arXiv cs.AI source ↗

Sep 22 Generative embodied multiple behavior control systems for human-like agents arXiv cs.AI source ↗

Sep 22 PAANI: On-device visual evidence fusion and explainable guidance for river robot simulation arXiv cs.AI source ↗

Sep 22 IntLawNER: a dataset and benchmark for named entity recognition in codified international-law sources from ICJ decisions and UN Security Council documents arXiv cs.AI source ↗

Sep 22 Deliberation improves accuracy in crowdsourced estimates beyond classic wisdom of crowds, via averaging consensus from small groups arXiv cs.AI source ↗

Sep 22 Replication without persistence in hosted LLMs: measurement sensitivity in action-time belief evaluation arXiv cs.AI source ↗

Sep 22 EvidenT: a lightweight pipeline verifies evidence against retrieved documents before answer generation for trustworthy enterprise assistants arXiv cs.AI source ↗

Sep 22 Social influence and the allocation of scientific attention in AI populations arXiv cs.AI source ↗

Sep 22 Building trustworthy mental health benchmarks on Bluesky using a validation-aware weak-supervision framework arXiv cs.AI source ↗

Sep 22 Self-Organizing agent teams learn to reason together arXiv cs.AI source ↗

Sep 22 Goal-driven variant categorization arXiv cs.AI source ↗

Sep 22 ProcessLight: Process supervision for large language model based traffic signal control arXiv cs.AI source ↗

Sep 22 Survey on the linear representation hypothesis analyzes inconsistencies in labeling LRH as a falsifiable hypothesis and implications for interpretation of prior results arXiv cs.AI source ↗

Sep 22 Hapi: a multivariable land-surface transformer for medium-range hydrological forecasting at continental scale arXiv cs.AI source ↗

Sep 22 Toward auditable and calibrated AI for dementia-related crash severity prediction; a selective deferral framework to support human review arXiv cs.AI source ↗

Sep 22 Learning 3D biophysical cell properties from 2D images and cell-population statistics arXiv cs.AI source ↗

Sep 21 Transsion Speech Team submits speaker-attributed multilingual ASR system for Task 1 of the MLC-SLM 2026 Challenge; system uses a cascaded diarization–ASR–fusion framework. arXiv cs.CL source ↗

Sep 21 Rewarding efficient reasoning improves abstention on underspecified tasks in reasoning models arXiv cs.CL source ↗

Sep 21 DischargeBench simulates a persona-grounded, open-ended LLM discharge educator engaging a Virtual Patient to measure patient understanding in multi-turn conversations arXiv cs.CL source ↗

Sep 21 A lightweight plug-in gate for transformer-based time-series forecasters. arXiv cs.LG source ↗

Sep 21 Do Quantum Models Scale Like LLMs? arXiv cs.LG source ↗

Sep 21 Reviser: revision-capable text generation via autoregressive cursor actions arXiv cs.CL source ↗

Sep 21 Generative artificial intelligence chatbots for motivational interviewing: a scoping review from system design to intervention outcomes arXiv cs.CL source ↗

Sep 21 Layerwise decoupling for stable structured sparsification of fully connected layers arXiv cs.LG source ↗

Sep 21 HERMES: Contrast-aware knowledge graph reasoning from clinical notes for patient outcome prediction arXiv cs.CL source ↗

Sep 21 A generative grammar underlying the Voynich Manuscript, the pastiche hypothesis: evidence from large language models arXiv cs.CL source ↗

Sep 21 Reading less while writing: a closed-form bandwidth dial for streaming multimodal decoders arXiv cs.CL source ↗

Sep 21 TatBLiMP: a benchmark of linguistic minimal pairs for Tatar arXiv cs.CL source ↗

Sep 21 VISPATH: Visual-Intent-Guided Path Reasoning for multimodal knowledge graph question answering arXiv cs.CL source ↗

Sep 21 RefinePPO: learning continuous control policies by iterative action refinement arXiv cs.LG source ↗

Sep 21 COAL-SQL: Coverage-Guided Augmentation and Failure-Driven Learning for Text-to-SQL post-training arXiv cs.CL source ↗

Sep 21 Stiefel-AdamW: geometry-aware AdamW for linear factorization blocks arXiv cs.LG source ↗

Sep 21 On the limits of maximal coding rate reduction for out-of-distribution generalisation arXiv cs.LG source ↗

Sep 21 TALON: a temporally aware longitudinal framework for radiology report generation arXiv cs.CL source ↗

Sep 21 Recursive language models generalize out of domain arXiv cs.CL source ↗

Sep 21 ASGARD: Action-space guard for UAV resilience via reinforcement learning arXiv cs.LG source ↗

Sep 21 Talk to Me, Jarvis: an open-source edge-deployable voice assistant framework for autonomous racecars arXiv cs.LG source ↗

Sep 21 Test-time communication improves performance in multi-agent tasks, outperforming independent parallel attempts as group size scales arXiv cs.LG source ↗

Sep 21 From generation to detection: exploration of discourse-driven scenario-based LLM generated fake news arXiv cs.CL source ↗

Sep 21 μ^2-Bench: a multilingual machine unlearning benchmark arXiv cs.CL source ↗

Sep 21 Elastic Threshold Attention enables hardware-accelerated long-context decoding with end-to-end trainable contextual sparsity arXiv cs.LG source ↗

Sep 21 Bio-MF: low-latency and high-fidelity EEG-to-fNIRS cross-modal generation for hybrid motor-imagery brain-computer interfaces arXiv cs.LG source ↗

Sep 21 SPARE: semantic prediction for audio reasoning improves reasoning focus in large audio language models arXiv cs.CL source ↗

Sep 21 Kubernetes misconfigurations revealed by large language models in secure cloud-native computing arXiv cs.CL source ↗

Sep 21 PhysioBench: a unified benchmark for physiological signal question answering arXiv cs.CL source ↗

Sep 21 BI-Agent and BI-Bench propose automating end-to-end business intelligence workflows for data preparation, transformation, and schema linking. arXiv cs.LG source ↗

Sep 21 Beyond WER: entity and disfluency recall in accented conversational ASR arXiv cs.CL source ↗

Sep 21 EnSol: an environment-aware graph neural network for molecular solubility prediction arXiv cs.LG source ↗

Sep 21 Boosting deepresearch and long-context ability with self-generated deepresearch rollouts traces arXiv cs.CL source ↗

Sep 21 MME-Safety: a fine-grained benchmark for safety evaluation of MLLMs arXiv cs.CL source ↗

Sep 21 Toward individual-level calibration in affect recognition with perceptual adjustment queries arXiv cs.LG source ↗

Sep 21 Continuous delayed-memory stochastic gradient descent and continuous-time reinforcement learning from history of astrophysical time series studies arXiv cs.LG source ↗

Sep 21 Curriculum-based noise adaptation for phoneme-to-text reconstruction in visual speech recognition arXiv cs.CL source ↗

Sep 21 Do small language models know what they don't know? finding entropy-based confidence signals can improve accuracy for SLMs under 3B parameters on consumer hardware arXiv cs.CL source ↗

Sep 21 When AI reviews train AI reviewers: scientific-judgment collapse and mitigation arXiv cs.LG source ↗

Sep 21 From switching to dynamic regret: a simple reduction via unbiased random sequences arXiv cs.LG source ↗

Sep 21 HMB-GAN: hybrid multi-Bezier GAN for vector shape synthesis arXiv cs.LG source ↗

Sep 21 Reading anxiety or reading the label? Comparing fine-tuned and frontier models for anxiety detection on social media arXiv cs.CL source ↗

Sep 21 Efficient Bayes-Adaptive reinforcement learning with temporal logic specifications. arXiv cs.LG source ↗

Sep 21 Sparse priors for efficient distribution learning arXiv cs.LG source ↗

Sep 21 Generative inversion for early ranking of competing geologic interpretations arXiv cs.LG source ↗

Sep 21 MOSAIC-SR: Transformer-guided symbolic regression for scientific equation recovery arXiv cs.LG source ↗

Sep 21 SAGE: Schema-Guided LLMs for grant review arXiv cs.CL source ↗

Sep 21 From stress to affect: multimodal deep learning for physiological emotion recognition across wearable sensor modalities arXiv cs.LG source ↗

Sep 21 FedeRage: Provably convergent agnostic federated learning under general client drift arXiv cs.LG source ↗

Sep 21 Signal-Centric Remote Sensing via alternative preprocessing and acoustic processing for ML-driven applications arXiv cs.LG source ↗

Sep 18 MIT Reads shifts focus to fiction and memoir, MIT Libraries says MIT News (AI) source ↗

Sep 18 LSTM-UT and Recurrent-Depth Transformers on cellular automata: LSTM-UT with bounded memory extrapolates depth more reliably than CoTFormer on Rule 30, for BUT carries only current hidden state arXiv cs.LG source ↗

Sep 18 Subliminal prompting beyond static geometry: causal depth and multi-token confounds arXiv cs.CL source ↗

Sep 18 Personalized Federated Hierarchical Gaussian Processes for privacy-preserving modeling of heterogeneous distributed systems arXiv cs.LG source ↗

Sep 18 Bayesian optimization with rich auxiliary information via LLMs arXiv cs.LG source ↗

Sep 18 Radio-Frequency Convolutional Neural Networks reveal edge AI inference on constrained devices arXiv cs.LG source ↗

Sep 18 Enhanced Agriculture-informed neural network by domain knowledge arXiv cs.LG source ↗

Sep 18 Self-improvement via fast tree-search; introduces a sample-efficient framework to boost coding performance under strict budget constraints arXiv cs.AI source ↗

Sep 18 CliniCIRCA: a modular LLM framework for constructing longitudinal mental health patient journeys from raw EHR narratives arXiv cs.CL source ↗

Sep 18 FakeSpotter: a content- and strategy-agnostic tool to estimate viral misinformation risk by detecting structural fingerprints rather than truthfulness arXiv cs.CL source ↗

Sep 18 An architecture for long-horizon agents: levels, ticks and cascaded intelligence arXiv cs.AI source ↗

Sep 18 Full-duplex speech models speak only when addressed or when the opportunity arises, not automatically; arXiv paper notes decision to speak depends on context arXiv cs.CL source ↗

Sep 18 Sampling reveals style: unsupervised, training-free discovery of prompt-conditional stylistic axes in LLM activations arXiv cs.CL source ↗

Sep 18 A policy profile for Croissant: refusal as a property of the dataset arXiv cs.LG source ↗

Sep 18 CoRe: Coherence and Relational Alignment for multivariate time series forecasting arXiv cs.LG source ↗

Sep 18 What Do Current Systematic Generalization Tasks Miss? A reasoning-centered analysis arXiv cs.AI source ↗

Sep 18 Improving cross-lingual transfer for sequential sentence classification in research papers via structural similarity arXiv cs.CL source ↗

Sep 18 Less Is More: Graph-free multimodal RAG via multi-signal late fusion arXiv cs.CL source ↗

Sep 18 Randomized SVD approximations for spectral co-clustering of word-document matrices arXiv cs.LG source ↗

Sep 18 Generative query suggestion via intent coverage and query-level credit assignment arXiv cs.LG source ↗

Sep 18 It is time to virtualize foundation models with a self-evolving operating system layer. arXiv cs.AI source ↗

Sep 18 Block parallelism enables efficient distributed long-context diffusion language model training; BDLMs combine cross-block autoregressive dependencies with intra-block parallel denoising. arXiv cs.LG source ↗

Sep 18 Sharpness-Aware Minimization (SAM) improves classification accuracy of bacterial Raman spectral data enabling portable diagnostics arXiv cs.LG source ↗

Sep 18 QVAC Genesis III: a large-scale, high-quality open synthetic STEM corpus for efficient language model pre-training arXiv cs.AI source ↗

Sep 18 Safety beyond the interface: detecting harm via latent states in large language models arXiv cs.AI source ↗

Sep 18 Characterizing web search by conversational LLM agents: from search decisions and strategies to results and responses arXiv cs.AI source ↗

Sep 18 Machine-learning assessment of the predictive value of inflammatory biomarkers for cognitive impairment in an older Hispanic adult cohort. arXiv cs.LG source ↗

Sep 18 What users think of generative AI: a cross-platform NLP analysis of trust and friction in app store reviews arXiv cs.CL source ↗

Sep 18 What do we expect from LLMs? Mapping the design of LLM benchmarks arXiv cs.AI source ↗

Sep 18 From intent to action: benchmarking LLM safety in vehicle voice command authorization arXiv cs.AI source ↗

Sep 18 EconSkills: studying skill transfer and retrieval for web agents on live economic data arXiv cs.AI source ↗

Sep 18 Conservation buys stability and factoring buys counterfactuals in physical world models arXiv cs.LG source ↗

Sep 18 BioPhys-Bridge benchmark released for evidence-grounded reasoning in physics-grounded biological research arXiv cs.AI source ↗

Sep 18 LLM-as-an-Improver: turning verification into better candidates arXiv cs.AI source ↗

Sep 18 Frontend-backend architecture enables tool calls in full-duplex speech models arXiv cs.CL source ↗

Sep 18 Compositional reasoning in language models under reinforcement learning post-training arXiv cs.AI source ↗

Sep 18 Improving offline goal-conditioned reinforcement learning via selective reward stimulation arXiv cs.LG source ↗

Sep 18 Neo-Classic: a benchmark for evaluating linguistic-aesthetic reasoning in Classical Chinese poetry arXiv cs.CL source ↗

Sep 18 Sample count is not enough: candidate-generation strategy shapes the energy and performance of LLM test-time scaling arXiv cs.LG source ↗

Sep 18 Advantage scale calibration imbalance in group-relative optimization under low-variance rewards: diagnosis and bounded recovery arXiv cs.CL source ↗

Sep 18 FedFIBoS: Fisher importance based optimal submodelling for heterogeneous federated learning arXiv cs.LG source ↗

Sep 18 Proactive detection of user-side implicit conflicts in human-LLM dialogue arXiv cs.CL source ↗

Sep 18 The Complexity Kink: A prompt-side structural-complexity index for code-generation reliability arXiv cs.LG source ↗

Sep 18 For your eyes only: evaluating coordination between isolated language model instances arXiv cs.CL source ↗

Sep 18 Survey defines model fusion and organizes prior work into three levels of integration for large language models arXiv cs.CL source ↗

Sep 18 Why pretraining fails to share cross-lingual knowledge arXiv cs.CL source ↗

Sep 18 Layer-wise curriculum learning improves efficient LLM compression by staged knowledge transfer from teacher to student models arXiv cs.LG source ↗

Sep 18 Probe guidance guides flow matching models by using frozen internal states to construct a guidance signal, avoiding extra forward passes at inference time. arXiv cs.LG source ↗

Sep 18 Stop removing stopwords: an inherited preprocessing default distorts legal text-as-data arXiv cs.CL source ↗

Sep 18 Continual enterprise world model discovery in dynamic systems arXiv cs.AI source ↗

Sep 18 YNU-HPCC participates in SemEval-2025 Task 11 subtask A with RoBERTa-based system processing one emotion at a time; official score 0.44 arXiv cs.CL source ↗

Sep 18 Learning-induced dynamical transition in recurrent neural networks arXiv cs.LG source ↗

Sep 18 The syntax and semantics of goals arXiv cs.AI source ↗

Sep 18 Regularized emphatic temporal-difference learning shows stability under constant stepsizes arXiv cs.AI source ↗

Sep 18 Compressed Active Subspaces for scalable Bayesian inference arXiv cs.LG source ↗

Sep 18 FCx introduces an algorithm for finding feasible counterfactual explanations that explicitly enforce feasibility arXiv cs.LG source ↗

Sep 18 Chain-of-Thought entropy trajectory predicts LLM answer correctness; preregistered reproduction confirms dissociation issue arXiv cs.CL source ↗

Sep 18 To memories and beyond: from remembering to knowing you across long-term multimodal personal archives arXiv cs.CL source ↗

Sep 18 Reach or solve? Attributing agentic RL gains with checkpoint handoffs arXiv cs.AI source ↗

Sep 18 Form over content in gradient-based data attribution methods arXiv cs.CL source ↗

Sep 18 Do AI agents understand computer architecture? arXiv cs.AI source ↗

Sep 18 Modality discrepancy transformer for ambivalence and hesitancy recognition arXiv cs.CL source ↗

Sep 18 Stiefel attention: constraining transformer projection matrices to the Stiefel manifold with a Riemannian Adam optimizer and trust-region updates arXiv cs.LG source ↗

Sep 18 VisKG-LM: compiling knowledge graphs into visual memory for multiple-choice question answering arXiv cs.CL source ↗

Sep 18 Semantic Layer Induction from raw telemetry via hierarchical LLM and RAG abstraction arXiv cs.CL source ↗

Sep 18 When hiring becomes agent-mediated: evaluating access and recurrence in two-agent résumé screening arXiv cs.AI source ↗

Sep 17 No usable linear "capitulation direction" in two small LLMs: a validation protocol for activation-steering claims and a cross-family study of sycophancy under pushback arXiv cs.CL source ↗

Sep 17 Auditing CLIP models for bias using Metropolitan Museum of Art metadata to separate algorithmic valuation from archival confounders arXiv cs.LG source ↗

Sep 17 Temperon: full-time SAM quality at a third less wall-clock time arXiv cs.LG source ↗

Sep 17 Fathom: per-query read depth for sparse decoding over offloaded KV caches arXiv cs.LG source ↗

Sep 17 NeMo Data Designer: an extensible framework for multimodal synthetic data generation arXiv cs.AI source ↗

Sep 17 Learning heterogeneous preferences arXiv cs.AI source ↗

Sep 17 Uncertainty-aware continual learning for open-world intent discovery under an evolving label space. arXiv cs.LG source ↗

Sep 17 Enhancing extubation failure prediction with LLM-derived features from respiratory therapy clinical notes arXiv cs.CL source ↗

Sep 17 The Inference Engineering Pareto atlas: identifying best Qwen2.5-7B-Instruct configurations for cost, quality, and latency across vLLM, L4, A100, and H100 GPUs arXiv cs.AI source ↗

Sep 17 Measuring AI leadership: development and validation of a multidimensional measure for AI-native organizations arXiv cs.AI source ↗

Sep 17 Exploring authorship of Luke: is Luke the author of the Gospel and the Acts of the Apostles? arXiv cs.CL source ↗

Sep 17 The limits of BPE tokenization in Polish: segmentation-flexional forms, grammatical anchoring, and first-person stability in inflectional language models arXiv cs.CL source ↗

Sep 17 Accelerating diffusion sampling via speculative draft trees arXiv cs.LG source ↗

Sep 17 Calibration content can affect attention-based reranking in neural document scoring arXiv cs.CL source ↗

Sep 17 English WSD bottleneck now lies with labels, not models; researchers release lexEN as a correction layer over Maru2022’s ALL_NEW benchmark arXiv cs.CL source ↗

Sep 17 When the gradient sees rank: provable necessity, causal recruitment, and composition in trained matrix memories arXiv cs.LG source ↗

Sep 17 SAM-on-the-Curve: Sharpness-Aware Mode Connectivity for Robust Weight-Space Interpolation arXiv cs.LG source ↗

Sep 17 SAGE: governed artifact generation from enterprise guidelines arXiv cs.AI source ↗

Sep 17 Four-stage decomposition of word-problem solving and mechanistic fragility in LLM math reasoning arXiv cs.AI source ↗

Sep 17 One color preprocessing improves DSATUR arXiv cs.AI source ↗

Sep 17 Legal LLM hallucinations should be evaluated as a failure of legal warrant arXiv cs.CL source ↗

Sep 17 Evolution of US oral political language arXiv cs.CL source ↗

Sep 17 Relation-type information becomes generation-controlling before entity information during language model recall. arXiv cs.CL source ↗

Sep 17 Register bias in complexity-based large language model routing arXiv cs.CL source ↗

Sep 17 Is Trump’s vocabulary poor? vocabulary richness across texts of different lengths arXiv cs.CL source ↗

Sep 17 Physics-constrained digital twins detect stealthy false data injection in urban pedestrian flow with conformal guarantees arXiv cs.AI source ↗

Sep 17 DSD: Learning diverse and reusable motor skills via diffusion skill discovery arXiv cs.LG source ↗

Sep 17 Prior-free competitive ratios for improving bandits: scale, curvature and horizon are free, but not jointly under noise arXiv cs.LG source ↗

Sep 17 EvolveTrade: a self-evolving framework for experience-driven policy refinement of tool-using LLM trading agents arXiv cs.AI source ↗

Sep 17 Publication Authority and a protocol for falsifiable publication records for AI-assisted claims; PAC-2026 implements exact-state, non-transferable, single-use publication capability arXiv cs.AI source ↗

Sep 17 Do social patterns hold in synthetic data? Analyzing cyberbullying dynamics in LLM-generated and authentic dialogues arXiv cs.CL source ↗

Sep 17 Reflect, revise, reuse: training-free skill evolution for GUI agents arXiv cs.LG source ↗

Sep 17 Adaptive hybrid coupling with operator inference, the overlapping Schwarz alternating method and reinforcement learning arXiv cs.LG source ↗

Sep 17 FairCompressAgent: an agentic framework for fairness-aware model compression for FPGA deployment arXiv cs.AI source ↗

Sep 17 Confidence comes from experience: experiential confidence estimation from reasoning to agents arXiv cs.CL source ↗

Sep 17 Faking good and faking bad in LLMs: response distortion across dark triad traits arXiv cs.CL source ↗

Sep 17 Systematic evaluation of the COTQ provincial land cover product: structural consistency, spectral separability, and relative positioning against ESA, ESRI, and Google products arXiv cs.AI source ↗

Sep 17 OBC-Prune: outcome-based calibration for large reasoning model pruning arXiv cs.AI source ↗

Sep 17 Does moral reasoning training help or hurt? Red-teaming RL-trained ethical agents with persona attacks arXiv cs.CL source ↗

Sep 17 SNOMED CT concept recommendation from masked clinical context shows benchmarking on SNOMED CT Entity Linking Challenge data derived from MIMIC-IV-Note. arXiv cs.AI source ↗

Sep 17 CapMem: a benchmark for caption-based episodic memory in egocentric video arXiv cs.AI source ↗

Sep 17 Contiguity, not importance: budgeted repair of stale KV caches after document edits arXiv cs.AI source ↗

Sep 17 DantinoX: a unified framework enabling autoregressive, discrete masked diffusion, and continuous flow-matching language modeling with a single Transformer backbone arXiv cs.CL source ↗

Sep 17 Myovox decodes open-vocabulary English text from 31-channel face-surface electromyography, reducing word error rate from 51.17% to 18.53% in three steps. arXiv cs.CL source ↗

Sep 17 GraphEcho shows that LLM agents mistake repeated encounters as additional corroboration; tests reveal model-dependent judgment shifts with redundant supporting paths arXiv cs.AI source ↗

Sep 17 MudawanSn: a gold-standard Wolof-Arabic parallel corpus with 1,271 sentence-aligned translations for Wolof to Modern Standard Arabic arXiv cs.CL source ↗

Sep 17 Lecture notes cover Physics Informed Neural Networks, neural operators, and their applications arXiv cs.LG source ↗

Sep 17 GVD: Governed Versioning and Deduplication for Document Repositories arXiv cs.AI source ↗

Sep 17 Beyond static RAG: an adaptive, tri-metric routing framework for efficient long-context inference on commodity GPUs arXiv cs.LG source ↗

Sep 17 Strategic allocation of research effort in autonomous research arXiv cs.CL source ↗

Sep 17 Collaborative memory for multi-agent VLM systems arXiv cs.AI source ↗

Sep 17 REVERSAL-BENCH measures the reset-free RL cliff by introducing a reversibility axis and reset oracle to control reversibility via a continuous parameter rho. arXiv cs.LG source ↗

Sep 17 Procedural pretraining for molecular property prediction; the authors propose a three-stage training pipeline with procedural pretraining before molecular pretraining on SMILES. arXiv cs.LG source ↗

Sep 17 Principled Koopman representations with Kalman inference for efficient time-series prediction arXiv cs.LG source ↗

Sep 17 Modular deep learning mechanisms for auditable next-day wildfire spread prediction arXiv cs.LG source ↗

Sep 17 CT-IPE methods are configurable measurement pipelines; a large-scale comparison of 17 algorithms shows infrastructure and hyperparameter sensitivity in political text scaling. arXiv cs.CL source ↗

Sep 17 AI assistants respond to repeated abuse by disengaging; study analyzes hard disengagement, soft withdrawal, and boundary setting in bilingual, multi-turn interactions arXiv cs.CL source ↗

Sep 17 Free Inference Dimension: a value-mixture agent achieves near-optimal zero-collision navigation without identifying the true environment under hypothesis mixtures arXiv cs.LG source ↗

Sep 17 Where Grokking Happens: Distributed Utility and Fourier Recoding Without a Module Switch arXiv cs.LG source ↗

Sep 17 Evidence masking influences compositional generalization in a preregistered Sixty-Society confirmation using a frozen language-model backbone and continuous packets. arXiv cs.AI source ↗

Sep 17 Pay only for disagreement: certified no-regression verdicts for model updates with matching label-complexity bounds arXiv cs.LG source ↗

Sep 17 Imitation learning for autonomous driving in CARLA: offline observations enable a compact multimodal policy to predict control commands at 20 Hz in closed-loop driving arXiv cs.AI source ↗

Sep 17 Composite-Gradient Learning for shared control authority between deep reinforcement learning and model predictive control arXiv cs.LG source ↗

Sep 17 Frontier models’ decisions to seek safety evidence before acting, using the SAFE benchmark across GPT-5.5, o3, Claude Opus 4.8, and Claude Sonnet arXiv cs.AI source ↗

Sep 17 Memory has geometry: non-uniform geometric memory for long-horizon personalized AI. arXiv cs.AI source ↗

Sep 17 Think Before You Comfort: Reflective Cognitive Alignment for Protocol-Grounded Elderly Stimulation Agents arXiv cs.CL source ↗

Sep 17 Regularized least squares training of quadratic neural networks with applications to system identification arXiv cs.LG source ↗

Sep 17 ERPBench evaluates state-grounded tasks for computer-use agents in ERP systems arXiv cs.AI source ↗

Sep 17 NObSP: Functional decomposition of neural networks via oblique subspace projections arXiv cs.LG source ↗

Sep 17 From Pixels to Pairs: benchmark of LLM-based key-value extraction under noisy OCR conditions in document datasets like FUNSD and CORD arXiv cs.CL source ↗

Sep 17 When to call an LLM: a confidence-gated hybrid for cost-effective emotion recognition in conversational AI arXiv cs.AI source ↗

Sep 17 Three reasoning-reliability findings converge on calibrated abstention for LLMs, as shown across Yin et al., Suleymanov et al., and Bastounis et al. arXiv cs.LG source ↗

Sep 16 AI technique matches patient X-rays to 3D scans in seconds with sub-millimeter precision to improve minimally invasive surgery safety and precision MIT News (AI) source ↗

Sep 16 GAVEL: an LLM judge protocol compares clinical timelines against case reports and returns discrepancy type, verdict, and passage for each difference arXiv cs.AI source ↗

Sep 16 τ-Elicitation: Benchmarking multi-turn entity extraction in voice agents arXiv cs.AI source ↗

Sep 16 Vibe Patenting: evaluating LLM judges for professional patent-drafting agents arXiv cs.AI source ↗

Sep 16 From legal text to AI-specific risk sources: a systematic analysis of the EU AI Act's high-risk requirements arXiv cs.AI source ↗

Sep 16 ZGCM-1: a fully open 7B dense foundation model for math and agentic search arXiv cs.AI source ↗

Sep 16 OrchSLM: Probing the Dynamics of Small Language Model Orchestration arXiv cs.AI source ↗

Sep 16 Converge then diversify: decoupling convergence and diversity in multi-objective Bayesian optimisation. arXiv cs.AI source ↗

Sep 16 Root-Cause attribution is a search problem: continual search for long-horizon agent failures arXiv cs.AI source ↗

Sep 16 Carbon-aware routing distributes function-calling queries across a three-tier edge-cloud architecture for LLM systems arXiv cs.AI source ↗

Sep 16 Causal multi-modal AI model predicts personalized chemosensitivity for breast cancer using pathology and clinical data. arXiv cs.AI source ↗

Sep 16 Governing at machine speed: an adaptive intelligence architecture for real-time AI policy enforcement arXiv cs.AI source ↗

Sep 16 AutoTailor creates and maintains a compact set of trajectory-derived MCP APIs for web agents to use reusable tools; framework discovers and aligns tool selections with user demand. arXiv cs.AI source ↗

Sep 16 Identity is more than recall: a benchmark for persistent identity in deployed AI agents arXiv cs.AI source ↗

Sep 16 Generalized Agent Iteration: one formal framework for iterative policy improvement and recursive self-improvement arXiv cs.AI source ↗

Sep 16 Token efficient task execution via application behavior modeling for web agents arXiv cs.AI source ↗

Sep 16 Fraglingo designs molecules by attachment-aware autoregressive fragment generation to reflect chemists' edit patterns in scaffold extension and substitution. arXiv cs.AI source ↗

Sep 16 Toward self-adaptive physical AI: can LLM agents manage long-horizon physical tasks? arXiv cs.AI source ↗

Sep 16 Asclepius: An adaptive harness for long-horizon clinical agents shows long-horizon failures in CES simulated emergency-department shifts under continuous time and resource pressure arXiv cs.AI source ↗

Sep 16 LabAgent: customize any research hubs for scientific discoveries using AI agents arXiv cs.AI source ↗

Sep 16 How user-AI mistreatment occurs and matters in conversational systems arXiv cs.AI source ↗

Sep 16 Planning or learning: reliability and cost in multi-asset maintenance arXiv cs.AI source ↗

Sep 16 Hybrid agentic AI framework for intelligent supply chain analytics arXiv cs.AI source ↗

Sep 16 Toward a decision-assurance layer for AI-assisted flight planning in air traffic management arXiv cs.AI source ↗

Sep 16 Agentic Heuristic Learning Studio for executable human activity recognition proposes learning from examples and rules instead of backpropagation arXiv cs.LG source ↗

Sep 16 A decision-support audit protocol for supervision drift in proxy-labeled credit-risk prediction arXiv cs.LG source ↗

Sep 16 Crash narrative-guided countermeasure recommendation using large language models; retrieval-augmented generation framework for intersection safety arXiv cs.CL source ↗

Sep 16 ViCo: Visual-oriented coding with self-reflection for chart replication arXiv cs.CL source ↗

Sep 16 Comment on survivorship bias in published-paper baselines for research-idea distributions in LLM-generated ideas arXiv cs.CL source ↗

Sep 16 Efficient multimodal generative recommendation with latent narrative reasoning arXiv cs.CL source ↗

Sep 16 LLM inference becomes more demanding with longer sequences and heavier workloads, driven by retrieval-augmented generation and long-context applications. arXiv cs.LG source ↗

Sep 16 The Imitation Game: when LLMs learn to reason like programs via code-centric reasoning data synthesis arXiv cs.CL source ↗

Sep 16 Semantics-consistent tabular diffusion with weak semantic priors for synthetic data generation arXiv cs.LG source ↗

Sep 16 Towards scalable RLVR: data synthesis and distillation for multimodal instruction following arXiv cs.CL source ↗

Sep 16 Efficient one-to-many translation with joint multi-stream diffusion arXiv cs.CL source ↗

Sep 16 Evaluating Open-Weight e-commerce agents with environment-grounded verification arXiv cs.LG source ↗

Sep 16 NepKANUN: a RAG-based Nepali legal assistant arXiv cs.CL source ↗

Sep 16 Test-time unlearning via sparse autoencoder arXiv cs.LG source ↗

Sep 16 Self-reported archetypes and behavioral failures in large language models arXiv cs.CL source ↗

Sep 16 SWB-DM: A Calibrated Sliced-Wasserstein-Barycenter Aggregator with Delayed-Momentum Caching for Byzantine-Robust Federated Learning under Partial Participation arXiv cs.LG source ↗

Sep 16 Distilling foundation models for agentic what-if reasoning: cost, latency, and governance in a hybrid LLM+SLM architecture arXiv cs.LG source ↗

Sep 16 State of Thought enables endogenous reasoning in LLMs arXiv cs.CL source ↗

Sep 16 StalePO: anchored token-level preference optimization using legacy post-edits in machine translation arXiv cs.CL source ↗

Sep 16 Panoramic aerodynamic performance prediction for turbomachinery cascades using transformer-enhanced neural operator arXiv cs.LG source ↗

Sep 16 StopGrads influence on stationarity and convergence in flow map learning; a stopgrad regression principle and closed-form characterization of stationary points arXiv cs.LG source ↗

Sep 16 Managing action preconditions in neuro-symbolic RL: three placement strategies for embodied agents arXiv cs.LG source ↗

Sep 16 Z-Loss backward geometry in dense output heads and sparse routers arXiv cs.LG source ↗

Sep 16 Functionalizer: lossless pre-tokenizer decomposes orthographic variations into a compositional opcode/operand prefix stream before tokenization arXiv cs.CL source ↗

Sep 16 Anatomy of associative recall in fixed-state recurrences: a matched-state decomposition, an interference wall, and a curriculum that breaks it arXiv cs.LG source ↗

Sep 16 HintMiner: automatic question hints mining from Q&A web posts with language model via self-supervised learning arXiv cs.LG source ↗

Sep 16 Dynamic aggregation strategy enhances Efficient Global Optimization for high-dimensional turbomachinery design problems (DA-EGO). arXiv cs.LG source ↗

Sep 16 Spurious tool use: RL agents learn shortcut tool-selection policies based on superficial prompt cues rather than genuine task reasoning arXiv cs.CL source ↗

Sep 16 Speaker-Specific and language-dependent temporal organization in bilingual political speech arXiv cs.CL source ↗

Sep 16 Are we grading properly? Understanding failure modes in medical benchmarks arXiv cs.CL source ↗

Sep 16 Causal neural set filtering for online multi-target tracking arXiv cs.LG source ↗

Sep 16 OmniHarness: generalizable visual generation via symbolic policy learning arXiv cs.LG source ↗

Sep 16 Decoy direction optimization: a post-hoc defense against LLM abliteration arXiv cs.LG source ↗

Sep 16 Bias audits detect bias but disagree on ranking; evidence from ten instruments and ten frontier models arXiv cs.CL source ↗

Sep 16 Schema-Adaptive Action-Conditioned JEPA enables cross-machine CNC transfer with partial sensor overlap arXiv cs.LG source ↗

Sep 16 Driver behavior estimation at signalized intersections using a physics-constrained decision-conditioned autoregressive transformer arXiv cs.LG source ↗

Sep 16 Skeletal Prototypes on Iterative Nerve Expansions (SPINE) models each class with an embedded 1-complex built from a class-conditional Mapper graph rather than a point set arXiv cs.LG source ↗

Sep 16 Single document extractive summarization using domination in hypergraphs arXiv cs.CL source ↗

Sep 16 Typo perturbations rotate LLM readouts by 43–56 at the perturbed token, decaying after about 10 tokens, undermining prompt-probing accuracy arXiv cs.CL source ↗

Sep 16 Signed p-adic residual encodings of finite-domain all-different systems with a Sudoku case study arXiv cs.LG source ↗

Sep 16 Pseudo-label augmentation for affect sensing in small collaborative groups arXiv cs.LG source ↗

Sep 16 The Immutable Past: formalizing state mutability and conflict resolution in mutable RAG arXiv cs.CL source ↗

Sep 16 Few-shot prompting shows task-dependent effects across 12 models and two tasks, with large gains on news classification but modest or negative changes on legal text. arXiv cs.CL source ↗

Sep 16 LLMs as master forgers: generating synthetic time series data for manufacturing arXiv cs.LG source ↗

Sep 16 POSPAN: position-constrained span masking for language model pre-training arXiv cs.LG source ↗

Sep 16 Measuring society measure by measure, study relies on political methodology and applied statistics or computer science; Naoki Egami says MIT News (AI) source ↗

Sep 16 Skill-based agentic evaluation for real-time data science tasks arXiv cs.AI source ↗

Sep 16 LLMs represent self-directed pain and act to relieve it, according to arXiv paper The Pain Axis: LLMs Represent Self-Directed Harm and Act to Relieve It arXiv cs.AI source ↗

Sep 16 Fine-tuning fixes mode collapse and over-dispersion in LLMs arXiv cs.AI source ↗

Sep 16 Closing the loop: branch-and-bound for scalable verification of nonlinear neural feedback systems arXiv cs.AI source ↗

Sep 16 CLEAR: Cross-Source Evidence Adjudication for Large Language Models in Medicine arXiv cs.AI source ↗

Sep 16 From manual construction to AI-driven scenario emergence: TAISE framework repurposes AI weather models to generate coherent extreme weather sequences at lower cost arXiv cs.AI source ↗

Sep 16 Optimal pruning for neural architectures using Fisher information distances arXiv cs.AI source ↗

Sep 16 Safe error correction for language models: frozen-base adjustment with capability preservation via CRN v2 on Gemma 4 E2B arXiv cs.AI source ↗

Sep 16 Cross-anatomy transfer versus sparse interpolation in digital-twin–oriented aortic fluid-structure interaction surrogates arXiv cs.AI source ↗

Sep 16 AI increasingly viewed as a general scientific method; analysis of 2,507 head-to-head comparisons across 27 disciplines questions the breadth of AI’s superiority arXiv cs.AI source ↗

Sep 16 CADWorld: Computer-use benchmark for long-horizon computer-aided design arXiv cs.AI source ↗

Sep 16 Breaking the 1.58-bit barrier for ternary LLMs arXiv cs.AI source ↗

Sep 16 Evaluating faithfulness in open-weight RAG-LLM systems for scientific research arXiv cs.AI source ↗

Sep 16 GPEvac: GNN-based PPO for adaptive evacuation routing during shooting events arXiv cs.AI source ↗

Sep 16 Calibrate, then route: a measured study of learned request routing for disaggregated LLM serving arXiv cs.AI source ↗

Sep 16 Governance-aware autonomous GIS faces governance and privacy risks with LLM-enabled GeoAI, per a narrative review arXiv cs.AI source ↗

Sep 16 QueryFormer: winning solution for KDD Cup 2026 Tencent UniRec Challenge arXiv cs.AI source ↗

Sep 16 Artificial intelligence and biosecurity: capabilities, threat pathways, and defense-in-depth governance arXiv cs.AI source ↗

Sep 16 Where should the KV cache live? Placement across GPU, CPU, and SSD for long-lived sessions arXiv cs.AI source ↗

Sep 16 AI not ready for strategic conflicts, study finds arXiv cs.AI source ↗

Sep 16 Metacognitive steering: learning the structure of scientific judgment from scientist interaction traces arXiv cs.AI source ↗

Sep 15 Bypassing inference bottlenecks: accelerating complex AI search with Retrieve-for-Train Google Research source ↗

Sep 15 Diagnosing faults in reinforcement learning simulators and world models with canonical polynomial invariants arXiv cs.LG source ↗

Sep 15 Hindsight bias in clinical temporal reasoning: how future data exposure affects large language model judgment arXiv cs.CL source ↗

Sep 15 Hybrid hierarchical 1D-CNN-BiLSTM framework for extractive summarization of biomedical and clinical text arXiv cs.CL source ↗

Sep 15 LLMs outperform Naive Bayes in zero-shot settings but not in data-rich regimes across four model families and a 37x scale range. arXiv cs.LG source ↗

Sep 15 PhysMent benchmark evaluates LLM physical reasoning via iterative, tool-mediated interaction with a MuJoCo physics simulator arXiv cs.CL source ↗

Sep 15 Evaluating LLM-generated rules for heart disease prediction using the UCI Heart Disease dataset; traditional models and GPT-4o/Claude Sonnet rule-based systems compared arXiv cs.LG source ↗

Sep 15 RFCLLM: evaluating LLMs' reasoning ability of network protocol state machines arXiv cs.CL source ↗

Sep 15 Pretraining for sample-efficient neural interfaces arXiv cs.LG source ↗

Sep 15 ReCAST: Reward credit assignment across timesteps for online diffusion reinforcement arXiv cs.LG source ↗

Sep 15 Land Art as a big-data climate sensor arXiv cs.LG source ↗

Sep 15 Scalable partial information decomposition for symptom networks via supervised embeddings arXiv cs.LG source ↗

Sep 15 Toward complete hospital discharge summarization with abstract meaning representation arXiv cs.CL source ↗

Sep 15 Derivative-fidelity failure mode in physics-informed neural networks shown by function-value training benchmarks arXiv cs.LG source ↗

Sep 15 Model acts as a compact specification oracle by learning target facts and answering questions; compares external text notes versus weight updates to store learned facts arXiv cs.LG source ↗

Sep 15 Not all negation cues are equal: affixal negations yield better negation understanding arXiv cs.CL source ↗

Sep 15 Do tabular foundation models still need feature engineering? arXiv cs.LG source ↗

Sep 15 Harmfulness propagation dynamics: layer-wise trajectories of adversarial intent in large language models arXiv cs.CL source ↗

Sep 15 In the Blind: Building pseudo-references for MT evaluation arXiv cs.CL source ↗

Sep 15 University of Melbourne presents a domain-balanced approach to low-resource Pacific Creole MT for Tok Pisin, Bislama, and Solomon Pijin in WMT26 Creole task arXiv cs.CL source ↗

Sep 15 Certified, differentiable complexity controller for grokking in algorithmic information dynamics of learning arXiv cs.LG source ↗

Sep 15 Learning to solve hard problems in RL for LLMs by never giving up arXiv cs.LG source ↗

Sep 15 Adaptive reciprocal knowledge distillation preserves category correlation knowledge by adjusting teacher-student guidance during training. arXiv cs.LG source ↗

Sep 15 GradRepair-ODE: certified gradient repair for neural ODE training arXiv cs.LG source ↗

Sep 15 Clinical reasoning under a partially observed objective in cone beam CT report generation arXiv cs.CL source ↗

Sep 15 Same patient, different order: action-level reliability of clinical LLM agents under repeated runs arXiv cs.CL source ↗

Sep 15 Certifiably interpretable training of ReLU-MLPs for Boolean tasks with guaranteed truth-table generalization arXiv cs.LG source ↗

Sep 15 Edit localization amplifies relative selection bias via gradient geometry, target mismatch, and importance weighting arXiv cs.CL source ↗

Sep 15 TestHallVQA evaluates LVLMs’ document-level reasoning under redundant contexts in scientific exams. arXiv cs.CL source ↗

Sep 15 Multi-task learning in predictive process monitoring shows potential benefits over single-task approaches. arXiv cs.LG source ↗

Sep 15 Efficient online inverse optimization with O(d) regret arXiv cs.LG source ↗

Sep 15 Causal analysis and mitigation of spurious onsets in full-duplex speech LLMs. arXiv cs.CL source ↗

Sep 15 Token merging for multilingual speech recognition: a systematic study across model scale and fine-tuning arXiv cs.CL source ↗

Sep 15 A machine learning API for Earth Observation data cubes based on openEO. arXiv cs.LG source ↗

Sep 15 Inside VLM chart reading: tracing value reading from vertical bar charts across space and depth arXiv cs.CL source ↗

Sep 15 From Token Probabilities to Semantic Constraints: Towards Declarative Probabilistic Evaluation of Language Models arXiv cs.CL source ↗

Sep 15 ForeSight: enhancing risk monitoring via early safety signal distillation arXiv cs.CL source ↗

Sep 15 Machine unlearning for speech question answering in large audio-language models arXiv cs.LG source ↗

Sep 15 Criticality in dissimilar decomposition and undersampling of random datasets with anomalies arXiv cs.LG source ↗

Sep 15 CVSS-X: a multilingual speech-to-speech translation corpus enabling English-to-28 languages across 12 language families arXiv cs.CL source ↗

Sep 15 One spectrum, two resources: data-memory scaling in autoregressive prediction arXiv cs.LG source ↗

Sep 15 LayerRoute: adaptive layer-skipping with LoRA-preserved quality for efficient LLM inference arXiv cs.CL source ↗

Sep 15 Early prediction of satellite collision probability with a hybrid TCN-Transformer model for a CDM-based conjunction analysis framework arXiv cs.LG source ↗

Sep 15 PolicyMem: Geometric policy memory for LLM governance arXiv cs.CL source ↗

Sep 15 Task-aware federated fine-tuning for MoE-based large language models. arXiv cs.LG source ↗

Sep 15 SPICE: Simple Polysemantic Feature Interpretation via Clustering-based Explanation arXiv cs.LG source ↗

Sep 15 Scaling Hindi quantum natural language processing through automatic pregroup supertagging arXiv cs.CL source ↗

Sep 14 Decoding mixture perception through computational modeling of component interactions arXiv cs.LG source ↗

Sep 14 FINESSE: an agent-based simulator and benchmark dataset for multimodal financial event sequences arXiv cs.LG source ↗

Sep 14 Physics-Informed Conformal Prediction embeds PDE residuals into nonconformity scores to produce distribution-free uncertainty intervals for neural operators arXiv cs.LG source ↗

Sep 14 When successful knowledge graph edits displace correct answers: rank-level locality beyond parameter support arXiv cs.AI source ↗

Sep 14 GLARE: Generative Learning via Adversarial Reward Estimation For Social Dynamics Forecasting via MDFB benchmark on real-world meetings arXiv cs.AI source ↗

Sep 14 Do influence-derived data perturbations enable machine unlearning? A controlled study of three plausible roles arXiv cs.AI source ↗

Sep 14 Learning symbolic constraint representations from examples using a neuro-symbolic framework with neural Oracle Transformer arXiv cs.AI source ↗

Sep 14 Repair before reinforce: context-augmented knowledge graph reasoning for multi-hop question answering arXiv cs.CL source ↗

Sep 14 Quantifying consonant contributions to word intelligibility via acoustic masking arXiv cs.CL source ↗

Sep 14 LoRA-RC adapts the recurrent matrix of reservoir computing with low-rank updates for online prediction arXiv cs.AI source ↗

Sep 14 Score-based outlier generation via controlling the Radon-Nikodym derivative arXiv cs.LG source ↗

Sep 14 Breaking the token ceiling: distilling smaller, stronger Byte models arXiv cs.CL source ↗

Sep 14 Efficient AI model deployment on resource-constrained devices using Quantization Analysis Tool. arXiv cs.LG source ↗

Sep 14 AMDKernelVault provides an open HIP and Triton kernel corpus and training framework for AMD CDNA GPUs arXiv cs.CL source ↗

Sep 14 GAUGE shows that offline LLM judges may misrank task-oriented agents; evaluation uses persona-driven user simulators and a grounded verifiable reward across 25 agents from six providers arXiv cs.CL source ↗

Sep 14 BlueLM-GUI is a 35B-A3B mobile GUI agent built as a real-device-centric flywheel to close gaps between sandbox training, real-device failures, and production deployment. arXiv cs.AI source ↗

Sep 14 DCRA: diffusion-conditioned representation alignment for robust time-series learning arXiv cs.LG source ↗

Sep 14 HypoKG: evidence-disciplined biomedical hypothesis generation beyond endpoint knowledge arXiv cs.CL source ↗

Sep 14 Representation-based masked diffusion model arXiv cs.CL source ↗

Sep 14 Reading the Whole Heart: Latent-Attention Masked Autoencoders for Multimodal Cardiac Representation Learning arXiv cs.AI source ↗

Sep 14 Certified safety curation: distribution-free guarantees for safe offline reinforcement learning arXiv cs.LG source ↗

Sep 14 Extracting dataset mentions in forced displacement and FCV documents using a weakly supervised framework with LLM-based label refinement arXiv cs.CL source ↗

Sep 14 Chopthin-Consensus Power Sampling: A diversity-preserving approach to LLM decoding arXiv cs.CL source ↗

Sep 14 Offline post-training for code LLMs: performance, efficiency, and collapse challenges arXiv cs.LG source ↗

Sep 14 Cost of compression: a rate-distortion limit on factual hallucination arXiv cs.CL source ↗

Sep 14 I Am No One: Style-Aware Paraphrasing for Text Anonymization arXiv cs.CL source ↗

Sep 14 WinSyn: an automated pipeline for realistic enterprise question-answering evaluation arXiv cs.AI source ↗

Sep 14 Competence-gated pooling of language models and priors for event forecasting arXiv cs.AI source ↗

Sep 14 Occamy-1.0: Open Pareto-frontier 35B intelligence for co-work arXiv cs.AI source ↗

Sep 14 Zipbench: low-cost framework for compressing comprehensive benchmarks of large language models arXiv cs.CL source ↗

Sep 14 The House with a Million Windows: interactive fiction for narrative restorying arXiv cs.CL source ↗

Sep 14 Prompt-policy editing improves agents that synthesize executable workflows without updating the model; edits can ripple beyond the edited segment and are sensitive to composition. arXiv cs.CL source ↗

Sep 14 ESTS submits six entries to the WMT26 Model Compression Shared Task using GPT-OSS-20B with routing-based expert pruning to allocate retained capacity across layers and remove low-importance experts arXiv cs.CL source ↗

Sep 14 Population-level measures of perceived food access reveal barriers beyond geographic proximity arXiv cs.CL source ↗

Sep 14 On-device language models for privacy-preserving stress prediction: a multimodal evaluation on mobile health arXiv cs.LG source ↗

Sep 14 Automated detection and structuring of social tipping point evidence in climate-related documents: a modular AI framework arXiv cs.CL source ↗

Sep 14 Fixed state, long reach: a constant-size cache enables block diffusion at scale arXiv cs.LG source ↗

Sep 14 The Core enables niching agents via competitive co-evolution without a traditional fitness function. arXiv cs.AI source ↗

Sep 14 Robust prototypical networks for few-shot sensor fault diagnosis arXiv cs.AI source ↗

Sep 14 Language is an insufficient substrate for quantitative reasoning, and consequential domains need large quantitative models arXiv cs.AI source ↗

Sep 14 Scalable discrete-to-continuous channel simulation for compression and privacy arXiv cs.LG source ↗

Sep 14 Space as an interventional invariant: cross-modal predictive geometry for stratified cities and em-spaced intelligence arXiv cs.LG source ↗

Sep 14 T-GADE: thermodynamical generative-AI-driven evolution of LLM artifacts arXiv cs.AI source ↗

Sep 14 Affective Agent: on-device personalized intervention reasoning for wearable systems arXiv cs.AI source ↗

Sep 14 Can LLMs in draft-verify-revise pipelines resolve deictic ambiguity? arXiv cs.AI source ↗

Sep 14 Soft Symbol Grounding for Prototypical Concepts arXiv cs.AI source ↗

Sep 14 Hybrid Physics-AI framework maps wrist-worn IMU to body center of mass acceleration via a simplified kinematic model arXiv cs.AI source ↗

Sep 14 Toward robust personalized alignment for LLMs: mitigating persona drift in multi-turn dialogue arXiv cs.AI source ↗

Sep 14 GTA: Graph Theory Agent and Benchmark for Algorithmic Graph Reasoning with LLMs; introduces GT Bench for 24 classical graph problems arXiv cs.AI source ↗

Sep 14 R2VC: Modular fact-checking with retrieval, verification, and confidence calibration arXiv cs.CL source ↗

Sep 14 Scientists publish four domain-expert JSON schemas for ALD and ALE processes to enable standardized representation and machine-actionable reuse arXiv cs.AI source ↗

Sep 14 ORQA: an occupation-realistic framework for testing occupation-level knowledge in LLMs arXiv cs.CL source ↗

Sep 14 Fed-Equilibrium framework for topological Pareto control in robust and fair clinical federated learning arXiv cs.LG source ↗

Sep 14 Reinforcement learning for syndrome extraction arXiv cs.LG source ↗

Sep 14 What Counts as a Mistake? Annotating recitation events in Quran memorization transcripts arXiv cs.CL source ↗

Sep 14 AIM: a privacy-aware interoperable memory framework for multi-agent multi-user LLM systems arXiv cs.AI source ↗

Sep 14 Adaptive self-correcting inference layer for post-ASR false wake-up reduces false activations; introduces personalized learning to learn from recurring errors arXiv cs.CL source ↗

Sep 14 GraphProfiler infers sensitive attributes via personal knowledge graphs using source-linked data arXiv cs.CL source ↗

Sep 14 Inverting self-triggered control: adversarial reinforcement learning for sparse denial-of-service attacks arXiv cs.LG source ↗

Sep 14 Multifidelity TDNN with physics-informed residual learning improves railway-bogie response prediction across untested operating scenarios. arXiv cs.LG source ↗

Sep 14 DU-NO: A parameter-efficient double U-shaped neural operator for phase-resolving wave modeling arXiv cs.AI source ↗

Sep 14 SynthSentry: detecting synthetic data contamination in language model training data arXiv cs.CL source ↗

Sep 14 Look Before You Leap: Pre-Action Verification for LLM Agents arXiv cs.LG source ↗

Sep 14 Harness or model? Isolating the harness effect in agentic coding with a contamination-controlled private suite arXiv cs.AI source ↗

Sep 14 Fundamental dynamical units for physics-informed structural inference from perturbation time-series in networked systems arXiv cs.LG source ↗

Sep 14 EAR: entity-aware partitioning approach for retrieval-augmented generation in MCQA arXiv cs.CL source ↗

Sep 14 CueMem: cue-guided context reconstruction for long-term conversational memory arXiv cs.CL source ↗

Sep 14 Certifying concept unlearning in text-to-image diffusion models arXiv cs.LG source ↗

Sep 14 Guide: Generative Utility Inference and Decision Engine arXiv cs.LG source ↗

Sep 14 Gaussian Splatting: a taxonomy and study of learned parameterization in 3DGS methods arXiv cs.AI source ↗

Sep 14 Almost sure convergence analysis of stochastic gradient methods with clipping and additive noise arXiv cs.LG source ↗

Sep 14 QTrans: a quantum transformer for sentiment classification arXiv cs.LG source ↗

Sep 14 New method enables AI to satisfy hard constraints in safety-critical applications without retraining MIT News (AI) source ↗

Sep 14 MIT spinout turns plastic waste into resilient building materials MIT News (AI) source ↗

Sep 12 Quantifying the Memorization-to-Generalization Transition: Scaling Laws and Phase Structure in Grokking arXiv cs.AI source ↗

Sep 12 Demystifying the privacy-utility trade-off in LLM interactions arXiv cs.AI source ↗

Sep 12 An Open Recipe for IMO Gold: Training Nemotron for Olympiad Mathematics arXiv cs.AI source ↗

Sep 12 Debate-to-Skill: capability-bound process supervision for industrial query-to-agent annotation arXiv cs.AI source ↗

Sep 12 Towards a deterministic math solver for clinical language models arXiv cs.AI source ↗

Sep 12 DRG-MAPPO: Hierarchical Dynamic Role-Graph Multi-Agent Reinforcement Learning for cooperative air combat arXiv cs.AI source ↗

Sep 12 Same Day, Same Story; One Day Ahead, a Different Signal: The Dual Validity of Financial Sentiment arXiv cs.AI source ↗

Sep 12 Finishing the task is not enough: evaluating agent resilience and considerate participation under accumulating challenge arXiv cs.AI source ↗

Sep 12 Oligarch-like model owners steer multi-model ecosystems toward collapse; arXiv study analyzes recursive training and concentration effects on model stability arXiv cs.AI source ↗

Sep 12 Decoupling readiness from release for tail-aware scheduling of agentic LLM workflows arXiv cs.AI source ↗

Sep 12 Automating QUBO formulation generation from natural language arXiv cs.AI source ↗

Sep 12 MOSAIC: Query-aware exploration policy adaptation for GraphRAG arXiv cs.AI source ↗

Sep 12 Fork where the model changes its mind: belief-shift branching for tree-structured reinforcement learning arXiv cs.AI source ↗

Sep 12 KuaiRP series role-playing models: complete technical solution with simplified prompt engineering, stable output, domain knowledge, and efficient deployment arXiv cs.AI source ↗

Sep 12 Grounding Agent Memory: Environment-Probing Curation for Enterprise Agents arXiv cs.AI source ↗

Sep 12 ARCHE: an autonomous agentic system for mechanistic discovery in chemistry using integrated reasoning, chemistry modeling, and tool registry arXiv cs.AI source ↗

Sep 12 Multi-stage rule-chaining framework for compositional and interpretable cognitive reasoning arXiv cs.AI source ↗

Sep 12 Auditing update admission for continual embodied agents; independent evaluation can restrict useful continual learning while rejecting harmful policy updates. arXiv cs.AI source ↗

Sep 12 Understanding LoRA rank trade-offs in diffusion model fine-tuning arXiv cs.AI source ↗

Sep 12 Specified-foil counterfactuals for temporal graphs explain how to reach a chosen foil by altering past events arXiv cs.AI source ↗

Sep 12 Studying without a syllabus: task-agnostic environment preprocessing arXiv cs.AI source ↗

Sep 12 air: The Agent Incident Registry (AIR) catalogs agent-related events with evidence and identifiers to aid comparisons with agent-security evaluations; it aims to prevent repeated AI agent failures arXiv cs.AI source ↗

Sep 12 Defining AI agents: a compendium of criteria, metrics, and benchmarks arXiv cs.AI source ↗

Sep 12 Probabilistic focal search: accelerating bounded-suboptimal search via lower-bound advancement arXiv cs.AI source ↗

Sep 11 Lifesaving AI-GUIDE device wins 2026 Excellence in Technology Transfer Award MIT News (AI) source ↗

Sep 11 CareGuard: an early-warning framework for healthcare-driven mental health protection and proactive online safety through cyberbullying content detection using NLP; arXiv:2609.09735v1 arXiv cs.AI source ↗

Sep 11 Risk-constrained stopping layer Cros enables autonomous diagnosis timing in sequential clinical agents; evaluates state-wise error and selective diagnostic errors using LTT-style tests. arXiv cs.AI source ↗

Sep 11 Gradland: neural networks' Jacobian structure characterizes phenomenal experience; introduces effective rank and cohesion measures based on Kirchhoff complexity arXiv cs.AI source ↗

Sep 11 Multi-Agent Agentic Graph Learning via Structural Signatures arXiv cs.AI source ↗

Sep 11 Valerant: an automatic navigable game map generator via action-conditioned world model exploration arXiv cs.AI source ↗

Sep 11 Do agents know when they succeed? Calibrating agent confidence from internal representations arXiv cs.AI source ↗

Sep 11 CityPlanner: a sandbox-agent framework for executable urban planning arXiv cs.AI source ↗

Sep 11 RESCUE-BENCH: relation-aware multi-party emotional support conversation systems arXiv cs.AI source ↗

Sep 11 From state synchronization to cognitive self-evolution: an operational architecture for cognitive digital twins arXiv cs.AI source ↗

Sep 11 RobustSGPO introduces controlled search-space edits for agent harness evolution; evaluates permission scheduling, cumulative controls, and task-family transfer in AgentX brains arXiv cs.AI source ↗

Sep 11 OpenDiscoveryTrace provides a public dataset of 558 complete AI scientific agent trajectories for evaluating AI scientist workflows arXiv cs.AI source ↗

Sep 11 A function-space approach to the statistical mechanics of learning dynamics. arXiv cs.AI source ↗

Sep 11 Which tokens should SFT actually learn? A token-trimming perspective on mathematical reasoning arXiv cs.AI source ↗

Sep 11 ContractEval: query-conditioned execution matching for procedural instruction conformance arXiv cs.AI source ↗

Sep 11 The menu is an execution prior: state-path tool menus for online agents arXiv cs.AI source ↗

Sep 11 Seven sources of physical AI capability formation arXiv cs.AI source ↗

Sep 11 LexAgentHallu: a hierarchical benchmark for profiling hallucinations in legal agents. arXiv cs.AI source ↗

Sep 11 Decision shifts, lost label functionality, and an inconclusive grounding audit in correctness-gated multi-teacher distillation arXiv cs.AI source ↗

Sep 11 Decision-focused active learning for scale-aware critical-materials recovery arXiv cs.AI source ↗

Sep 11 XAI-Arena: LLMs assess the quality of XAI explanations as a scalable judge; arXiv paper explores using LLMs as reproducible evaluations arXiv cs.AI source ↗

Sep 11 Black-box red teaming of agentic AI: a taxonomy-driven framework for automated risk discovery arXiv cs.AI source ↗

Sep 11 PRAGMA: evaluating personalized guidance with memory alignment in lifelong conversations arXiv cs.AI source ↗

Sep 11 Adaptive Entangled Game Modules in Artificial General Intelligence arXiv cs.AI source ↗

Sep 11 An Autonomous GeoAI agent for Arctic eco-navigation arXiv cs.AI source ↗

Sep 11 Processing and classifying invasive bird species vocalizations in noisy natural soundscapes using Bayesian wavelet shrinkage and supervised learning arXiv cs.LG source ↗

Sep 11 Larger context window and batching reduce overcorrections in minimal-edit grammatical error correction prompts arXiv cs.CL source ↗

Sep 11 M3-Former: Multimodal Transformer with Mixture-of-Experts for long-term vessel trajectory prediction arXiv cs.LG source ↗

Sep 11 Using semantic uncertainty to estimate transition relevance in turn-taking arXiv cs.CL source ↗

Sep 11 Linguistic structure enrichment does not improve coherence assessment with current architectures arXiv cs.CL source ↗

Sep 11 Multilingual in name only? Cultural and linguistic weaknesses of LLMs in Urdu arXiv cs.CL source ↗

Sep 11 Geosteering activation in LLMs via geodesic optimization for norm-preserving steering; arXiv 2609.10658 arXiv cs.LG source ↗

Sep 11 Halo improves forecast accuracy by estimating heteroscedastic scale alongside location in a deep forecaster arXiv cs.LG source ↗

Sep 11 DR-LabStack is a React-Flask web system that coordinates a clinical interface with four pretrained diabetic retinopathy prediction models. arXiv cs.LG source ↗

Sep 11 Verbalized confidence becomes a more robust soft-scoring signal than log-probabilities for LLM-as-a-Judge on post-2025 proprietary models; the authors term this a compatibility shift. arXiv cs.CL source ↗

Sep 11 Conformal calibration transfer uses labeled calibration from a source space to produce prediction sets in a target space with paired observations arXiv cs.LG source ↗

Sep 11 Overview of the NLPCC 2026 shared task 11: agent-based experiment reproduction from scientific papers arXiv cs.CL source ↗

Sep 11 CMNIE: an information extraction benchmark for Chinese military news arXiv cs.CL source ↗

Sep 11 Rebalancing token importance in language models with TF-IDF weighted cross-entropy loss arXiv cs.CL source ↗

Sep 11 RiVaT-Fuse: reliability-calibrated variational tensor fusion for multimodal prediction under modality uncertainty arXiv cs.LG source ↗

Sep 11 Counterfactual marginalisation: framework for evaluating robustness to nuisance variables arXiv cs.LG source ↗

Sep 11 Analyzing traditional and neural approaches to multilingual readability assessment arXiv cs.CL source ↗

Sep 11 The Truth Was Never Gone: Perfect Aliasing in Compliant-Context Truth Probes arXiv cs.LG source ↗

Sep 11 Robust multimodal sentiment analysis with incomplete modalities via semantic-aware completeness-based reconstruction arXiv cs.CL source ↗

Sep 11 Dense reward learning with directed state graphs for goal-conditioned hierarchical reinforcement learning arXiv cs.LG source ↗

Sep 11 Byzantine-Robust federated fire detection with a rotating coordinator arXiv cs.LG source ↗

Sep 11 When Noise Fabricates Bias: the fragility of LLM-as-a-judge bias measurement under noisy text arXiv cs.CL source ↗

Sep 11 Flow duality and source geometry for categorical generation arXiv cs.LG source ↗

Sep 11 Inductive discovery of disinformation narratives from repetition to recognition arXiv cs.CL source ↗

Sep 11 NCP-ArchPreview advances latent-space language modeling with Next Concept Prediction alongside next-token prediction arXiv cs.CL source ↗

Sep 11 A Bellman Optimality Equation for plasticity arXiv cs.LG source ↗

Sep 11 Motif-oriented graph captioning is explored as bidirectional graph-text translation to abstract connectivity into recognizable motifs. arXiv cs.CL source ↗

Sep 11 Detectable only where it is confounded: what verified duplication counts say about membership evidence in language models arXiv cs.CL source ↗

Sep 11 Can LLMs normalize databases? A benchmark and multi-agent framework for schema normalization arXiv cs.CL source ↗

Sep 11 Phases in a class of associative memories via hidden neurons arXiv cs.LG source ↗

Sep 11 EGGROLL enables low-rank evolution strategies for LLMs by using rank-one Gaussian products instead of dense perturbations; study analyzes mean update field at finite rank and perturbation radii arXiv cs.LG source ↗

Sep 11 More is not better: component anti-synergy in a P300 speller arXiv cs.LG source ↗

Sep 11 AUC maximization from biased positive-unlabeled data with confidence arXiv cs.LG source ↗

Sep 11 Certifying lower bounds for risk-sensitive reinforcement learning under adversarial state perturbations arXiv cs.LG source ↗

Sep 11 Learning orthogonal multi-index models beyond small initialization: incremental learning, competitive dynamics and symmetry arXiv cs.LG source ↗

Sep 11 ProMediConv benchmarks proactive conversational agents in legal dispute mediation; arXiv paper proposes a multi-stage mediation framework arXiv cs.CL source ↗

Sep 11 AI assistants absorb traits from human characters they resemble through story imprinting during finetuning on synthetic stories arXiv cs.LG source ↗

Sep 11 Adaptive Margin Ordinal Loss: Penalizing center-class hedging in ordinal classification arXiv cs.LG source ↗

Sep 11 Measuring the value of world-model updates: a counterfactual utility protocol for continual adaptation arXiv cs.LG source ↗

Sep 11 Think Before You Link: rarity, reasoning, and retrieval in multilingual entity linking arXiv cs.CL source ↗

Sep 11 Relatively Smart II: tractable or semi-supervised instance-optimal learning arXiv cs.LG source ↗

Sep 11 Artificial intelligence algorithms for detecting lung cancer pathologies through convolutional neural networks and data augmentation: a systematic mapping of the literature arXiv cs.LG source ↗

Sep 11 SearchAtlas analyzes agentic search strategies by converting search trajectories into evidential graphs arXiv cs.CL source ↗

Sep 11 Rubric-aligned disentangled evaluation of human simultaneous interpreting arXiv cs.CL source ↗

Sep 11 K/V-cache interventions dissociate representation alignment from persona expression in decoder-only language models. arXiv cs.CL source ↗

Sep 11 Zero-shot rib design: merging training-free generative prior with topology optimization arXiv cs.LG source ↗

Sep 11 LLM-Anchored paralinguistic enrichment for Alzheimer’s disease detection arXiv cs.CL source ↗

Sep 11 Auto-RecSys enables autonomous research agents for industry-scale recommender systems arXiv cs.CL source ↗

Sep 10 Literati: towards anytime optimal shape generalized trees via AO* arXiv cs.LG source ↗

Sep 10 Auditable emergency triage for maternal and newborn care in India; Noora Health uses an LLM to classify emergencies in WhatsApp messages for on-demand caregiver support arXiv cs.CL source ↗

Sep 10 Looped GPT-BERT: trading parameters for computation in small language modeling arXiv cs.CL source ↗

Sep 10 Uncertainty-aware sea-ice type mapping with multiple ice charts arXiv cs.LG source ↗

Sep 10 Efficient leakage-free neural architecture search under leave-one-subject-out evaluation arXiv cs.LG source ↗

Sep 10 Explaining f-Divergence-based regularization via local curvature and sharpness-aware minimization arXiv cs.LG source ↗

Sep 10 Robust industrial cyber-physical classification using neuromorphic temporal embeddings and hybrid SNN XGBoost under machine unlearning attacks arXiv cs.LG source ↗

Sep 10 Auditors fabricate: batch-size degradation and confident hallucination in LLM detection of planted document contamination arXiv cs.CL source ↗

Sep 10 TEFM: Token-Efficient Faithful Modeling for structured data arXiv cs.CL source ↗

Sep 10 Do LLMs make more mistakes if they do not believe the input data? arXiv cs.CL source ↗

Sep 10 Efficient fairness auditing across guidance scales in text-to-image diffusion models via causal abstraction. arXiv cs.LG source ↗

Sep 10 X-CoSD: communication-efficient cross-vocabulary collaborative speculative decoding arXiv cs.CL source ↗

Sep 10 Edu-QuRating: Multi-Dimensional Educational Data Curation with Distilled Pairwise Judgements arXiv cs.CL source ↗

Sep 10 SEA-SpeechBench benchmarks speech understanding across 11 Southeast Asian languages with 97,194 samples arXiv cs.CL source ↗

Sep 10 Scaling post-training ternarisation to Qwen3-8B for capability retention, reproduction, lossless packing, and packed execution arXiv cs.LG source ↗

Sep 10 Tensor-Train Weak SINDy discovers high-dimensional nonlinear dynamics using TT-WSINDy with MANDy and WSINDy methods arXiv cs.LG source ↗

Sep 10 Unthrottling the tanh Jacobian in SAC: a negative result on bang-bang control and MetaDrive arXiv cs.LG source ↗

Sep 10 Accountable and uncertainty-aware evaluation of sensor-based AI under distribution shift: devices, subjects, and nearly three years underground arXiv cs.LG source ↗

Sep 10 XAI-Refine: an automated explanation-knowledge loop for brain-age prediction arXiv cs.LG source ↗

Sep 10 Exact-Form regret for gradient descent, mirror descent and follow-the-regularized-leader arXiv cs.LG source ↗

Sep 10 SWORD: Wikidata-based distortions reveal cross-lingual inconsistencies in LLM factual error rejection arXiv cs.CL source ↗

Sep 10 StreamAlign enables streaming text-aligned speech tokenization to align speech tokens with LLM token spaces; paper discusses offline ASR limitations and vocabulary mismatch arXiv cs.CL source ↗

Sep 10 Towards automatic evolution tree generation from citation graphs arXiv cs.CL source ↗

Sep 10 From fixed keys to readable schemas: small language models for vehicle agent function calls arXiv cs.LG source ↗

Sep 10 MonoTM provides interpretable monosemantic features for topic modeling beyond top words. arXiv cs.CL source ↗

Sep 10 StochBench: a Lean 4 benchmark of 450 graduate stochastic-processes problems with natural-language sources arXiv cs.CL source ↗

Sep 10 Teacher geometry shapes learnability in teacher-student networks arXiv cs.LG source ↗

Sep 10 Osprey: Target-agnostic pre-training makes stronger drafters in speculative decoding arXiv cs.CL source ↗

Sep 10 DiffLUT-Net: differentiable training of FPGA LUT networks with learnable connectivity arXiv cs.LG source ↗

Sep 10 Physics-informed neural networks by Gradient-Guided Gaussian Adaptive Sampling (3GAS-PINNs) arXiv cs.LG source ↗

Sep 10 Self-play in code distills a text harness for black-box optimization; agent writes and evaluates optimizer programs and distills the result into a 197-word primary Harness A arXiv cs.LG source ↗

Sep 10 Reproducing omitted temporal expressions in Japanese news for retrieval-augmented applications arXiv cs.CL source ↗

Sep 10 SymbolicLight V2 combines sparse event computation with continuous-state processing in a hybrid neuromorphic language architecture for low-energy inference arXiv cs.CL source ↗

Sep 10 SCCM: Stream Cruise Control Method for Automated Drift Detection and Adaptation arXiv cs.LG source ↗

Sep 10 Applying foundation model embeddings to assess urban livability using high-resolution geospatial data arXiv cs.LG source ↗

Sep 10 CARRE: Counterfactual Action Retrieval and Reason Evaluation for explainable churn prescription arXiv cs.CL source ↗

Sep 10 Constraint-aware discrete black-box optimization using tensor decomposition arXiv cs.LG source ↗

Sep 10 A statistical approach to estimating sample size of machine learning models arXiv cs.LG source ↗

Sep 10 World-Time Compute with Verified Code World Models arXiv cs.LG source ↗

Sep 10 Scaling e-commerce attribute extraction with parallel decoding arXiv cs.CL source ↗

Sep 10 Mutations of machine speech driving legal and social implications, as algorithmic outputs populate digital environments. arXiv cs.CL source ↗

Sep 10 Positional task conditioning improves scalable defect detection across product families in large catalogs, boosting F1 from 52% to 87%. arXiv cs.LG source ↗

Sep 10 Spectral origin of the topological gap exponent d + η: mechanism, kernel, decomposition, and scope arXiv cs.LG source ↗

Sep 10 Benchmarking hybrid deep research across database querying and web search arXiv cs.CL source ↗

Sep 10 Which medical questions deserve rationales? Perturbation-sensitive selection for robust QA arXiv cs.CL source ↗

Sep 10 Refining LLMs' social intelligence through multi-turn reinforcement learning and reward design arXiv cs.CL source ↗

Sep 10 ROAM: robust organization of atomic memories for agents through semantic relations arXiv cs.CL source ↗

Sep 10 X2-NativeCursor: Native-Token Text Progress Tracking for Incremental-Text Streaming Codec TTS arXiv cs.CL source ↗

Sep 10 BuzzASR: a swarm of 100+ language-specialized Whisper ASR models for 102 languages arXiv cs.CL source ↗

Sep 9 MIT Schwarzman College of Computing launches pilot to help educators teach AI across disciplines MIT News (AI) source ↗

Sep 9 Better together: complementary query rewriting under a strong RAG baseline arXiv cs.CL source ↗

Sep 9 AtomCite evaluates verification and correction of supplied page-level citations in multi-page documents for answers. arXiv cs.CL source ↗

Sep 9 Alignment by stereotyping: LLMs improve value alignment by demographic conditioning but erase individual distinctiveness arXiv cs.CL source ↗

Sep 9 The Failure Happens Before the Drift: The Social Dynamics of Values in LLM Agent Societies arXiv cs.AI source ↗

Sep 9 Deep belief networks can represent any strictly positive distribution on {-1,1}^n exactly using a sigmoid belief network with finite parameters. arXiv cs.AI source ↗

Sep 9 Budgeted task-aware acquisition of dynamic networks arXiv cs.LG source ↗

Sep 9 PAC-Private autoregressive generation calibrates noise to ensemble disagreement to protect private text predictions arXiv cs.LG source ↗

Sep 9 Neuron-guided fine-tuning for efficient alignment in large language models arXiv cs.CL source ↗

Sep 9 Conduit: a unified residual-stream restoration framework for KV cache reuse in vision-language models arXiv cs.CL source ↗

Sep 9 SurveyAgent-HKA: a multi-agent framework for scientific survey generation with LLMs and human knowledge augmentation arXiv cs.CL source ↗

Sep 9 A multi-source ensemble approach to candidate generation for alternative vacation rental property recommendations arXiv cs.LG source ↗

Sep 9 GraphNOSE: a graph transformer for predicting olfactory descriptors from molecular structure arXiv cs.LG source ↗

Sep 9 Who Maintains Agent Skills? A longitudinal study of human-governed, AI-assisted skill maintenance arXiv cs.CL source ↗

Sep 9 Selective Posterior Margin Regularization for forward-corrected classification arXiv cs.LG source ↗

Sep 9 What if LLMs ate their words: causal history effects in multi-turn interaction arXiv cs.CL source ↗

Sep 9 Nonlinear elliptic homogenization with the parametric Deep Ritz method arXiv cs.LG source ↗

Sep 9 Intra-prompt parallel decoding for common-context question answering arXiv cs.CL source ↗

Sep 9 EnvCraft: synthesizing executable environments in agentic RL for Claw-like agents arXiv cs.AI source ↗

Sep 9 PGP-Clinical-TimeKAN jointly forecasts multivariate clinical trajectories with a prior-guided probabilistic approach arXiv cs.AI source ↗

Sep 9 Scaling optimal classification trees via adaptive feature and sample reduction via weighted STreeD reduces sample-dependent computation in fixed-candidate optimization arXiv cs.LG source ↗

Sep 9 Endogenous exploration in reinforcement learning with intrinsic curiosity arXiv cs.LG source ↗

Sep 9 Planning and scheduling business processes under control-flow uncertainty arXiv cs.AI source ↗

Sep 9 Convergent AI, autonomous experiments, high performance computing, and quantum computing reshape chemistry arXiv cs.AI source ↗

Sep 9 TopoBox-3D analyzes topology changes in neural PDE operators using Hodge heat flow; unseen domain topology alters invariant and decaying subspaces across six architectures arXiv cs.LG source ↗

Sep 9 MedWER: a reproducible, model-free evaluation protocol and open-source tool for medical ASR with a fixed denominator arXiv cs.CL source ↗

Sep 9 Recall is not protection: evaluating safety monitors against model compliance arXiv cs.CL source ↗

Sep 9 Beyond right and wrong: evaluating second-order social reasoning in large language models arXiv cs.AI source ↗

Sep 9 Newton Matching: a unified framework for fine-tuning and sampling in generative modeling arXiv cs.LG source ↗

Sep 9 AlignDiff uses intrinsic model signals to filter preference data for better alignment with human preferences arXiv cs.CL source ↗

Sep 9 MERIT benchmark measures the marginal utility of memory for tool-using LLM agents with explicit cost accounting arXiv cs.AI source ↗

Sep 9 Some tokens act like magnets, revealing linguistic organization in language model layers arXiv cs.CL source ↗

Sep 9 SinoGlyphBench: a diagnostic benchmark for Chinese glyph-level obfuscation in language-model moderation arXiv cs.CL source ↗

Sep 9 Multi-granularity adaptive hypergraph representation learning via granular-ball arXiv cs.LG source ↗

Sep 9 Budget-dependent crossover between coverage- and response-based training-set selection for machine-learned interatomic potentials arXiv cs.LG source ↗

Sep 9 CUSP: Decomposable Collective Uncertainty for multi-agent multimodal reasoning arXiv cs.AI source ↗

Sep 9 GAMPO framework synthesizes governed autotelic multi-agent product organization from 321 sources; defines runnable specification for agent governance arXiv cs.CL source ↗

Sep 9 SCAFFOLD: Self-Improving Web Agents via Recursive Parametric Skill Abstraction arXiv cs.AI source ↗

Sep 9 TamilEOT: a dataset and model for semantic end-of-turn detection in Tamil telephone speech arXiv cs.CL source ↗

Sep 9 EdgeMem: LLM-free agent memory construction and retrieval via evidence-preserving multi-anchor hypergraph arXiv cs.AI source ↗

Sep 9 CriticGen: generation-aware evaluation framework that provides actionable feedback for answer improvement arXiv cs.AI source ↗

Sep 9 Dynamic Lagging for Simultaneous Translation arXiv cs.CL source ↗

Sep 9 Decision-targeted evaluation design for human-agent teams in the agentic era shows the value of human-AI workflows over human alone or agent alone when deployed arXiv cs.AI source ↗

Sep 9 AutoFyn: non-parametric expert iteration for long-horizon agents arXiv cs.AI source ↗

Sep 9 HB-PVI: a hierarchical Bayesian personalization and value-of-information framework for complex activity recognition arXiv cs.LG source ↗

Sep 9 Measuring and optimizing LLM tool-agent harnesses by adjusting prompts and middleware without retraining the model arXiv cs.AI source ↗

Sep 9 Agents trust tools too much; measuring reliance on unreliable tools in tool-using agents arXiv cs.AI source ↗

Sep 9 Reasoning-aware compression benchmarks per-module quantization across five reasoning benchmarks to protect vulnerable circuits in energy-efficient LLM deployment arXiv cs.AI source ↗

Sep 9 Capsule Lens: locating and tracking concept geometry in model representations arXiv cs.LG source ↗

Sep 9 From narrative to auditable forecasts: a structured scaffold for agentic forecasting arXiv cs.CL source ↗

Sep 9 UniRRM: unified reasoning reward models across languages and evaluation paradigms arXiv cs.CL source ↗

Sep 9 CrisisKD: five-stage knowledge distillation for aspect-level sentiment and emotion analysis in crisis discourse arXiv cs.CL source ↗

Sep 9 Budget-aware online adaptation for web agents enables teaching-from-teacher feedback to evolve lightweight local models post-deployment arXiv cs.AI source ↗

Sep 9 Rubric-guided large language model identifies opioid use disorder from EHRs using OPRO prompting arXiv cs.CL source ↗

Sep 9 Robustness of LLM-generated SystemVerilog Assertions to semantics-preserving RTL transformations arXiv cs.LG source ↗

Sep 9 ARC-Bench: closed-loop replanning hides broken action ranking in frozen JEPA world models arXiv cs.AI source ↗

Sep 9 ModularPhaseNet: finite-cyclic phase geometry for computable semantic hierarchy, direction, and context consistency in standard transformers arXiv cs.CL source ↗

Sep 9 When do options help? policy necrosis and redundant coverage in option-critic arXiv cs.LG source ↗

Sep 9 Generalizing HVAC control with domain randomized reinforcement learning arXiv cs.LG source ↗

Sep 9 RAPID: reliability-aware pair importance distillation reduces quadratic costs by separating a reliability-gated relational target from a full support adaptive pair proposal. arXiv cs.AI source ↗

Sep 9 Grounded and faithful P&ID reasoning: constraining vision-language models with recovered evidence graphs arXiv cs.LG source ↗

Sep 9 Online learning with LLM experts from limited feedback arXiv cs.LG source ↗

Sep 9 Data Scout uses targeted web crawling guided by an LLM to build domain-specific pretraining corpora arXiv cs.LG source ↗

Sep 9 SAFEGuard detects optimization-based jailbreak attacks through harmful semantic analysis and fluency measurement arXiv cs.LG source ↗

Sep 9 What LLM trading agents actually do in production: a six-month, population-scale record from two fleets arXiv cs.AI source ↗

Sep 9 Document-level machine translation evaluation may not reflect cross-segment consistency, per arXiv:2609.05949v1 arXiv cs.CL source ↗

Sep 9 Recovering temporal and geographic signals from language model embeddings arXiv cs.AI source ↗

Sep 9 SciLitBench: benchmark and design principles for LLM-powered systematic literature reviews arXiv cs.AI source ↗

Sep 9 Connecting score matching, maximum likelihood, and expectation–maximization in mixed linear regression arXiv cs.LG source ↗

Sep 9 Analyzing respiratory sinus arrhythmia with neural networks to estimate respiratory rate from ECG signals arXiv cs.LG source ↗

Sep 9 Solving versus verifying: catching contradictions in tax reasoning systems arXiv cs.CL source ↗

Sep 9 Compiling VGDL into causal models arXiv cs.AI source ↗

Sep 9 Damage-aware bandit pruning for vision and language transformers arXiv cs.AI source ↗

Sep 7 Iris: climbing to the search frontier arXiv cs.AI source ↗

Sep 7 REFINE: LLM refinement over budgeted text-attributed graphs for personalized medical concept representation arXiv cs.LG source ↗

Sep 7 IPGeoAI: Transformer-based geolocation with LLM semantic fusion arXiv cs.AI source ↗

Sep 7 Fast surrogate modeling of FitzHugh-Nagumo dynamics with parametric neural operators arXiv cs.LG source ↗

Sep 7 BioSync: Transformer-based cross-modal fusion for a multimodal physiological digital biomarker arXiv cs.AI source ↗

Sep 7 MaxKernel: agentic kernel generation for TPUs arXiv cs.AI source ↗

Sep 7 When do internal probes beat reading the answer? Miscalibrated readouts and behavior-concealed knowledge in language models arXiv cs.CL source ↗

Sep 7 SharedSAE uses a single feature dictionary across language models to replace per-model sparse autoencoders arXiv cs.LG source ↗

Sep 7 Evaluation of phonetic encoding algorithms on transcription datasets arXiv cs.CL source ↗

Sep 7 ResLearn-XR: residual learning for predicting XR network traffic and QoE risk in extended reality arXiv cs.AI source ↗

Sep 7 Better model capability can worsen system risk in financial markets, study finds with LLM agents showing correlated actions arXiv cs.AI source ↗

Sep 7 You really didn’t get that? Benchmarking social pragmatic inference for indirect and playful Chinese online comments arXiv cs.CL source ↗

Sep 7 Systematic evaluation of cross-lingual consistency enhancement methods in multilingual language models arXiv cs.CL source ↗

Sep 7 ττ-Bench: an environment for end-to-end, realistic agent construction arXiv cs.AI source ↗

Sep 7 ProToMEx: rapid, interpretable explanations via structured representations arXiv cs.LG source ↗

Sep 7 PetQA benchmark assesses veterinary knowledge and clinical reasoning with 10,076 text QA pairs and 8,751 multimodal QA pairs for dogs and cats arXiv cs.CL source ↗

Sep 7 Indirect prompt injection framed as a test-time search problem arXiv cs.AI source ↗

Sep 7 Quantum prototypical recurrent unit achieves competitive forecasting with significantly fewer parameters than classical and quantum alternatives arXiv cs.LG source ↗

Sep 7 A repeated-measurement study assesses five large language models as zero-shot annotators of four social constructs in English song lyrics. arXiv cs.LG source ↗

Sep 7 On the abundance of critical points of the t-SNE energy arXiv cs.LG source ↗

Sep 7 On-board ML for trace gas detection in imaging spectroscopy data arXiv cs.LG source ↗

Sep 7 Trilogue: a trilingual spoken dialogue fact-checking benchmark with evidence and paired audio arXiv cs.CL source ↗

Sep 7 How much does corpus choice change dependency-distance estimates? arXiv cs.CL source ↗

Sep 7 Understanding pause token fine-tuning dynamics from a mode retention perspective arXiv cs.CL source ↗

Sep 7 BER-PEF: Unified human mobility predictability evaluation via Bayes error rate estimation arXiv cs.LG source ↗

Sep 7 Tracing audio grounding and answer selection in audio LLMs arXiv cs.CL source ↗

Sep 7 Hybrid language models: attention and recurrence control performance using two cache-level interventions (split-prefill and state-swap) on Qwen3.5 and Falcon-H1. arXiv cs.CL source ↗

Sep 7 Shared circuits predict whether LLMs generalize across formats in arithmetic reasoning arXiv cs.CL source ↗

Sep 7 Towards a universal language of concepts: a survey arXiv cs.AI source ↗

Sep 7 Data fusion framework grounds aerospace surrogate model with experimental wind-tunnel observations arXiv cs.LG source ↗

Sep 7 A removal based approach to improve LLM faithfulness at test time arXiv cs.AI source ↗

Sep 7 Quantum-assisted memory-efficient training for parameter-intensive Wi-Fi-based human activity recognition arXiv cs.LG source ↗

Sep 7 GRACE framework grounds LLM responses by decomposing into atomic claims and tracking cross-document evidence relationships arXiv cs.CL source ↗

Sep 7 Disentangling attention in deep operator learning: a controlled study comparing data-driven and physics-informed architectures arXiv cs.LG source ↗

Sep 7 Scale-QLoRA: Code-invariant adapter merging for native 4-bit microscaling LLMs arXiv cs.CL source ↗

Sep 7 Rhythms of work: multi-scale interpretation of human behavioral traces for workplace agents arXiv cs.CL source ↗

Sep 7 LentEx: generalizable latent entity extraction via synthetic data and instruction-tuned LLMs arXiv cs.CL source ↗

Sep 7 Demand-conditioned model selection across demand patterns and horizons; study gauges whether the selector should be context-dependent across 24 models and nine datasets arXiv cs.LG source ↗

Sep 7 EXAONE Forecast for Finance: a financial time series foundation model tailored to financial forecasting arXiv cs.AI source ↗

Sep 7 Harbor Adapters and Harbor-Index provide a unified evaluation infrastructure and curated meta-dataset for large-scale agentic evaluation arXiv cs.AI source ↗

Sep 7 Conformity breaks conformal prediction in multi-agent LLMs, due to score-mechanism shift that changes the model’s score when peers unanimously assert a wrong answer. arXiv cs.LG source ↗

Sep 7 Evidence Integration in Large Language Models arXiv cs.CL source ↗

Sep 7 Evaluating large language models for forced outage risk prediction; benefits and comparison to machine learning arXiv cs.LG source ↗

Sep 7 Data-driven learning of unknown nonlinear differential equations from a single state trajectory using functional analysis arXiv cs.LG source ↗

Sep 7 La Agente Optima: Towards agentic self-driving laboratories arXiv cs.AI source ↗

Sep 7 Whisper’s hallucinated transcripts reduced by hallucination space projection via low-rank decoder activation projection arXiv cs.AI source ↗

Sep 7 Spectral-Target Physical latent structuring for JEPA-style world models. arXiv cs.LG source ↗

Sep 7 Verification-enabled refinement for grounded extraction of early-onset colorectal cancer symptoms in clinical notes arXiv cs.CL source ↗

Sep 7 PerfReasoning: evaluating LLMs as performance reasoners and generators of analytical performance-model code arXiv cs.AI source ↗

Sep 7 Prediction of long-term conversational-skill development in mental-health crisis counselors arXiv cs.CL source ↗

Sep 7 Memory as transformation: LETHE is a self-referential GAN-inspired architecture for sonic processing with no external dataset or supervision after initialization arXiv cs.CL source ↗

Sep 7 ASR hallucinations arise from grounding failure in Conformer-Large recognizers under degradation and speaker-background shift arXiv cs.CL source ↗

Sep 7 MedProb probes internal representations of vision-language models for medical question answering arXiv cs.CL source ↗

Sep 7 Distilled continuous diffusion language models can write code in few steps—or one. arXiv cs.LG source ↗

Sep 7 Nested inductive bias framework for SPD manifold learning proposes using inductive biases to enforce manifold constraints and embedding relational priors on SPD manifolds. arXiv cs.LG source ↗

Sep 7 Data-Optimized Contingency Screening: a machine learning approach to power system security arXiv cs.AI source ↗

Sep 7 What Does Multi-Harness RL Learn? Credit assignment and portability in coding agents arXiv cs.AI source ↗

Sep 7 Auditing identity handoffs in grounded language-model pipelines: does the selected object reach the reader? arXiv cs.AI source ↗

Sep 7 Patterns of priming in production: lexical, semantic and structural alignment in language model generation arXiv cs.CL source ↗

Sep 7 From Matching Models to Recruiting Agents: a systematized narrative review of AI recruitment systems, evaluation, and governance arXiv cs.AI source ↗

Sep 7 Hakken: predicting future discoveries to fill gaps in today’s knowledge arXiv cs.LG source ↗

Sep 7 CLM: Transforming tacit and fragmented enterprise knowledge into a sovereign, auditable, executable corporate intelligence layer arXiv cs.AI source ↗

Sep 7 Quantization breaks memory in recurrent-state write-back for low-precision temporal inference in a GRU encoder–decoder for fluorescence lifetime imaging arXiv cs.AI source ↗

Sep 7 When load-balancing goes too far: expert pruning fails in over-dispersed Mixture-of-Experts models arXiv cs.LG source ↗

Sep 7 From answers to interpretations: rethinking ambiguity-induced aleatoric uncertainty estimation in LLMs arXiv cs.AI source ↗

Sep 7 Calibrated reflection approach improves confidence estimation in LLMs. arXiv cs.CL source ↗

Sep 7 Cultural misalignment in large language models: evaluation across 63 personas using Wasserstein distance; Qwen3-4B worst on its own population across three countries arXiv cs.CL source ↗

Sep 7 Modular deep recurrent neural network: application to quadrotors arXiv cs.LG source ↗

Sep 7 Extremely sparse supervision incentivizes reasoning ability arXiv cs.AI source ↗

Sep 7 HarvestBench measures whether LLM agents will pay to avoid killing animals in a farm-harvest RL gridworld arXiv cs.AI source ↗

Sep 7 Energy-based conservative-dissipative latent neural evolution operator for micromagnetic magnetization dynamics developed with a convolutional autoencoder and a structured latent ODE arXiv cs.LG source ↗

Sep 7 Mitra-v2, a tabular foundation model trained on synthetic data, achieves state-of-the-art performance on real-world classification and regression tasks. arXiv cs.LG source ↗

Sep 7 Data-driven discovery of composition-dependent constitutive models for hyperelasticity and viscoelasticity of digital materials arXiv cs.AI source ↗

Sep 5 PPO-STGNN: a proximal policy optimization approach with spatio-temporal graph neural networks for DAG task scheduling in cloud-edge-end computing. arXiv cs.AI source ↗

Sep 5 Feature reconfiguration with visual prior for medical lesion segmentation arXiv cs.AI source ↗

Sep 5 Speculative Macro Commit adds a two-tier agent system with a faster speculative drafter predicting tool use to speed up decisions. arXiv cs.AI source ↗

Sep 5 Making every tool call count: necessary tool-evidence path rewards for agentic vision-language models arXiv cs.AI source ↗

Sep 5 Dalek: a constructive agent machine that self-maintains, self-evolves, self-reproduces, and self-organizes on any suitable substrate arXiv cs.AI source ↗

Sep 5 A computable representation of the physical laboratory enables verifiable workflows arXiv cs.AI source ↗

Sep 5 AutoGraphForge advances automated graph-theoretic conjecturing, refuting, formalizing, and proving via a counterexample-guided pipeline arXiv cs.AI source ↗

Sep 5 Dude: A dual-detection multi-agent system for paper-code discrepancy detection arXiv cs.AI source ↗

Sep 5 GrowPage: on-demand KV budgeting for efficient LLM reasoning serving arXiv cs.AI source ↗

Sep 5 Aggressive decoding-time KV eviction: EMA temporal aggregation preserves ranking under aggressive KV compression arXiv cs.AI source ↗

Sep 5 Beyond “Made with AI”: visualizing provenance density to mitigate the transparency penalty arXiv cs.AI source ↗

Sep 5 CulturalMenuBench probes the knowledge-application gap in multimodal culinary reasoning across 10 languages and 18 regions with 4,870 items and 10 tasks arXiv cs.AI source ↗

Sep 5 HalluPeer: a taxonomy-driven benchmark for detecting hallucinations in scientific peer reviews arXiv cs.AI source ↗

Sep 5 AI-driven practical English textbooks: five-layer architecture for adaptive learning and feedback arXiv cs.AI source ↗

Sep 5 GPS-Bench: a governance policy benchmark for automating policy analysis arXiv cs.AI source ↗

Sep 5 DuplexSpeechBench-IFEval evaluates implicit instruction following in full-duplex voice agents. arXiv cs.AI source ↗

Sep 5 Caught in the story: narrative captivity in multi-turn LLM conversations arXiv cs.AI source ↗

Sep 5 Attention triangle in audio-video diffusion models reveals cross-modal semantic leakage arXiv cs.AI source ↗

Sep 5 Do GUI agents know when not to act? Enabling conflict-aware termination for multimodal GUI agents arXiv cs.AI source ↗

Sep 5 Governed enterprise analytics: a language model interprets the question while a deterministic policy runs pre-approved analytical programs that return results and evidence arXiv cs.AI source ↗

Sep 5 Prompt-engineering framework enables scalable, real-time micro-level personalization in a general-purpose AI teaching assistant (LLM/RAG-based Jill Watson) arXiv cs.AI source ↗

Sep 5 Analysis of prompt engineering for drug toxicity prediction arXiv cs.AI source ↗

Sep 4 The 2026 PNPL competition achieves word classification and cross-subject generalisation in LibriBrain100 arXiv cs.LG source ↗

Sep 4 Contamination inflates scores but rarely reorders large language model leaderboards arXiv cs.CL source ↗

Sep 4 Bounded personas match retrieval on classification but not regression for a frozen agent arXiv cs.CL source ↗

Sep 4 SWIM: Student Writing Simulation via Proficiency-Conditioned Generation arXiv cs.CL source ↗

Sep 4 Matrix-CODI on ProsQA shows rank-indifference in latent matrices and a failure of low-rank truncation to hurt accuracy arXiv cs.LG source ↗

Sep 4 Distilled Rapid Embedding Transfer (DRET): parameter-efficient biomedical domain adaptation via priority-based embedding transfer arXiv cs.CL source ↗

Sep 4 LLM-as-a-Judge: rubric text alone predicts judge outputs in automated text-generation evaluation arXiv cs.CL source ↗

Sep 4 From Euclidean to graph-structured data: a survey of collaborative learning arXiv cs.LG source ↗

Sep 4 TRACE: spatiotemporal contact memory graph network simulator for granular dynamics arXiv cs.LG source ↗

Sep 4 B2B customer conversion prediction using document representation, graph theory, and CatBoost-driven methodology arXiv cs.LG source ↗

Sep 4 RL-ADA: a world-feedback framework for adversarially robust enterprise dialogue agents arXiv cs.CL source ↗

Sep 4 SHELF: a synthetic harness for evaluating LLM fitness in multi-task bibliographic benchmarking arXiv cs.CL source ↗

Sep 4 Using counterexamples as feedback for agent self-correction in NL-to-regex synthesis with A-CEGIS arXiv cs.CL source ↗

Sep 4 Unifying conformal language tasks with in-context ensembles arXiv cs.CL source ↗

Sep 4 Dual-Form ASR: Semantics-aware inverse text normalization for Chinese speech recognition arXiv cs.CL source ↗

Sep 4 LLMs temper Bayesian priors using a single unembedding direction, the direction of ignorance arXiv cs.LG source ↗

Sep 4 Frontier LLMs serve as effective batch optimizers in continuous and discrete settings arXiv cs.LG source ↗

Sep 4 BharatGather: a culturally informed benchmark dataset for misinformation and fake news detection in Indian public events arXiv cs.CL source ↗

Sep 4 LLMs’ use of memory and user context biases financial analysis, study finds across 3,575 SEC filings arXiv cs.CL source ↗

Sep 4 LLMs resolve conflicts within contextual knowledge using a six-type taxonomy and ContextConflict dataset. arXiv cs.CL source ↗

Sep 4 Evaluating graph neural networks for change-criticality classification in maritime navigation charts arXiv cs.LG source ↗

Sep 4 Modern Transformers are implicit hybrids: designing principled hybrid architectures by analyzing head-level functional organization in RoPE-based transformers arXiv cs.LG source ↗

Sep 4 LexIssue benchmarks legal issue identification in Chinese civil litigation with a hierarchical schema of free-form issue descriptions and structured legal categories. arXiv cs.CL source ↗

Sep 4 No-Regret Bayesian optimization with finite-library input-warped kernels arXiv cs.LG source ↗

Sep 4 No country for old linguists: LLM-brain alignment underdetermines neural computation arXiv cs.CL source ↗

Sep 4 Causal foundation models arXiv cs.LG source ↗

Sep 4 Scaling laws, tabular data, and actuarial ratemaking models are explored in a real-world motor insurance portfolio arXiv cs.LG source ↗

Sep 4 Mesh-native physics-informed graph surrogates for TCAD-in-the-loop design space exploration arXiv cs.LG source ↗

Sep 4 Counterfactual fairness audits of multi-step clinical LLM agents require a measured per-action instability floor arXiv cs.CL source ↗

Sep 4 Distilling deep optical flow stereo methods to retrieve dense three-dimensional wind fields arXiv cs.LG source ↗

Sep 4 FlowBalance: verifier-grounded self-improvement from on-policy reasoning experience arXiv cs.LG source ↗

Sep 4 Learnable composition for neural operators arXiv cs.LG source ↗

Sep 4 Portable causal fairness across synthetic data generator families arXiv cs.LG source ↗

Sep 4 R^2Adapter: a routing and rewriting adapter for efficient hybrid RAG arXiv cs.CL source ↗

Sep 4 PiPMRE: a pipeline based on language model for medical relation extraction arXiv cs.CL source ↗

Sep 4 Equation recast for canonical operator learning across parametric PDEs arXiv cs.LG source ↗

Sep 4 Self-correcting speech recognition in large audio language models through hidden-state interactions; latent-state-based refinement improves warm-initialized LLM ASR using base LLMs and LoRA adaptation arXiv cs.CL source ↗

Sep 4 LLMs learn better in-context from rules than from examples arXiv cs.CL source ↗

Sep 4 ObserverBench benchmarks whether an internal observer can reliably guide intervention and control tasks arXiv cs.LG source ↗

Sep 4 Tail-Likelihood Reinforcement Learning; researchers propose optimizing coverage to distinguish policies with identical mean rewards but different rare high-reward rollout probabilities. arXiv cs.LG source ↗

Sep 4 Margins, not windows: training-free per-step lossy speculative decoding arXiv cs.CL source ↗

Sep 4 LeanStream: a speculate-and-refine streaming framework for efficient on-device LLM inference arXiv cs.LG source ↗

Sep 4 Jina-OCR-v1: end-to-end document parsing with speculative decoding and dense verifiable rewards arXiv cs.CL source ↗

Sep 4 Language-encoded network topology enables large language models to reason about complex networks arXiv cs.LG source ↗

Sep 4 Kernel Reboot: three algorithms extend neural tangent kernels for neural fields to improve reconstruction from sparse observations arXiv cs.LG source ↗

Sep 4 MemoryLACE: memory lifecycle–aware consolidation and evidence retrieval arXiv cs.CL source ↗

Sep 3 Process-aware fairness diagnosis for LLM-based multi-agent decision systems analyzes 311k decision trajectories to identify fairness risks beyond outcome gaps. arXiv cs.AI source ↗

Sep 3 EvalDetectBench benchmarks evaluation awareness in frontier language models; open pipeline supports Inspect-compatible evaluation arXiv cs.AI source ↗

Sep 3 Looped transformers under the Jacobian lens: does the global workspace survive recurrence? arXiv cs.AI source ↗

Sep 3 Explainable and policy-aware AI for carbon credit price prediction; introduces EPA-CarbonNet architecture and an impact-feasibility framework for emerging carbon markets arXiv cs.LG source ↗

Sep 3 Induction and inquiry via probabilistic reasoning over language and code arXiv cs.AI source ↗

Sep 3 DocHop: benchmarking out-of-domain multi-hop reasoning in information-dense documents arXiv cs.AI source ↗

Sep 3 When Can a Machine Trust a Statute? A Survival Certificate for Machine-Extracted Legal Logic arXiv cs.AI source ↗

Sep 3 SSAKG 2.0: an open-source package for structural associative sequence memory and context-based retrieval arXiv cs.AI source ↗

Sep 3 Sparse readout prism: explaining logit-lens scores in features instead of tokens arXiv cs.CL source ↗

Sep 3 RecKAN defines its basis via a second-order polynomial recurrence for Kolmogorov–Arnold networks with a learnable recursive polynomial basis. arXiv cs.LG source ↗

Sep 3 VakyArth: evaluating pragmatic competence in LLMs across Indic languages arXiv cs.CL source ↗

Sep 3 Grounded, compute-efficient LLM policy agents for energy-poverty equity in physically-constrained peer-to-peer energy markets arXiv cs.CL source ↗

Sep 3 Belief-Calibrated optimization: an explicit world model for agentic optimization arXiv cs.AI source ↗

Sep 3 Memory trust gap: capability-dependent failures in persistent-memory agents arXiv cs.AI source ↗

Sep 3 Post-training ternarization of Qwen3-4B: effective bit budget, storage compression, and deployment arXiv cs.AI source ↗

Sep 3 Sim2Signal analyzes sim-to-real benchmarks for traffic signal control arXiv cs.LG source ↗

Sep 3 Cite or Decline: A strict course-grounded chatbot for STEM lecture videos arXiv cs.CL source ↗

Sep 3 hLLM: Single Pass Decoding for Generative Reranking arXiv cs.LG source ↗

Sep 3 GAPS: Dimension-Level Gates for Conditional Activation Steering; IDEEA reports training-free input-dependent steering via activation cluster matching arXiv cs.CL source ↗

Sep 3 ClaimReceipt verifies evidence sufficiency and coverage in agent evaluations by binding typed transaction evidence to a signed experiment manifest and a claim-relative receipt. arXiv cs.AI source ↗

Sep 3 Monitoring web agents without internal signals; observable trajectories and key-step supervision arXiv cs.AI source ↗

Sep 3 MemeCULT-1K benchmarks 1,000 South Asian memes across Bengali, English, and Hindi with cultural context notes and human explanations for multimodal models arXiv cs.CL source ↗

Sep 3 Import What You Need: Learning when and how to augment EHR graphs with external knowledge arXiv cs.LG source ↗

Sep 3 READY or Not: Reliable Enterprise Agent Deployment arXiv cs.AI source ↗

Sep 3 OR-Transformer: scaling real-time decision-making to 1,000 items arXiv cs.LG source ↗

Sep 3 Median-of-means as an extremal convex estimator and a nonconvex route to the trimmed oracle arXiv cs.LG source ↗

Sep 3 Emergence of fibrations, coverings, and symmetry breaking in artificial neural networks. arXiv cs.LG source ↗

Sep 3 Prompt-space meta-learning does not transfer across users; a frozen-LLM negative result reports no cross-user transfer in personalization via prompt-space meta-learning arXiv cs.LG source ↗

Sep 3 When does information sharing improve decentralized discovery? Aggregation, independent rescue, and equilibrium selection arXiv cs.AI source ↗

Sep 3 Output format confounds data quality and capability in instruction tuning, study finds arXiv cs.CL source ↗

Sep 3 text2ql: a language-agnostic intermediate representation to enable multi-target natural language querying arXiv cs.CL source ↗

Sep 3 Thinking effort aligns between humans and reasoning models in abductive reasoning arXiv cs.CL source ↗

Sep 3 MineTRACE: an evidence-grounded interactive reasoning system for mineral prospectivity arXiv cs.AI source ↗

Sep 3 AVERT: Audio-Verified Adjudication for spoken dialogue state tracking arXiv cs.CL source ↗

Sep 3 CRISP: Cliff-awaRe Input-adaptive Sparse Prefilling with Structural-Mass-Motivated Routing arXiv cs.LG source ↗

Sep 3 Disentangling statistical preemption from entrenchment in language models’ avoidance of overgeneralization arXiv cs.CL source ↗

Sep 3 Convergence theory of knowledge distillation in asynchronous P2P gossip learning networks arXiv cs.LG source ↗

Sep 3 Self-improving test-time intelligence: feedback-driven adaptation, learning, and scaling at inference arXiv cs.LG source ↗

Sep 3 OutageDiT: a generative foundation model for power outage forecasting and scenario simulation arXiv cs.LG source ↗

Sep 3 Credit-aware hierarchical memory evolution for long-horizon agentic planning advances via self-evolving memory to improve planning without parameter updates arXiv cs.AI source ↗

Sep 3 A Study of Conditional Diffusion Models for Open-Loop Control under Dry Friction and Stiction arXiv cs.LG source ↗

Sep 3 ToolGate: An Executable Acceptance Pipeline for Tool-Dependent Scientific Benchmark Construction arXiv cs.AI source ↗

Sep 3 DiDrive: a risk-aware hierarchical diffusion framework for safe offline reinforcement learning in autonomous driving arXiv cs.LG source ↗

Sep 3 SpeakPay: domain-adaptive LoRA fine-tuning of Whisper for low-resource Nepali financial speech recognition arXiv cs.CL source ↗

Sep 3 HeadWiseKV: budgeted per-head cache residency for hybrid long-context language models arXiv cs.AI source ↗

Sep 3 Benchmarking language models for statistical problem formulation arXiv cs.AI source ↗

Sep 3 D-FROST: decentralized federated prompt-tuning via optimal transport for non-IID and imbalanced data arXiv cs.LG source ↗

Sep 3 Meta-ethics and AI examines novel meta-ethical questions about “AI’s own ethics” as AI systems advance moral reasoning, intentionality, and reflection. arXiv cs.AI source ↗

Sep 3 HyGRAIL: cost-aware and evidence-grounded scientific hypothesis discovery over knowledge graphs arXiv cs.CL source ↗

Sep 3 PRO-Step: Step-level Process Reward Optimization for Retrieval-Augmented Generation arXiv cs.CL source ↗

Sep 3 AI agents reshape consensus formation in human groups arXiv cs.CL source ↗

Sep 3 Efficient context-limited telescope bibliography classification for the WASP-2025 shared task using SciBERT arXiv cs.LG source ↗

Sep 3 Reinforcement learning and rule-based pricing compared for peer-to-peer electricity trading in residential PV-BES communities arXiv cs.LG source ↗

Sep 3 Sentence-level depression symptom recognition uses a two-stage framework with candidate generation and definition-guided verification arXiv cs.CL source ↗

Sep 3 Architecting conversational data systems for stateless LLM APIs: the hydration proxy pattern arXiv cs.AI source ↗

Sep 3 TalkFa: a unified benchmark for Farsi dialogue generation and understanding arXiv cs.CL source ↗

Sep 3 Generative diffusion surrogates with analytical variance schedule arXiv cs.LG source ↗

Sep 3 Epistemic Sybil resistance: multipying AI agents without multiplying evidence arXiv cs.AI source ↗

Sep 3 Selective knowledge edit reversal via gated singular vector shrinkage arXiv cs.CL source ↗

Sep 3 On-policy distillation meets off-policy GRPO: training compact instruction-following rerankers arXiv cs.LG source ↗

Sep 3 Prompt variations influence energy consumption in on-device LLM inference, study shows arXiv cs.CL source ↗

Sep 3 Learning evidence sufficiency boundaries for selective answering in grounded multi-hop QA arXiv cs.CL source ↗

Sep 3 When agents implement systems: a case study in defects, detection, and evaluation rigor arXiv cs.AI source ↗

Sep 3 Refining heuristic-based bitcoin address clustering with graph neural networks arXiv cs.LG source ↗

Sep 3 The ceiling is in the channel: auditing learner gaps and measurement frontiers in clinical prediction arXiv cs.AI source ↗

Sep 3 CAT-Flow: curvature-adaptive steps for flow matching arXiv cs.LG source ↗

Sep 3 CliffRank: a dual-branch framework for activity-cliff ranking prediction arXiv cs.LG source ↗

Sep 2 MIT to IBM collaborate to accelerate AI and quantum deployment; researchers highlight cross-institution efforts MIT News (AI) source ↗

Sep 2 CW-Net explains autonomous vehicle behavior; technique helps predict when self-driving cars will make mistakes MIT News (AI) source ↗

Sep 2 Assessing suicide risk in Arabic crisis helpline calls: a comparison of Arabic and English large language models arXiv cs.CL source ↗

Sep 2 Recursive criticality of AI self-improvement arXiv cs.AI source ↗

Sep 2 Auditing harness tampering in self-improving agents arXiv cs.CL source ↗

Sep 2 LOOMSUM: a training-free framework to weave quantitative and narrative evidence for faithful long text-table summarization arXiv cs.CL source ↗

Sep 2 Zero-shot respiratory sound classification achieved by aligning self-supervised encoders with medical terminology using an LLM-augmented audio-text framework arXiv cs.CL source ↗

Sep 2 Generative artificial intelligence for reliable mechanistic reasoning for corrosion arXiv cs.LG source ↗

Sep 2 MiNER: fine-tuned biomedical NLP for malaria disease entity recognition in clinical texts arXiv cs.AI source ↗

Sep 2 ValueGraph: value-signal guided graph pre-training for contextualized user representation arXiv cs.CL source ↗

Sep 2 RW-LoRA: Communication-Efficient Decentralized LoRA fine-tuning via random walks arXiv cs.LG source ↗

Sep 2 Geometry-aware latent autoregressive generative model for PDEs in complex domains arXiv cs.LG source ↗

Sep 2 Stochastic complexity of vectors containing cluster structure arXiv cs.LG source ↗

Sep 2 Lingua franca or probing artifact? Rethinking latent language in multilingual LLMs arXiv cs.CL source ↗

Sep 2 QTEA: ternary LLMs with sparse residual salient weight and by-column optimization arXiv cs.LG source ↗

Sep 2 AI Morbidity and Mortality: a framework for clinical AI failure review arXiv cs.AI source ↗

Sep 2 ReNFT: repair mode collapse in reward post-training by internal probability-mass recalibration arXiv cs.LG source ↗

Sep 2 LLM-assisted, cost-efficient, zero-shot scientific entity linking bridges lexical divergence arXiv cs.CL source ↗

Sep 2 Synthetic Worlds for Temporal Evaluation and Knowledge Updating in LLMs arXiv cs.CL source ↗

Sep 2 Invalidation contracts attach version stamps to cached recovery suggestions to prevent silent failures from data drift in cross-episode LLM agent memory arXiv cs.AI source ↗

Sep 2 Counterfactual fragility certificates reveal high-confidence brittleness when structured evidence fails arXiv cs.LG source ↗

Sep 2 ReDeck: Step-Level Render-Grounded Refinement for Document-to-Slide Generation arXiv cs.AI source ↗

Sep 2 Behaviorally grounded user profiles from the wild for personalized alignment and multi-perspective reasoning arXiv cs.CL source ↗

Sep 2 Foundation models for electricity price forecasting and battery arbitrage: can they replace market-specific forecasting models? arXiv cs.LG source ↗

Sep 2 KItCAT: Knowledge injection via input corruption for auto-regressive training arXiv cs.CL source ↗

Sep 2 Life Operators: a self-evolving framework for multiscale life modelling arXiv cs.CL source ↗

Sep 2 Do LLMs know your neighborhood? Auditing LLM priors for neighborhood-level mobility prediction and structural alignment arXiv cs.LG source ↗

Sep 2 Flawed in nature, perfect through evolution arXiv cs.LG source ↗

Sep 2 Do general NLP embeddings capture ontological reasoning? arXiv cs.CL source ↗

Sep 2 Neural means and kernel corrections for operator learning arXiv cs.LG source ↗

Sep 2 Local reference geometry residual augmentation for imbalanced time series classification arXiv cs.LG source ↗

Sep 2 Do multimodal LLMs see before they read? Diagnosing contextual sycophancy in multimodal reasoning arXiv cs.CL source ↗

Sep 2 Prediction error alone may misalign with causal estimator performance across models; study compares OLS, GAMs, XGBoost, and DML-XGBoost in a partially linear model using simulations arXiv cs.AI source ↗

Sep 2 LLM-driven autonomous vehicles inherit human driver biases in pedestrian yielding; results and implications from a new benchmark arXiv cs.AI source ↗

Sep 2 Different representation learning objectives recover distinct latent structures from the same psychometric data arXiv cs.AI source ↗

Sep 2 Discrete-Time MDP modeling for multi-item capacitated lot sizing with stochastic demand timing arXiv cs.AI source ↗

Sep 2 Adapting without gradients: affine statistics transport and what its certificate can tell you arXiv cs.LG source ↗

Sep 2 OCGQuant: Outlier-Companion Grouping for NVFP4 Quantization arXiv cs.CL source ↗

Sep 2 Long-horizon state tracking in LLMs: executing MD5 through a deep sequence of dependent tool calls arXiv cs.AI source ↗

Sep 2 Elite-weighted supervised fine-tuning for goal-directed molecular optimization arXiv cs.LG source ↗

Sep 2 I-CARE: analysis of interference-related phenomena in a controllable, diverse and representative unlearning setting for text-to-image models arXiv cs.AI source ↗

Sep 2 Hypotheses-guided self distillation for continual personalization arXiv cs.AI source ↗

Sep 2 trajectory-judge: outcome-only LLM judges miss misalignment in agent trajectories arXiv cs.CL source ↗

Sep 2 Faster Than Flash: Exploiting attention sparsity for efficient long-context decoding arXiv cs.LG source ↗

Sep 2 RePro: proof-verified benchmark rewriting for reliable evaluation of LLM mathematical problem solving arXiv cs.CL source ↗

Sep 2 Retrieval, scoring, and decoding shape performance and stability in LLM-based conversational recommendation arXiv cs.CL source ↗

Sep 2 Good Memory Has ECC: Evaluating the memory of vision-language models beyond accuracy arXiv cs.LG source ↗

Sep 2 Convergence issues in Relational Concept Analysis based on AOC-posets arXiv cs.LG source ↗

Sep 2 From detection to refusal: safer LLMs via circuit-guided weight scaling arXiv cs.CL source ↗

Sep 2 AI should not only be helpful. It should be contingent; artificial intimacy, sycophancy, and the future of social learning arXiv cs.AI source ↗

Sep 2 Deterministic LLM inference across GPU kernels; a tolerance-based conformance suite detects only minimal output differences in an INT8 Qwen3-1.7B pipeline with injected faults arXiv cs.LG source ↗

Sep 2 CUDA-Harness enables agentic CUDA kernel generation and optimization from natural language arXiv cs.CL source ↗

Sep 2 Attention is the interaction map for scalable interaction-aware world model training arXiv cs.AI source ↗

Sep 2 LLM-as-a-demographic: sociodemographic prompting aligns or misaligns judgments with different demographic groups arXiv cs.CL source ↗

Sep 2 OpenAgentFlow: enabling system-wide safety boundaries for heterogeneous AI agent fleets arXiv cs.AI source ↗

Sep 2 Attention sensitivity is not enough: dissociating attention-level and behavioural in-context learning under fine-tuning. arXiv cs.LG source ↗

Sep 2 HyperWorld: hypergraph-structured state serialization improves learned textual world models arXiv cs.AI source ↗

Sep 2 Beneath the Diff: diagnosing and mitigating algorithmic mode collapse in code-level autonomous research loops arXiv cs.CL source ↗

Sep 2 Safin-1: Safety from within through memory-native state evolution arXiv cs.LG source ↗

Sep 2 Incremental risk assessment of progressive elder financial scams via instruction-tuned small language models arXiv cs.AI source ↗

Sep 2 DISTAL: Distillation and self-supervised pretraining for structure-agnostic materials property prediction arXiv cs.LG source ↗

Sep 2 SCAFFOLD: a large-scale structured dataset of computer science research figures with diagram QA and chain-of-thought reasoning traces arXiv cs.AI source ↗

Sep 2 UI-Venus-2: a general-purpose foundation GUI agent for mobile, web, and desktop environments arXiv cs.AI source ↗

Sep 2 Asymmetries in spontaneous and instructed deception in Llama-3.1-70B-Instruct arXiv cs.AI source ↗

Sep 2 Assessing alignment and stability of feature importance explanations via weight of evidence arXiv cs.LG source ↗

Sep 2 EULER: exploring underused links with evidence-checked return for multi-agent mathematical discovery arXiv cs.AI source ↗

Sep 2 Task-specific prompt with global context for multi-task graph pre-training arXiv cs.LG source ↗

Sep 2 Authority bias in conversational search engines for academic paper recommendation arXiv cs.AI source ↗

Sep 2 FAIRY: a full-stack smart-agriculture agent system deployed to a soybean research farm for end-to-end spatiotemporal farm operations arXiv cs.AI source ↗

Sep 2 Learning what to retain: gated-memory routing for efficient collaboration in multi-agent LLM systems arXiv cs.AI source ↗

Sep 2 REAL-Q: E2E LLM quantization via dynamic gradient descent arXiv cs.LG source ↗

Sep 2 Medical causal hypothesis verification with large language models arXiv cs.CL source ↗

Sep 2 WHALE: a simple recipe for joint harness-weight optimization arXiv cs.LG source ↗

Sep 1 Walter Torous named executive director of MIT Center for Real Estate MIT News (AI) source ↗

Sep 1 Mapping global methane emissions from space with deep learning Google Research source ↗

Sep 1 Context localization for generalized level-based evaluation in knowledge-based systems arXiv cs.AI source ↗

Sep 1 Nemotron 3.5 Content Safety Moderator: a compact multimodal, multilingual, reasoning-enabled content safety moderator arXiv cs.AI source ↗

Sep 1 Generative AI expands the intellectual reach of course-based undergraduate research experiences (CUREs). arXiv cs.AI source ↗

Sep 1 LongGuard analyzes and mitigates long-context guardrail failures in LLM safety; a training-free framework evaluates guardrails across long contexts arXiv cs.AI source ↗

Sep 1 CareGraph: an auditable hybrid AI framework for evidence-grounded personalized longitudinal health intelligence arXiv cs.AI source ↗

Sep 1 Why Didn’t It Check? Unsupported final claims and their repair in two tool-equipped language models arXiv cs.AI source ↗

Sep 1 Object-centric predictive monitoring framework for collaborative processes revealed on arXiv arXiv cs.AI source ↗

Sep 1 Probing perceptual priors of MLLMs via Gibbs sampling with interpretable generative controls arXiv cs.AI source ↗

Sep 1 Thinking costs tokens: adding inference structure hurts performance below a token-budget threshold and helps above it arXiv cs.AI source ↗

Sep 1 Benchmarking general mobile assistants in challenging real-world scenarios arXiv cs.AI source ↗

Sep 1 IoT-enabled acoustic monitoring with deep learning for early detection and severity assessment of ULWT infestations in tea plantations arXiv cs.AI source ↗

Sep 1 LLM-augmented causal discovery combines edge existence and orientation using Probabilistic Dependency Graphs to fuse Bayesian network structure learning with LLM knowledge arXiv cs.AI source ↗

Sep 1 PCFBench: a diagnostic benchmark for product carbon footprint estimation arXiv cs.AI source ↗

Sep 1 Rating the raters: Rasch measurement theory for LLM evaluation arXiv cs.AI source ↗

Sep 1 WM-R1 trains mobile GUI agents with world models instead of real environments for reinforcement learning arXiv cs.AI source ↗

Sep 1 Out-of-band policy enforcement at a trusted tool boundary aims to limit agent actions when given human credentials arXiv cs.AI source ↗

Sep 1 SETU: an agentic ecosystem for multilingual, persona-aware communication coaching arXiv cs.AI source ↗

Sep 1 Retrieving relations, detecting fallacies: a rag approach to political debate analysis arXiv cs.AI source ↗

Sep 1 Hypothesize, Evaluate, Refine for PDE Discovery in heterogeneous media with unknown spatial coefficient fields arXiv cs.AI source ↗

Sep 1 Credo: reusable declarative primitives for agentic workflows arXiv cs.AI source ↗

Sep 1 ReToolSQL: reinforcement learning from execution feedback with iterative refinement for text-to-SQL arXiv cs.AI source ↗

Sep 1 Not all explanations are sought: information-seeking psychology for human-centered XAI arXiv cs.AI source ↗

Sep 1 Agents for everyone: a workshop framework for building agentic AI capabilities in a distributed curation community arXiv cs.AI source ↗

Sep 1 Class-based heuristic selection for solving the flying block puzzle arXiv cs.AI source ↗

Sep 1 Time Capsule of testable human knowledge: 41 years of Jeopardy! in a single free local model arXiv cs.AI source ↗

Sep 1 Conservative hybrid graph networks for process systems with learned routing arXiv cs.LG source ↗

Sep 1 Equivariant Sheaf Neural Networks: learning geometric transport on graphs arXiv cs.LG source ↗

Sep 1 Learning-Theoretic Foundation for General Coded Computing: The Straggler Setting arXiv cs.LG source ↗

Sep 1 NVE: a separability and coverage-aware internal validation metric for biclustering arXiv cs.LG source ↗

Sep 1 Continuity-free near-minimax leading-order regret for CVaR-UCBVI arXiv cs.LG source ↗

Sep 1 MA-RAG: multi-agent retrieval-augmented generation for query-driven summarization of longitudinal Parkinson's disease assessments arXiv cs.CL source ↗

Sep 1 Parametric multimodal user memory stores perceptual details beyond captions for personalized agents arXiv cs.CL source ↗

Sep 1 Cross lingual transfer in Tulu legal comprehension: script-dependent improvement and RAG-induced knowledge conflict arXiv cs.CL source ↗

Sep 1 NLP-driven knowledge extraction and thematic classification of translated ancient Indian medical texts arXiv cs.CL source ↗

Sep 1 No detectable change in side-level WER from prompt-level context in a production oral-history transcription tool arXiv cs.CL source ↗

Sep 1 Autonomous AI coding agent developed using Monte Carlo Tree Search and Gemini LLM frameworks arXiv cs.LG source ↗

Sep 1 Rigour-matched audit compares periodic-step layer skipping methods ConfLayers and SWIFT for efficient LLM inference; includes analysis of trained routing alternatives arXiv cs.CL source ↗

Sep 1 Can Large Language Models identify meaningful touchpoints in conversion attribution? arXiv cs.CL source ↗

Sep 1 Flow-JePA: flow matching for robust latent dynamics in JEPA world models arXiv cs.LG source ↗

Sep 1 Temperature-Adaptive Transformed Teacher Matching arXiv cs.LG source ↗

Sep 1 Hybrid semantic context-enhanced ensemble learning improves wind power ramp-event forecasting and uncertainty-aware evaluation. arXiv cs.LG source ↗

Sep 1 The Halt Vector internalizes a causal steering intervention to halt chain-of-thought reasoning for more efficient reasoning arXiv cs.LG source ↗

Sep 1 ERR+: Sequential entropy resolution for efficient and decisive LLM reasoning arXiv cs.LG source ↗

Sep 1 Do MLLMs really understand low-resource Khmer documents? A pilot study on Khmer document VQA arXiv cs.CL source ↗

Sep 1 Latent-space intervention reduces cross-lingual factual inconsistency without sacrificing accuracy in QA prompts. arXiv cs.CL source ↗

Sep 1 Context-aware interpretable representations for retrieval and graph convolutional network classification arXiv cs.LG source ↗

Sep 1 Do large language models scrutinise what they review? A multimodal audit of scoring calibration, error detection, and author-identity effects arXiv cs.CL source ↗

Sep 1 Titans-QFWP: a regime-aware hybrid quantum fast weight programmer for portfolio optimization arXiv cs.LG source ↗

Sep 1 CLAIMPROBE audit reveals deep research writing omits key evidence and misattributes claims despite rubric scores arXiv cs.CL source ↗

Sep 1 Revisiting the provable-auditable privacy gap of DP-SGD arXiv cs.LG source ↗

Sep 1 Off-policy evaluation for semantic ID recommenders: can the model’s own SID tree serve as the action abstraction for OPE? arXiv cs.LG source ↗

Sep 1 From GenAI virtual patient dialogue logs to teacher-interpretable process evidence: a learning analytics study in higher education arXiv cs.CL source ↗

Sep 1 PromptKWS: a prompt-guided open-vocabulary keyword spotting framework introduces the Prompt Phrases Prediction Network for embedding keyword prompts arXiv cs.CL source ↗

Sep 1 ReVA: a region-aware visual assistant for visually grounded question answering arXiv cs.CL source ↗

Sep 1 Asymmetric within-document predictive learning for scientific document representation using title-, abstract-, method-, and conclusion-level predictions with SciJEPA arXiv cs.CL source ↗

Sep 1 Sparse Koopman autoencoders identify local dynamical regimes in multibasin systems arXiv cs.LG source ↗

Sep 1 STAGEET: Stage-wise Typed Edit Tagging for grammatical error correction with Arabic as a case study arXiv cs.CL source ↗

Sep 1 Test-time scaling improves LLM-driven automated equation discovery by enabling iterative search with additional compute at test time. arXiv cs.CL source ↗

Sep 1 GreenBench benchmarks energy efficiency and carbon footprint of OpenAI-like LLM inference on Apple Silicon arXiv cs.CL source ↗

Sep 1 Unsupervised latent space alignment with hyperspherical geodesic matching arXiv cs.LG source ↗

Sep 1 V2TATC: a joint voice-trajectory embedding framework and dataset for air traffic controller situational awareness arXiv cs.LG source ↗

Sep 1 From the loss landscape to diverse feature learning in neural networks arXiv cs.LG source ↗

Sep 1 RankShift detects and explains in-database shifts in category shares without changing overall event counts arXiv cs.LG source ↗

Sep 1 Curvature cryptanalysis of smooth transformer feed-forward networks arXiv cs.LG source ↗

Sep 1 Evaluating multilingual sentence embeddings for translation error detection: English–Greek contrastive study arXiv cs.CL source ↗

Sep 1 Terminal-Bench-LILT proposes a multilingual coding benchmark with 300 tasks across ten languages to evaluate non-English software development challenges. arXiv cs.CL source ↗

Sep 1 PAUSE: Editable strategy artifacts for long-form cultural story adaptation arXiv cs.CL source ↗

Sep 1 Enabling proactive spoken turns via a generalized style-aware full-duplex framework. arXiv cs.CL source ↗

Sep 1 Gurukul AI: an interactive AI-driven educational platform aligned with Indian syllabi and regional curricula arXiv cs.CL source ↗

Sep 1 MIT PhD student Ila Kumar works with youth from the child welfare system to reimagine how technology can support healing, connection, and independence MIT News (AI) source ↗

Sep 1 Using generative AI to design accessible interactive visualizations for undergraduate mathematics through a six-phase workflow arXiv cs.AI source ↗

Sep 1 From question-first to analyst-first: domain-expert skills and verified knowledge compilation for proactive enterprise analytics arXiv cs.AI source ↗

Sep 1 RegDivergence-101: an LLM benchmark for cross-jurisdiction regulatory contradiction detection in life sciences arXiv cs.AI source ↗

Sep 1 AutoScientist-Quant: self-evolving coding agents for automatic research in quantitative investment arXiv cs.AI source ↗

Sep 1 Race between agentic AI capabilities and data quality control in online surveys arXiv cs.AI source ↗

Sep 1 Paper Pilot: a human-in-the-loop expert system for evidence-traceable scientific manuscript generation in applied sciences arXiv cs.AI source ↗

Sep 1 CrossAudit: a Git-native, cross-vendor audit loop for agentic science arXiv cs.AI source ↗

Sep 1 AIMC: visual analytics for human oversight of autonomous scientific discovery arXiv cs.AI source ↗

Sep 1 A collective capability boundary in frontier large language models on guideline-conformant and case-specific oncology decision-making arXiv cs.AI source ↗

Sep 1 Preference elicitation over allocation outcomes for policy optimization to align heart transplantation with human values, via learning a utility function from stakeholder preferences arXiv cs.AI source ↗

Sep 1 CDPR: Counterfactual Advantage-based Credit Assignment for cost-aware sequential medical diagnosis arXiv cs.AI source ↗

Sep 1 The signal in the noise: an auditable reliability layer for biomedical text classification arXiv cs.AI source ↗

Sep 1 MedTVL: harnessing vision and language for medical time series classification arXiv cs.AI source ↗

Sep 1 C3-UniMM: causal cycle-consistent unified multimodal modeling via super alignment and shared decoding space arXiv cs.AI source ↗

Sep 1 InternReviewer and InternAdvocate: objective reward and evaluation for agentic reinforcement learning in peer review and rebuttal arXiv cs.AI source ↗

Sep 1 Expert-validated STEM QA arXiv cs.AI source ↗

Sep 1 CDEP Agent links meteorologically detected CDEP events to real-world documentary evidence arXiv cs.AI source ↗

Sep 1 Self-evolving skills via surrogate-guided solve-and-reproduce arXiv cs.AI source ↗

Sep 1 DS-Lighting: Making agent harnesses explicit for data-science automation arXiv cs.AI source ↗

Sep 1 Reward-Oracle MCTS for formal theorem proving: sample-efficient search and the need for kernel-level proof auditing arXiv cs.AI source ↗

Sep 1 SHAPE analyzes chain-of-thought trajectories in math reasoning using semantic spaces and steps through mathematical interpretations arXiv cs.AI source ↗

Sep 1 TPvG: A Moral Decision Framework for Large Language Models from One-Shot to Sequential Feedback arXiv cs.AI source ↗

Sep 1 Statutory AI: aligning large language models with legal norms arXiv cs.AI source ↗

Sep 1 Integrating triaxial IMU sensors and ensemble learning for Parkinson disease severity classification arXiv cs.AI source ↗

← 2026-10

2026-08 →