LIVE · refreshes every 20 min

updated Sep 16, 04:23 UTC

110010.art / archive / research / 2026-08

Research — August 2026

Aug 19 PIKFNO: an interpretable neural operator that uses physics informed kernel functions derived from governing equations to constrain the trunk network. arXiv cs.LG source ↗

Aug 19 DUET: Dual-Teacher On-Policy Distillation via Same-Weight Disagreement for Prohibition Compliance arXiv cs.LG source ↗

Aug 19 Pushing the limits of high-resolution weather forecasting through data scaling arXiv cs.LG source ↗

Aug 19 Early cycle charge trajectory generative prediction and full life cycle health management of iron-chromium flow batteries using FlowBD-E1 arXiv cs.LG source ↗

Aug 19 Randomly initialized autoencoders: fixed points and edge-of-chaos arXiv cs.LG source ↗

Aug 19 DumpsterCluster: using retired GPUs to serve LLaMA-70B on 60 GPUs arXiv cs.LG source ↗

Aug 19 Efficient neural-network-based high-resolution radiative transfer for CO2 retrieval, and application to interferometric sensing arXiv cs.LG source ↗

Aug 19 iFuzz-Meta: an interpretable fuzzy rule-based learning framework bridging top-down and bottom-up knowledge integration arXiv cs.LG source ↗

Aug 19 Training and evaluating ethical reinforcement learning agents on per-episode distributions arXiv cs.LG source ↗

Aug 19 Coarse-to-fine multi-resolution diffusion models for trajectory generation in urban systems arXiv cs.LG source ↗

Aug 19 Forward Pass Domain Adaptation (Without Cross-Layer Backpropagation) achieves 2.7–3.2x throughput and ~40% lower peak memory for fine-tuning without backward passes; off-domain benchmarks remain within seed-noise of baseline. arXiv cs.LG source ↗

Aug 19 Valid per-field selective risk control for document extraction; three failure modes, a validity ladder, and when conditioning pays arXiv cs.LG source ↗

Aug 19 Discrete diffusion language models are training-free multi-label classifiers arXiv cs.LG source ↗

Aug 19 Metaplasticity as adaptive gradient preconditioning for incremental learning arXiv cs.LG source ↗

Aug 19 Explanations using Alternative Realities for Reinforcement Learning (EARL) introduced for self-adaptive systems arXiv cs.LG source ↗

Aug 19 Learning discrete Riemannian metrics for physical fields with cochain-frame equivariance arXiv cs.LG source ↗

Aug 19 Geometry Is Not Robustness: a trajectory-level study of PGD evaluation arXiv cs.LG source ↗

Aug 19 Calibrated trust, not sharper prediction, shown in uncertainty fusion for legal outcome predictions arXiv cs.LG source ↗

Aug 19 Paired exact-reset evaluation of a prediction-derived medium-to-full world-model cascade arXiv cs.LG source ↗

Aug 19 Fractional optimizers meet fractal activation functions: an empirical study of multi-scale optimization in neural networks arXiv cs.LG source ↗

Aug 19 BDIP-Net: dual-interaction graph learning for property prediction of bilayer materials arXiv cs.LG source ↗

Aug 19 LiveHouse-TS: an open-world living benchmark for time series foundation models arXiv cs.AI source ↗

Aug 19 KernelArc: a multi-agent framework for autonomous GPU kernel optimization across heterogeneous workloads arXiv cs.AI source ↗

Aug 19 Fool's Gold: Defensive deception against safety-removal attacks on open-weight models arXiv cs.AI source ↗

Aug 19 Institution-specific LLM prompting recovers PHI that de-identification systems and their gold standards miss arXiv cs.CL source ↗

Aug 19 Token optimization and context-window management in multi-agent AI workflows arXiv cs.CL source ↗

Aug 19 Synthesizing feature extractors: an agentic approach for algorithm selection. arXiv cs.AI source ↗

Aug 19 TileMix: Tile-Centric mixed-precision attention for LLM inference acceleration arXiv cs.AI source ↗

Aug 19 Can LLMs reason in a legally meaningful manner? A small-scale study on European Court of Human Rights cases arXiv cs.CL source ↗

Aug 19 Margin-regularized structured semantic alignment for brain-language correspondence arXiv cs.CL source ↗

Aug 19 Temporal leakage in financial news NLP: a multi-architecture audit with a regime-specific M&A signal arXiv cs.CL source ↗

Aug 19 KnowSim: evaluating information calibration in LLM assistants with user simulators that learn arXiv cs.AI source ↗

Aug 19 Wuying-Browser-Agent: real-world centric, long-horizon browser agents for sustained decision making on live websites arXiv cs.AI source ↗

Aug 19 Which source wins? Task-dependent reliance in vision-language models arXiv cs.CL source ↗

Aug 19 ICD-Deepre predicts future diagnosis codes from longitudinal records using structured EHR foundation models and language models arXiv cs.CL source ↗

Aug 19 ArguLens: an open-source system for automated essay scoring and label-aware feedback generation arXiv cs.CL source ↗

Aug 19 From entity mentions to tone: an LLM-based pipeline for media bias analysis arXiv cs.CL source ↗

Aug 19 Polaris: learning to generate table descriptions from retrieval feedback arXiv cs.CL source ↗

Aug 19 LLMs for medical consultation evaluated after problem framing; study analyzes three API models across multi-turn vignettes and standardized-patient simulations arXiv cs.AI source ↗

Aug 19 AISA: AI Safety Assistant Framework for Continuous Improvement of Highway Construction arXiv cs.CL source ↗

Aug 19 The plot thins: uniformity and linearity in literary summaries arXiv cs.CL source ↗

Aug 19 Explicit state elicitation is not enough: a controlled audit of memory-policy classification arXiv cs.AI source ↗

Aug 19 Benchmarking the Benchmarks: evaluating automated safety benchmarks for small language models arXiv cs.AI source ↗

Aug 19 DeAR: decentralized agentic reasoning via capability grounding and collaborative thought navigation arXiv cs.AI source ↗

Aug 19 GxP-Agent: Process-DAG topologies for reliable clinical trial programming with LLM agents arXiv cs.AI source ↗

Aug 19 Cross-model memory transfer via target-side reader adaptation arXiv cs.CL source ↗

Aug 19 There is no theoretical curse of multilinguality for embedding space structure. arXiv cs.CL source ↗

Aug 19 PlanPO: Group planning-aware policy optimization for multi-turn agentic LLMs arXiv cs.AI source ↗

Aug 19 Logit-based energy scoring outperforms prompted LLM-as-judge for ranking scientific hypotheses arXiv cs.AI source ↗

Aug 19 Towards Safer RAG: only agents capable of System 2 thinking may access untrusted documents arXiv cs.CL source ↗

Aug 19 Memory Is Communication: allocation of memory and messaging under resource limits in bounded agents arXiv cs.AI source ↗

Aug 19 DiSCO: Defending text-to-image generation through distribution-guided contrastive prompt optimization arXiv cs.AI source ↗

Aug 19 ArborMem: navigating interaction states with memory forests arXiv cs.CL source ↗

Aug 19 Emotion Across Speech and Faces: Shared Affective Mechanisms in Multimodal Foundation Models arXiv cs.CL source ↗

Aug 19 SignalReasoner: assessing the upper bound of 3B models for signal mathematical reasoning arXiv cs.AI source ↗

Aug 19 A glyph is not a letter, a token is not a word, and a space is not a space in Voynichese units; study tests three assumptions against Zandbergen-Landini transliteration arXiv cs.CL source ↗

Aug 19 FedPref: federated preference learning for structured radiology report extraction arXiv cs.AI source ↗

Aug 19 Q-Interference: memory-efficient phase-aware quantum-inspired attention for autoregressive language modeling arXiv cs.CL source ↗

Aug 19 The problem is the problem: towards scalable mathematical discovery arXiv cs.AI source ↗

Aug 19 The price of thinking: reasoning effort as a model-specific API contract arXiv cs.AI source ↗

Aug 19 PTXBench benchmarks and adapts LLMs to architecture-specific PTX for GPU kernel optimization, measuring functional correctness and speedup on GEMM and attention workloads across H100 and B200 GPUs. arXiv cs.CL source ↗

Aug 19 Toward personal intelligence through cooperative observation arXiv cs.AI source ↗

Aug 19 Children accelerate word learning while language models do not arXiv cs.CL source ↗

Aug 19 SkillEffect: Checked lowering for memory-bounded agent tools arXiv cs.AI source ↗

Aug 19 ASI-Bench: at the dawn of artificial superintelligence arXiv cs.AI source ↗

Aug 18 Study finds generated AI art often cannot be traced to training data when authors are removed from the dataset MIT News (AI) source ↗

Aug 18 TeachMateGPT: a multi-agent knowledge-grounded framework for pedagogical assessment generation from science curriculum materials arXiv cs.CL source ↗

Aug 18 Jais 2 is a family of Arabic-centric large language models developed by MBZUAI, Cerebras, and Inception, including a 70B open LLM and an 8B-parameter variant. arXiv cs.CL source ↗

Aug 18 StreamHear: domain-adapted pseudo-labeling for semi-supervised streaming speech recognition arXiv cs.CL source ↗

Aug 18 Legal RAG systems still hallucinate, study finds across eight systems on English GDPR and French civil-law corpora arXiv cs.CL source ↗

Aug 18 Repair, not improvement: decomposing constrained decoding in tool-call abstention arXiv cs.CL source ↗

Aug 18 When lexical change misleads: evaluating dynamic topic models with traditional and LLM-based metrics across NYT, DBLP, and arXiv topics arXiv cs.CL source ↗

Aug 18 S2Dialog: Multimodal dialogue retrieval with semantic and acoustic-style modeling arXiv cs.CL source ↗

Aug 18 The conditional superiority of fast silicon sampling arXiv cs.CL source ↗

Aug 18 Leading-Silence augmentation and multi-stage synthetic supervision used for the second MLC-SLM challenge arXiv cs.CL source ↗

Aug 18 Not all tokens are equal: inflation-aware routing for agentic LLM systems arXiv cs.CL source ↗

Aug 18 Does a Language Server save tokens for coding agents? A measurement methodology and preliminary study arXiv cs.CL source ↗

Aug 18 GALA: Generation-Aware Cross-Modal Alignment for Text-to-Time-Series Synthesis arXiv cs.CL source ↗

Aug 18 Scaling creative writing beyond story-centric data with attribute-guided genre expansion arXiv cs.CL source ↗

Aug 18 Amplified does not mean predictive: reasoning behaviors in thinking models arXiv cs.CL source ↗

Aug 18 GRPO beyond English: a large-scale study of GRPO in non-English and multilingual settings arXiv cs.CL source ↗

Aug 18 Bootstrapping niche multilingual code translation via reinforcement learning with execution-based verifiable supervision arXiv cs.CL source ↗

Aug 18 IterCOMP: reasoning-aware adaptive prompt compression for multi-hop question answering arXiv cs.CL source ↗

Aug 18 Blockwise Causal Memory Transformer aims to model long-context language with local block attention and block-level global context propagation arXiv cs.CL source ↗

Aug 18 CLAIR-Fin: an adversarial multi-agent framework for claim-level verification and adaptive debate in cross-modal financial QA arXiv cs.CL source ↗

Aug 18 ASSERT: A measurement pipeline for GenAI audits arXiv cs.CL source ↗

Aug 18 BM25-augmented many-shot translation for low-resource north-eastern Indian languages arXiv cs.CL source ↗

Aug 18 Measuring fairness in large audio language models via semantic-aware bias estimation arXiv cs.CL source ↗

Aug 18 Think in latent, explain in language: self-explainable latent reasoning arXiv cs.CL source ↗

Aug 18 Medical AI neglects actual treatment outcomes in favor of opinions and syntheses, hindering its ability to predict treatment effectiveness. arXiv cs.AI source ↗

Aug 18 Longitudinal and graph-augmented prediction of adolescent substance use onset in the ABCD study. arXiv cs.AI source ↗

Aug 18 Do LLM agents negotiate rationally? A mechanism-design framework for verifiable multi-agent interaction over A2A/MCP arXiv cs.AI source ↗

Aug 18 FLOPs do not reliably predict execution time; replication is important for AI efficiency assessment arXiv cs.AI source ↗

Aug 18 Learning agent execution for KV-cache management in agentic serving arXiv cs.AI source ↗

Aug 18 MecEng benchmark measures large language models on creating multibody simulation models from parameterized text descriptions arXiv cs.AI source ↗

Aug 18 Global AI regulations for FAIR and ethics in high-risk use cases: a comparative review arXiv cs.AI source ↗

Aug 18 A human-centred approach to benchmarking LLMs for parenting advice arXiv cs.AI source ↗

Aug 18 Euclid-Omni: a unified neuro-symbolic framework for plane geometry arXiv cs.AI source ↗

Aug 18 Auxiliary uncertainty signals improve LLM-assisted systematic review screening; benchmark across eight Cohen drug-class reviews shows improved efficiency arXiv cs.CL source ↗

Aug 18 AI governance needs ISO-like interoperability protocols, not just laws arXiv cs.AI source ↗

Aug 18 Toward Safe LLM agents: a survey of specification, verification, and enforcement arXiv cs.AI source ↗

Aug 18 AI lock-in is in progress, and we must be prepared arXiv cs.AI source ↗

Aug 18 The Hallucination Snowball: modeling error propagation as state transitions in multi-agent LLM pipelines arXiv cs.AI source ↗

Aug 18 Agentic framework using rules and LLMs for embedding and annotating descriptive document layouts in plant science use case arXiv cs.AI source ↗

Aug 18 The Unwritten Benchmark: a new challenge for multimodal machine learning in abstract perceptual reasoning arXiv cs.AI source ↗

Aug 18 Characterizing rhetorical misalignment in decision-making with language models arXiv cs.CL source ↗

Aug 18 OGX: an open-source, vendor-neutral Generative AI application server implements APIs of major labs with pluggable backends arXiv cs.AI source ↗

Aug 18 SKILL: Self-correcting knowledge-guided iterative LLM agent for logic optimization arXiv cs.AI source ↗

Aug 18 Neural constraint reasoning should prioritize symbolic integration over pure learning to ensure certified correctness under distribution shifts. arXiv cs.AI source ↗

Aug 18 From Doyle to AGM: a survey and an implementation roadmap for belief change arXiv cs.AI source ↗

Aug 18 When do LLMs apply the wrong law? Diagnosing temporal applicable-law failures in legal reasoning arXiv cs.AI source ↗

Aug 18 Want better ML reviews? switch to a credit-based incentivization system arXiv cs.AI source ↗

Aug 18 Evaluations of AI moral reasoning still miss half the picture arXiv cs.AI source ↗

Aug 18 When to communicate: belief distributions and KL divergence for principled gating in multi-agent RL arXiv cs.AI source ↗

Aug 18 Which question is your attention metric answering? attention rows as compositional data arXiv cs.CL source ↗

Aug 18 Wiola 13M, a gated spiral attention architecture for parameter efficient small language models arXiv cs.CL source ↗

Aug 18 DeMTS: denoising trajectories as multivariate time series for hallucination detection in diffusion language models arXiv cs.CL source ↗

Aug 18 Automatic or controlled? Repetition priming reveals divergent processing in base LLMs, Instruct LLMs, and humans arXiv cs.CL source ↗

Aug 18 Multi-modal generative fuzzy system: fuzzy inference guides large model interactive question answering framework arXiv cs.CL source ↗

Aug 18 Evaluation of retrieval failure in RAG architectures for Bengali agricultural advisory; study analyzes query-type and language-condition gaps across 1,000 farmer queries and 2,882 knowledge nodes arXiv cs.CL source ↗

Aug 18 Training leaves traces: centered residual signatures for language model lineage verification arXiv cs.CL source ↗

Aug 18 Beyond the pale: assessing prevalence and contents of extremist speech in LLM training data arXiv cs.CL source ↗

Aug 18 HarmProfile benchmarks harmful outputs of frontier LLMs by characterizing distributions of misbehavior and related content. arXiv cs.CL source ↗

Aug 18 How Do Agents Fail on AutoResearch: End-to-End Diagnostic Evaluation on 100 Real-World Frontier Research Tasks arXiv cs.CL source ↗

Aug 18 What to Forget in Unlearning? Forget Set Curation for Language Models arXiv cs.CL source ↗

Aug 18 Beyond tokens: a survey on decoding methods for large language and vision-language models arXiv cs.CL source ↗

Aug 18 Interpretable cross-lingual alignment in small language models; probing cultural and pragmatic reasoning in Japanese-English bilingual LLMs arXiv cs.CL source ↗

Aug 18 Domain Agnostic text redaction from natural language rules using instruction tuning arXiv cs.CL source ↗

Aug 18 LLM safety alignment in low-resource languages: a systematic literature review arXiv cs.CL source ↗

Aug 18 RamseyGadgets: a graph construction dataset for LLMs arXiv cs.CL source ↗

Aug 18 Prompting is not enough: supervised baselines and leakage control for measuring shared decision-making with LLMs in pediatric encounters arXiv cs.CL source ↗

Aug 18 Writing style similarity reflects academic genealogy arXiv cs.CL source ↗

Aug 18 DA-RAC: distance-aware calibration of LLM judges for trustworthy AI auditing arXiv cs.CL source ↗

Aug 17 Rethinking how innovation happens: The Invisible Engine argues that innovation unfolds through the integration of market applications, technology, and implementation; Eugene Fitzgerald discusses how society should invest in it. MIT News (AI) source ↗

Aug 17 Agentao: a governed local-first runtime for tool-using LLM agents arXiv cs.AI source ↗

Aug 17 Training-free knowledge transfer across model scales through activation-guided pruning arXiv cs.LG source ↗

Aug 17 Depth-Aware sensitivity analysis of mixture-of-experts models via magnitude-based expert masking arXiv cs.AI source ↗

Aug 17 Don't claim benchmark-oriented optimization improves general coding capability; diverse evaluation is required arXiv cs.LG source ↗

Aug 17 Fashion outfit generation via unified sequential composition models arXiv cs.LG source ↗

Aug 17 Coverage aware active evaluation for failure discovery with paired systems arXiv cs.AI source ↗

Aug 17 Stable miscalibration in large language models: a practical view of high-confidence errors arXiv cs.AI source ↗

Aug 17 ConceptFlow: a scikit-learn-compatible Python library for Formal Concept Analysis builds nested line diagrams from many-valued contexts and outputs interactive visualizations arXiv cs.AI source ↗

Aug 17 Query-derived erase direction (QED) added to mitigate interference in linear attention reads arXiv cs.LG source ↗

Aug 17 Inducing reward-free judging rubrics that reduce over-crediting in agent evaluation arXiv cs.AI source ↗

Aug 17 ARC: Fair relative advantage comparison in open-ended real-world interaction arXiv cs.AI source ↗

Aug 17 Robust XGBoosting for regression arXiv cs.LG source ↗

Aug 17 Geometric filtering of LLM-generated samples for few-shot text classification arXiv cs.LG source ↗

Aug 17 Advances in deep learning-based drug-target binding affinity prediction arXiv cs.LG source ↗

Aug 17 HI-MeshGraphNets: Efficient and accurate mesh-based physics learning with hierarchical multi-scale graph neural networks arXiv cs.LG source ↗

Aug 17 Capacity-dependent effects of data selection for reasoning arXiv cs.LG source ↗

Aug 17 Robust dual-model collaborative random vector functional link network arXiv cs.LG source ↗

Aug 17 From BERT to frontier agents: eight years of language-model progress, the collapse of the capability-cost curve, and the rise of task-targeted models arXiv cs.LG source ↗

Aug 17 Localizing cross-kernel divergence in INT8-quantized LLM inference across CUTLASS and Triton kernels arXiv cs.LG source ↗

Aug 17 A year in LLM serving: workload evolution, caching and load-balancing arXiv cs.AI source ↗

Aug 17 Reward machines for signal temporal logic arXiv cs.AI source ↗

Aug 17 AI evaluation should focus on human–AI teams rather than autonomous superhuman performance arXiv cs.AI source ↗

Aug 17 Learning to assemble novel structures with unfamiliar parts under semantic constraints arXiv cs.AI source ↗

Aug 17 Active perception for embodied disambiguation arXiv cs.AI source ↗

Aug 17 Evaluating agentic learning harness capabilities without labels via the scaling hypothesis arXiv cs.AI source ↗

Aug 17 Dynamic multi-depot vehicle routing with online requests using event-driven transformer--DRL and rolling-horizon benchmarking arXiv cs.LG source ↗

Aug 17 Measuring cross-task behavioral consistency in language model agents. arXiv cs.AI source ↗

Aug 17 Algorithm design and physician liability for disparate AI clinical algorithms</analysis> arXiv cs.AI source ↗

Aug 17 Cross-disciplinary taxonomy and modeling of misunderstanding generation, amplification, and detection, from pragmatics to AI agents arXiv cs.AI source ↗

Aug 17 L-FNO: Lorentzian Fourier Neural Operator for stochastic event dynamics arXiv cs.LG source ↗

Aug 17 Federated prompt learning: a unified framework, empirical analysis, and future directions arXiv cs.LG source ↗

Aug 17 How compliant is sepsis treatment? An expert-guided neuro-symbolic pipeline for generating clinical compliance insights arXiv cs.AI source ↗

Aug 17 Contrastive learning for interpretable anomaly detection at collider experiments arXiv cs.LG source ↗

Aug 17 No universal signal predicts sample-level LLM regression under version updates arXiv cs.AI source ↗

Aug 17 CutClean: neural network pruning for privacy-preserving inference arXiv cs.LG source ↗

Aug 17 SAGE: Surrogate-gradient adaptation via attention-guided entropy for spiking transformers arXiv cs.LG source ↗

Aug 17 Second Thought: reasoning in parallel as LLM agents act and observe arXiv cs.AI source ↗

Aug 17 Modular cognitive architecture emerges in large language models arXiv cs.AI source ↗

Aug 17 PPAPlace: Differentiable cross-stage objectives for chip placement optimization arXiv cs.LG source ↗

Aug 17 EEG-PRISM: physiologically grounded interpretability of predictions by EEG foundation models arXiv cs.LG source ↗

Aug 17 Stochastic control policies for robust molecular transition path sampling. arXiv cs.LG source ↗

Aug 17 Your Probabilistic JEPA is secretly a hidden Markov model: a state-space interpretation of Joint-Embedding Predictive Learning arXiv cs.AI source ↗

Aug 17 SemPlan benchmarks structured semantic planning for LLM-based queries over enterprise data with a bilingual synthetic benchmark and a frozen scientific evaluation subset arXiv cs.AI source ↗

Aug 17 Hybrid Quantum-inspired Kolmogorov-Arnold networks for privacy-aware federated biosignal learning arXiv cs.LG source ↗

Aug 17 MobileMem: learning from a year of mobile experiences arXiv cs.AI source ↗

Aug 17 Variation Brownian Kernel Ladder introduces a path-atomic function-space framework that separates nonlinear recursive dictionary construction from linear variation, using Brownian RKHS unit-ball profiles to build VBKL space. arXiv cs.LG source ↗

Aug 17 MedMix: specialization-consistent federated sparse MoEs under modality heterogeneity arXiv cs.LG source ↗

Aug 17 Hard Cases, Bad Labels: Testing error exposure and error location in uncertainty sampling under bounded label noise arXiv cs.LG source ↗

Aug 15 CAS: a causal attribution score for local and global explainable AI arXiv cs.AI source ↗

Aug 15 Auditable agentic AI coordinates diagnostic tools and stores outputs as auditable case-level evidence for thyroid ultrasound diagnosis and reporting arXiv cs.AI source ↗

Aug 15 Learning to adapt cross-domain preferences via meta-LoRA for LLM personalization arXiv cs.AI source ↗

Aug 15 Alignment methods can be misused for censorship and manipulation, as a position paper argues that modern AI alignment is dual-use and may provide malicious actors with better tools. arXiv cs.AI source ↗

Aug 15 AstraZeneca develops Research Assistant, an internal LLM-based system to explore biomedical questions across literature, knowledge graphs, chemistry, and clinical data arXiv cs.AI source ↗

Aug 15 Trie automata for constrained decoding over large finite sets arXiv cs.AI source ↗

Aug 15 SteerBench-Work: a benchmark for agent steering at action boundaries across workplace domains arXiv cs.AI source ↗

Aug 15 Dual-flow transformers decouple the primary prefill path from additional decode computation arXiv cs.AI source ↗

Aug 15 Governed Persistent Memory introduces a source-bound, auditable bitemporal state-transition model with fail-closed release for long-horizon agents arXiv cs.AI source ↗

Aug 15 Epsilon-MemEvo enables cross-task memory transfer in LLM program evolution by storing prior experience as task-agnostic tactic memories arXiv cs.AI source ↗

Aug 15 IntegrityBench benchmarks LLMs on misconduct classification, ethical action reasoning, and artifact-grounded decision making across 36 paired tasks under pressure; evaluates 18 frontier model variants arXiv cs.AI source ↗

Aug 15 Dead text or binding clause? Measuring and restoring constraint influence in black-box LLM dialogues arXiv cs.AI source ↗

Aug 15 MindMemOS: a portable and self-evolving memory operating layer for AI agents arXiv cs.AI source ↗

Aug 15 DiG-bench: discovery in games arXiv cs.AI source ↗

Aug 15 Jagged judges: Epistemic stability under silence, pressure, and persistence arXiv cs.AI source ↗

Aug 15 Designing AI pipelines for decision-ready ITSM intelligence arXiv cs.AI source ↗

Aug 15 Reasoning is a learnable rule-based process arXiv cs.AI source ↗

Aug 15 Asking an LLM in Japanese can change its willingness to advise a nuclear strike, study finds arXiv cs.AI source ↗

Aug 15 Multi-Agent Scheduling with LLM-assisted contract net negotiation for stream processing in mobile edge computing arXiv cs.AI source ↗

Aug 15 Agreement is not alignment: divergent moral grounds in human and LLM ethical judgments arXiv cs.AI source ↗

Aug 15 Probabilities of causation with causal knowledge tightened for binary PNS bounds by incorporating additional information arXiv cs.AI source ↗

Aug 15 Reasoning Jury: multi-model consensus for evaluating reasoning traces arXiv cs.AI source ↗

Aug 15 Attention is all you have; arXiv:2608.12610v1 reports that 56,804 public agent skills exist and installation bundles content, persistence, and trigger management, limiting long-tail usage. arXiv cs.AI source ↗

Aug 15 Large language models can follow instructions, but struggle with many constraints simultaneously; phase transitions in compositional constraint satisfaction arXiv cs.AI source ↗

Aug 14 InfraBench: evaluating infrastructure agents across layers, lifecycle, and risk; building infrastructure for financially autonomous AI agents arXiv cs.AI source ↗

Aug 14 RecSys Factory: bounding LLM agent autonomy to decision points in the industrial recommender lifecycle arXiv cs.AI source ↗

Aug 14 Towards query-agnostic RAG evaluation via query coverage and claim verifiability arXiv cs.AI source ↗

Aug 14 From monolithic to modular: segment-level automatic prompt optimization arXiv cs.AI source ↗

Aug 14 Cutting AI datacenter energy with reinforcement learning: measured power control of LLM training from one GPU to the fleet arXiv cs.AI source ↗

Aug 14 Detecting a route flip is easier than knowing whether to fix it; causal route-mediated damage in quantized mixture-of-experts arXiv cs.AI source ↗

Aug 14 Distribird automates literature-informed prior distribution design for Bayesian model calibration. arXiv cs.AI source ↗

Aug 14 VQ-bench: a composable vector quantization framework arXiv cs.AI source ↗

Aug 14 Harnessing agent memory to build lifelong AI partners for materials scientists arXiv cs.AI source ↗

Aug 14 Edge-based contiguous p-median problem with connections to logistics districting arXiv cs.AI source ↗

Aug 14 MaSRead addresses the read to content in replicated latent stores; content-addressed reading improves access in a conflict-free replicated cache arXiv cs.AI source ↗

Aug 14 Identity from the outside: a framework for AI personality clones and observed identity factors arXiv cs.AI source ↗

Aug 14 Geometry-aware Incremental Neural Operator for Long-Horizon PDE prediction arXiv cs.AI source ↗

Aug 14 Forced-Structure reduction shows no circulant graph on Z/99 satisfies more than 68.0% of constraints for Conway's 99-graph; verifiable bounds presented arXiv cs.AI source ↗

Aug 14 Poor Man's Agentic Modeling: simulating large LLM-agent societies on a laptop arXiv cs.AI source ↗

Aug 14 A Conceptual Framework for Refining Influence Knowledge from Simulation Evidence in Cyber-Physical Systems arXiv cs.AI source ↗

Aug 14 Synchronizing beliefs with second-order theory-of-mind in human-autonomy teams (extended version) arXiv cs.AI source ↗

Aug 14 Forecasting side effects of activation steering arXiv cs.AI source ↗

Aug 14 LinearKV: One cached state suffices for position-independent caching in hybrid LLMs arXiv cs.AI source ↗

Aug 14 AutoWorldModel-Bench: a state-centric benchmark for automated world-model research arXiv cs.AI source ↗

Aug 14 Dynamic governance of multi-LLM agent systems for collaborative conversational outcomes arXiv cs.AI source ↗

Aug 14 LLMs in process diagram engineering: from optimal PFDs to validated P&IDs arXiv cs.AI source ↗

Aug 14 BEST-KAG: Enhancing question answering of building engineering standards with multimodal knowledge graph modeling and large language model arXiv cs.AI source ↗

Aug 14 Off-support barrier: semantic safety constraints are not learning-problem invariants and implications for design, containment, and verification arXiv cs.AI source ↗

Aug 14 The Boolean power of ReLU: ReLU-MPLang expresses more Boolean queries than Σ-MPLang with eventually constant activations on finite graphs, resolving an open question arXiv cs.LG source ↗

Aug 14 Geometric and behavioral stratification in transformer residual streams arXiv cs.LG source ↗

Aug 14 LoKiFormer: Locality-aware attention with decoupled knowledge memory for efficient large language model pretraining arXiv cs.LG source ↗

Aug 14 Dual spatial-temporal attribution for architecture-aligned post-hoc explainability in recurrent graph anomaly detection arXiv cs.LG source ↗

Aug 14 Learning under treatment-induced label indeterminacy with expert annotations of counterfactual outcomes in neurological prognostication arXiv cs.LG source ↗

Aug 14 Scaling recurrent memory with content-routed state anchors arXiv cs.LG source ↗

Aug 14 Unifying generative models with path integrals arXiv cs.LG source ↗

Aug 14 Finding the Needle in a Haystack: Test-time analog circuit representation adaptation for Bayesian optimization arXiv cs.LG source ↗

Aug 14 GENADA: efficient generative time series adversarial attack framework arXiv cs.LG source ↗

Aug 14 Demand transfer estimation at scale via restricted logit modeling arXiv cs.LG source ↗

Aug 14 Prob-K: probabilistic one-pass filtering for efficient top-k selection arXiv cs.LG source ↗

Aug 14 Predicting when random low-dimensional reparameterizations train neural networks arXiv cs.LG source ↗

Aug 14 Free satellite signals indicate glacial-lake destabilization and timing of risk in Nepal Himalaya, using radar interferometry and weather data arXiv cs.LG source ↗

Aug 14 Impact of temporal context length and encoding strategies on self-supervised ECG representation learning arXiv cs.LG source ↗

Aug 14 Personalized scorer modeling: a learning-based framework for deriving robust sleep stage labels from multiple experts arXiv cs.LG source ↗

Aug 14 Multi-AUV ad-hoc network-based target tracking via a value gradient guidance multi-agent diffusion reinforcement learning approach arXiv cs.LG source ↗

Aug 14 Interpretable causal discovery via causal-effect constraints arXiv cs.LG source ↗

Aug 14 Training Under Challenge: Executable certificates and challenge-closed optimality for neural networks arXiv cs.LG source ↗

Aug 14 Represent, then generate: multimodal-conditioned time-series generation under irregular missingness arXiv cs.LG source ↗

Aug 14 Exemplar-based objective classification of gust-induced loads across multiple flight conditions arXiv cs.LG source ↗

Aug 14 Exploring oversmoothing with Householder matrices. arXiv cs.LG source ↗

Aug 14 When can you trust offline evaluation of equal-cost top-k allocation? a controlled, reproducible benchmark and practitioner's guide arXiv cs.LG source ↗

Aug 14 CAKE: compiler-agent co-design for frontier kernel evolution arXiv cs.LG source ↗

Aug 14 Structure-preserving uncertainty quantification for GENERIC dynamics arXiv cs.LG source ↗

Aug 13 Basin: efficient and extensible numerical optimization in Rust arXiv cs.LG source ↗

Aug 13 Gloss-free representation learning for cross-dataset sign spotting arXiv cs.CL source ↗

Aug 13 Dual-primal graph VAEs for noisy label aggregation arXiv cs.LG source ↗

Aug 13 Diffuse to compress: leveraging diffusion LMs for lossless compression arXiv cs.CL source ↗

Aug 13 Trajectory-adapted uncertainty quantification for LLM agents arXiv cs.CL source ↗

Aug 13 Click2Poly: a VLM for vector mapping of buildings and walls arXiv cs.LG source ↗

Aug 13 Weak reversible bisimilarities in CCSK: two variants (directional and mixed) proposed for tau actions; study differences among strong/weak and forward-only/reversible notions. arXiv cs.CL source ↗

Aug 13 Market-information-aware gated-LoRA of foundation models for transferable day-ahead electricity price forecasting arXiv cs.LG source ↗

Aug 13 Hybrid-Policy Self-Editing for Composable Unstructured Knowledge Editing arXiv cs.CL source ↗

Aug 13 XGBoost outperforms LSTM in forecasting transmitted heat energy in district heating systems, with lower computational cost and environmental impact. arXiv cs.LG source ↗

Aug 13 Learning to persuade exposes how easily LLMs abandon correct beliefs arXiv cs.CL source ↗

Aug 13 AutoGrable proposes a minimal table-to-graph abstraction where each row becomes a node; no trained graph model required arXiv cs.LG source ↗

Aug 13 PAC-Bayes beyond parameter space: behavioral equivalence, Z-information, and exact complexity decomposition arXiv cs.LG source ↗

Aug 13 PAIR: Pairwise-aware inclusion reweighting for adaptive rollout allocation in RLVR arXiv cs.LG source ↗

Aug 13 Federated learning for distributed CNC tool wear prediction arXiv cs.LG source ↗

Aug 13 DonorRank: learning-to-rank donor language suitability for low-resource cross-lingual speech recognition arXiv cs.CL source ↗

Aug 13 Convergence is likely learned during pretraining, with semantic convergence observed from the first alignment stage (instruction-tuning); output homogeneity traces back to base models arXiv cs.CL source ↗

Aug 13 Three tokens force exponential feature rank in nonnegative kernel attention arXiv cs.LG source ↗

Aug 13 Stigma and support in online sexual violence narratives on Reddit arXiv cs.CL source ↗

Aug 13 CT-ΔBench: a benchmark for longitudinal 3D medical imaging difference reporting with vision-language models arXiv cs.CL source ↗

Aug 13 Reoptimization algorithms for contextual bandits with knapsack constraints arXiv cs.LG source ↗

Aug 13 Principal Trait Analysis: deriving “skills” for human-AI collaboration arXiv cs.CL source ↗

Aug 13 TRACE Bench evaluates task-driven agentic roleplay using a fixed checklist and a user agent to reveal tested requirements, failures, and supporting dialogue evidence. arXiv cs.CL source ↗

Aug 13 Weightless fine-tuning uses logit-space transport to personalize LLMs without weight updates arXiv cs.LG source ↗

Aug 13 Unmasking Toxic Mimicry in medical offline reinforcement learning for ICU sepsis management via counterfactual clinical audits arXiv cs.LG source ↗

Aug 13 Self-Evolving Embodied Agents via Skill-Harness Evolution arXiv cs.CL source ↗

Aug 13 Easper: an accessible ASR pipeline for language documentation arXiv cs.CL source ↗

Aug 13 Mechanism design for generative engines: from exploitation toward win-win outcomes arXiv cs.LG source ↗

Aug 13 Backtrader-Bench benchmarks LLM agents on algorithmic trading with self-generated MCQs arXiv cs.CL source ↗

Aug 13 Programmatic skill learning best reduces agent cost, the arXiv paper argues, by treating skills as programs to improve cost efficiency. arXiv cs.CL source ↗

Aug 13 FarSky: Task-aware latent-space coupling for generative intra-hour solar forecasting arXiv cs.LG source ↗

Aug 13 Federated aggregation robustness evaluated across five methods, five datasets, five architectures, and four attack conditions including clean, sign-flipping, Gaussian, and BadNets (reconstructed 500-cell seed-1 evaluation). arXiv cs.LG source ↗

Aug 13 Auditing political alignment in LLMs: an Italian case-study arXiv cs.CL source ↗

Aug 13 Retrofitting recurrent depth into a pretrained language model improves iterative latent transitions at two parameter budgets; Qwen2.5-0.5B-Instruct uses a recurrent block with an identity-preserving loop. arXiv cs.CL source ↗

Aug 13 Physics-aware latent-space surrogates enable variational parameter calibration arXiv cs.LG source ↗

Aug 13 Group Alignment-Induced Sycophancy: A Two-Sided Evaluation of Steerable Pluralistic Alignment arXiv cs.CL source ↗

Aug 13 Measure, Don't Optimize: Forecasting Recovery in LLM Unlearning arXiv cs.CL source ↗

Aug 13 Semantic Lenia: emergence of homeostatic solitons within the semantic space of large language models arXiv cs.CL source ↗

Aug 13 The Wording Effect: rephrasing a problem with the same meaning can flip an LLM’s answer, causing bidirectional drift in benchmark performance. arXiv cs.CL source ↗

Aug 13 Diffusion-based data-driven assortment optimization arXiv cs.LG source ↗

Aug 13 Terminal symmetry as a decision resource: statewise refinement for anytime verified construction arXiv cs.LG source ↗

Aug 13 Multivariate outlier detection methods tested on district heating system data, including temperature and energy transmission indicators arXiv cs.LG source ↗

Aug 13 Lifecycle-Optimal Tokenization: vocabulary size is deployment-regime dependent infrastructure parameter arXiv cs.LG source ↗

Aug 13 Contextual quality-diversity evolutionary reinforcement learning for HVAC control in tropical commercial buildings arXiv cs.LG source ↗

Aug 13 Self-Fix Step-DPO strengthens step-level reasoning for self-correction in LLMs using reinforcement learning arXiv cs.CL source ↗

Aug 13 ODE-based transformer decoders improve iterative sign language translation without increasing model size arXiv cs.CL source ↗

Aug 13 Evaluating side-constraint loss under context compaction; COMPINT assesses how session constraints are dropped during context compression arXiv cs.CL source ↗

Aug 13 Dynamics models for offline hyperparameter selection in real-world RL arXiv cs.LG source ↗

Aug 13 Long-horizon forecasting of complete financial statements with Forma arXiv cs.LG source ↗

Aug 12 MindTopo reveals VLMs’ spatial reasoning abilities Microsoft Research source ↗

Aug 12 Recall is the bottleneck for parametric factuality in generative AI Google Research source ↗

Aug 12 Controlled memory interference in continual LLM agents arXiv cs.AI source ↗

Aug 12 Towards an argumentative foundation for evaluative AI arXiv cs.AI source ↗

Aug 12 IntelliAudit: using large language models to evaluate audit controls arXiv cs.AI source ↗

Aug 12 Emotion in an active inference model of human driving arXiv cs.AI source ↗

Aug 12 Linear probes detect corrupted context in language models but do not reliably predict final answer correctness across multi-hop arithmetic chains. arXiv cs.AI source ↗

Aug 12 Towards researcher agents for knowledge-graph question answering arXiv cs.AI source ↗

Aug 12 When LLM agents negotiate: private information and dynamic bargaining in supply chains arXiv cs.AI source ↗

Aug 12 QuantumMind proposes an auditable agentic workflow to generate and conservatively screen quantum-acceleration hypotheses for speedup analysis in quantum computing arXiv cs.AI source ↗

Aug 12 MetaSpace: metamorphic testing for spatial cognition in embodied agents arXiv cs.AI source ↗

Aug 12 Determinization in structure theories: a unified framework via closure, comparability, and joint admissibility arXiv cs.AI source ↗

Aug 12 CMU-Drive and V2V-VLA: cooperative multi-agent unified driving with reasoning benchmark and vehicle-to-vehicle vision-language-action models arXiv cs.AI source ↗

Aug 12 Dynamic coalition formation and communication pricing in skill-based agentic AI systems arXiv cs.AI source ↗

Aug 12 Cross-dimensional geometry from navigation to black holes: a continuous metric field learned via causal contrastive loss encodes scenes into Lie algebra elements and yields Riemannian or Lorentzian metrics arXiv cs.AI source ↗

Aug 12 Contextual value alignment via multilayer combinatorial fusion arXiv cs.AI source ↗

Aug 12 Mendel Gödel Machine: self-improving coding agents use comparative evolution to rewrite their own source code arXiv cs.AI source ↗

Aug 12 AI scientist for studying generalization in quadruped robot navigation policies in simulation; addresses drift in autonomous research loops arXiv cs.AI source ↗

Aug 12 Training Variable Long Sequences with Data-Centric Parallel arXiv cs.AI source ↗

Aug 12 From single chatbots to governed agent ecosystems: an agentic AI pattern catalogue and orchestration framework for mission-critical hospital information management systems arXiv cs.AI source ↗

Aug 12 Flow-by-Flow: Content-Judgment Bypass for Governing AI Output in High-Loss Domains arXiv cs.AI source ↗

Aug 12 TeXFix-Bench: an empirically grounded multi-format benchmark for LLM-based full-source document repair arXiv cs.AI source ↗

Aug 12 NL2SHACL-Bench: a benchmark suite for translating natural language to SHACL shapes arXiv cs.AI source ↗

Aug 12 TREAT: evaluating access to formal knowledge across equivalent mathematical representations arXiv cs.AI source ↗

Aug 12 Protecting patient privacy in clinical foundation models: technical and legal perspectives arXiv cs.AI source ↗

Aug 12 Agent-MD: Selective LLM intervention with event-driven escalation for stateful GCMC–MD campaigns arXiv cs.AI source ↗

Aug 12 TAF-MED measures multi-turn safety refusal collapse in LLMs when users express self-treatment intent arXiv cs.CL source ↗

Aug 12 Boundary-Seeking Policy Gradient for safe reinforcement learning arXiv cs.LG source ↗

Aug 12 MD-ProTector: positioning multiple data-driven prototypes for LLM-generated text detection arXiv cs.CL source ↗

Aug 12 ChronoSSM: training for temporally aware representations in autoregressive state space models arXiv cs.LG source ↗

Aug 12 Transformer Geometry Observatory TGO-IV: Developmental topology of representations across transformer layers explored via inductive analysis arXiv cs.LG source ↗

Aug 12 Similarity gates approve reversals: a validity audit of embedding-cosine thresholds in agent systems arXiv cs.CL source ↗

Aug 12 REATS: LLM reasoning-based ensemble learning for adaptive time series forecasting arXiv cs.LG source ↗

Aug 12 PERCEPT: a corpus for POS tagging and analysis of Persian-English code-mixing arXiv cs.CL source ↗

Aug 12 ASR-roundtrip evaluation can mask context- and convention-dependent reading errors in Chinese news TTS arXiv cs.CL source ↗

Aug 12 Procedural fairness failures in RLHF from preference averaging arXiv cs.LG source ↗

Aug 12 Sheaf-based federated representation learning arXiv cs.LG source ↗

Aug 12 Matched-integrator evaluation of Hamiltonian neural networks on pendulum and Kepler dynamics arXiv cs.LG source ↗

Aug 12 Finding the signal in the spam: jointly learning rewards and worker reliability from pairwise comparisons arXiv cs.LG source ↗

Aug 12 Power law graph attention: exact generalization of scaled dot-product attention; empirical collapse at inference arXiv cs.LG source ↗

Aug 12 Cross-contextual consistency as a measure of LLM credibility arXiv cs.CL source ↗

Aug 12 When chain-of-thought helps and when it hurts: an empirical investigation of the serial-depth bottleneck in LLM reasoning arXiv cs.CL source ↗

Aug 12 Multimodal item parameter estimation using simulated response probabilities arXiv cs.CL source ↗

Aug 12 DOCSCHISEL: Adaptive tool documentation optimization framework for LLM agents arXiv cs.LG source ↗

Aug 12 Simplex Relaxation for Discrete Diffusion; arXiv:2608.10615 introduces Simplax, an exact Dirichlet–categorical augmentation that enriches training objectives and reverse transitions without changing the underlying corruption process arXiv cs.CL source ↗

Aug 12 ELMER: Evolutionary Language Model that explores and refines arXiv cs.LG source ↗

Aug 12 How robust are LLMs to Vietnamese dialects? arXiv cs.CL source ↗

Aug 12 SeFoRA uses sketch-aggregated federated LoRA to handle heterogeneous client ranks in federated parameter-efficient fine-tuning with low-rank adaptation. arXiv cs.LG source ↗

Aug 12 Position encoding in transformers: from absolute and relative methods to rotary position embeddings and long-context scaling arXiv cs.CL source ↗

Aug 12 UserToolBench tests personalized decision making in tool-use LLMs by inferring latent user preferences from interaction history arXiv cs.LG source ↗

Aug 12 Analyzing LLM alignment across single- and multi-cultural settings using cultural consensus theory arXiv cs.CL source ↗

Aug 12 Detecting soft skills in ML engineering roles CVs arXiv cs.LG source ↗

Aug 12 FlowScout: from execution feedback to reliable tool-using agent workflows arXiv cs.LG source ↗

Aug 12 From prediction to incrementality: causal optimization for large-scale targeting and recommendation arXiv cs.LG source ↗

Aug 12 Calibrating post-training feature shifts for LLM data contamination detection arXiv cs.CL source ↗

Aug 12 Dual-Loop Self-Evolution via Verifiable Emotion Feedback for Multi-Turn Empathetic Dialogue arXiv cs.CL source ↗

Aug 12 CurveFP: rational-radix logarithmic datatypes with closed products for language models arXiv cs.LG source ↗

Aug 12 Asymmetric framing of La France insoumise and Rassemblement National in French news headlines, 2022–2025 arXiv cs.CL source ↗

Aug 12 Every Token Counts: exact Likert-scale distributions for measuring LLM attitudes and biases arXiv cs.CL source ↗

Aug 12 Lightweight bias correction in constrained decoding for grammar-constrained decoding masks; online sampling reduces distortion in LM outputs arXiv cs.CL source ↗

Aug 12 Independent reproduction of LeWorldModel achieves 94.0% goal offset in TwoRoom; evaluation on a single laptop CPU is reported arXiv cs.LG source ↗

Aug 12 Uncertainty-aware ensemble deep randomized neural networks for classification arXiv cs.LG source ↗

Aug 12 Physics-informed machine learning in prognostics and health management: a systematic literature review arXiv cs.LG source ↗

Aug 12 STCAD: scalable trajectory clustering and anomaly detection on terabyte-scale AIS data arXiv cs.LG source ↗

Aug 12 Off-Axis, On Purpose: where a Transformer computes concepts and why it does so arXiv cs.CL source ↗

Aug 12 Multilingual quantization tax: edge SLMs suffer performance degradation from 4-bit weight quantization across languages, study finds across Gemma 4 and Qwen 3.5 using MMLU ProX Lite and GlobalPIQA arXiv cs.CL source ↗

Aug 12 Toward human rights benchmarking for LLMs: a pilot methodology arXiv cs.LG source ↗

Aug 12 Observational policy ranking for SMB financial guidance from multi-action accounting logs arXiv cs.LG source ↗

Aug 12 CRHT: A Continuous Regression Hybrid Transformer for vessel trajectory prediction with online cluster sampling arXiv cs.LG source ↗

Aug 12 LLM Agents Factory: retrieval-based framework builds domain-specific, Wikipedia-grounded agents on demand arXiv cs.CL source ↗

Aug 12 Intrinsic Structure: Spectral identifiability for mechanistic interpretability arXiv cs.LG source ↗

Aug 12 Cracks in the foundation: four minor architectural decisions reduce long context extensibility across Olmo, Llama, and Qwen dense models arXiv cs.CL source ↗

Aug 12 Edge Phoneme Recognition for Children's Speech through Age-Aware Training arXiv cs.AI source ↗

Aug 12 Mitigating bus bunching with reinforcement learning enhanced by semantic stop embedding arXiv cs.AI source ↗

Aug 12 MESA: Task-adaptive multi-structure evidence selection for long-horizon agent memory arXiv cs.AI source ↗

Aug 12 CHORUS: complementary experts for high-coverage testbench stimulus generation arXiv cs.AI source ↗

Aug 12 MIDAS: mutual information disentanglement with uncertainty-aware fusion for incomplete multimodal sentiment analysis arXiv cs.AI source ↗

Aug 12 Automating and scaling behavioral scientific research on AI agents with AEROBAT arXiv cs.AI source ↗

Aug 12 Closed-loop LLM co-pilots for digital agriculture arXiv cs.AI source ↗

Aug 12 ReCBM: uncertainty-gated relational reasoning for concept bottleneck models arXiv cs.AI source ↗

Aug 12 TRACE: Trustworthy retrieval-augmented conversational engine discusses improving reliability of public-service chatbot recommendations with retrieval augmentation arXiv cs.AI source ↗

Aug 12 Decodable but not detachable: training data granularity determines parametric modularity in large language models arXiv cs.AI source ↗

Aug 12 Logit-boundary geometric belief interfaces and sparse sheaf-enclave protocols for secure EHR interoperability arXiv cs.AI source ↗

Aug 12 SPOT: Sampling Policy Observation Tree provides lookahead explanations for deep reinforcement learning policies. arXiv cs.AI source ↗

Aug 12 Mind viruses: self-propagating ideas in multi-agent LLM systems arXiv cs.AI source ↗

Aug 12 Interpreting language model hidden states at scale arXiv cs.AI source ↗

Aug 12 Post-hoc sparse coding of latent communication between vision-language model agents arXiv cs.AI source ↗

Aug 12 Beyond detection: evaluating defensive LLMs against AI-generated social engineering in live turn-by-turn interaction arXiv cs.AI source ↗

Aug 12 Generating attacks for LLMs with GFlowNets arXiv cs.AI source ↗

Aug 12 Evaluation-Conditioned Training: teaching models to generalize to stronger oversight regimes arXiv cs.AI source ↗

Aug 12 Towards Sustainable AI: a comprehensive review and comparative analysis of deep learning models’ carbon footprint arXiv cs.AI source ↗

Aug 12 Relational geometry attacks on contrastive embedding manifolds emerge in new study arXiv cs.AI source ↗

Aug 12 Neuroevolution Arena: nested ecological evaluation of update-and-inheritance regimes across neural architectures arXiv cs.AI source ↗

Aug 12 LinkedIn presents a self-evolving agentic support system that uses retrieval-augmented generation, evolutionary auto-prompting, and a modular evaluation framework for continuous, production-aligned improvement. arXiv cs.AI source ↗

Aug 11 CARE-X aims for clinically useful radiology VLMs with auxiliary supervision, reward-aligned learning, and tool-augmented measurement Microsoft Research source ↗

Aug 11 LLMCompiler generates program embeddings from source and IR code for program analysis and optimization tasks. arXiv cs.LG source ↗

Aug 11 WuYuEval: a multi-level benchmark for evaluating large language models in solid waste management across foundational knowledge, domain reasoning, and expert decision-making arXiv cs.CL source ↗

Aug 11 STEMMA: an adversarial multi-agent framework for evaluating self-identity consistency in LLMs arXiv cs.CL source ↗

Aug 11 Scaling inherently interpretable language models by training with interpretability as a constraint alongside language modeling objective arXiv cs.CL source ↗

Aug 11 Grounded search agents use representation-based intrinsic rewards to guide retrieval and grounding of language model outputs arXiv cs.CL source ↗

Aug 11 LUCID: Latent-Skill Unified Control via Imagined Dynamics for long-horizon humanoid loco-manipulation arXiv cs.LG source ↗

Aug 11 DialectS2S: End-to-end speech dialogue modeling for low-resource Chinese dialects arXiv cs.CL source ↗

Aug 11 From token probabilities to calibrated confidence: an empirical study of mathematical question answering arXiv cs.LG source ↗

Aug 11 Many Are My Names": The Anatomy of the Assistant and Its Personas via Sparse Autoencoders arXiv cs.CL source ↗

Aug 11 Data-driven fire-zone segmentation improves short-term wildfire prediction arXiv cs.LG source ↗

Aug 11 Multilingual supervision improves figurative language identification in proverbs, evaluated across seven languages and 742 proverb concepts arXiv cs.CL source ↗

Aug 11 Adaptive KappaSharp: Condition-Number Shaping for Preferential Bayesian Optimization arXiv cs.LG source ↗

Aug 11 Information Routing across Batch Boundaries: Memory–Batch Tradeoffs in Lipschitz Bandits arXiv cs.LG source ↗

Aug 11 DocAtlas treats long-document understanding as mutable-state interaction arXiv cs.CL source ↗

Aug 11 Finite constant frontiers and auditable regret certificates for average-reward reinforcement learning arXiv cs.LG source ↗

Aug 11 Interpreting reasoning mechanisms of large language models via sparse autoencoders: separating Thinking from NoThinking in CoT-enabled models arXiv cs.CL source ↗

Aug 11 Polish vision-language evaluation PoVisLE assesses cultural grounding in vision-language models arXiv cs.CL source ↗

Aug 11 Foundation models in cognitive science: evaluating their cognitive alignment and developmental trajectories arXiv cs.CL source ↗

Aug 11 Tracing sources of epistemic uncertainty in deep learning predictions with homo- and heteroscedastic linearized estimators arXiv cs.LG source ↗

Aug 11 Policy learning with mu-resets: the sample complexity under policy realizability for the Kakade–Langford interaction protocol arXiv cs.LG source ↗

Aug 11 Trace-driven evaluation can mislead MoE expert caching due to replay semantics, workload contamination, and operating regimes. arXiv cs.LG source ↗

Aug 11 NeuPAT: neuron-aware plasticity allocation tuning for language-preserving mllms arXiv cs.CL source ↗

Aug 11 Neural operators for immersed-boundary soft swimmers locomotion arXiv cs.LG source ↗

Aug 11 From benchmark performance to tool deployment: human-in-the-loop anomaly detection arXiv cs.LG source ↗

Aug 11 Spectral outliers reveal dominant learned structure in transformer attention. arXiv cs.LG source ↗

Aug 11 CODS: Iterative Bellman-residual data selection for reusable offline reinforcement learning arXiv cs.LG source ↗

Aug 11 APEX-VW: a document-level English-Spanish post-editing dataset in the healthcare domain arXiv cs.CL source ↗

Aug 11 SkillConsist detects inconsistencies in agent skills via bidirectional graph alignment arXiv cs.LG source ↗

Aug 11 SPECTRA: pushing the KV cache beyond the 2-bit cliff via spectral transform coding arXiv cs.LG source ↗

Aug 11 SurakshaEval: a safety benchmark for multilingual LLMs covering ten major Indian languages arXiv cs.CL source ↗

Aug 11 Applying artificial intelligence to recognize fraudulent banking operations; arXiv:2608.07471v1 discusses using machine learning for online banking fraud detection arXiv cs.LG source ↗

Aug 11 TEMPER: Tensorized Efficient Manifold-constrained Parameterization for Expressive Residual Routing arXiv cs.LG source ↗

Aug 11 V-Simba improves architectural efficiency for reinforcement learning in visual continuous control arXiv cs.LG source ↗

Aug 11 Shape mutating expert compression: LorExperts and BTExperts arXiv cs.LG source ↗

Aug 11 CONFER: conflict-aware evidence negotiation for regime-calibrated weak supervision in multimodal emotion recognition arXiv cs.LG source ↗

Aug 11 Evaluating dedicated monolingual and joint multilingual causal models for Dravidian languages arXiv cs.CL source ↗

Aug 11 SurveyReview: a reviewer-aligned benchmark for survey evaluators arXiv cs.CL source ↗

Aug 11 Unified hallucination fuzzing for multimodal large language models arXiv cs.CL source ↗

Aug 11 Prompt Embedding Probes detects hallucinations from hidden states of a frozen LLM arXiv cs.CL source ↗

Aug 11 Reasoning models fail to ration test-time compute across questions arXiv cs.CL source ↗

Aug 11 Embedding initialization for unseen low-resource languages in multilingual NMT: a case study on Limbum-English translation arXiv cs.CL source ↗

Aug 11 Classifier-free guidance becomes unnecessary at times in masked diffusion language models, study finds arXiv cs.CL source ↗

Aug 11 Detection of self-introductions in legislative testimony arXiv cs.CL source ↗

Aug 11 PhysAttNet: enhancing predictive performance in industrial and astrophysical time series via physics-informed attention arXiv cs.LG source ↗

Aug 11 The no-meaning falsity: the structural impossibility of the arbitrary sign in Classical Arabic arXiv cs.CL source ↗

Aug 10 AI models simulate a wider range of real-world scenarios with a physics feel, enabling engineers to test vehicle blueprints with fewer physical experiments, says MIT CSAIL researcher Haixu Wu MIT News (AI) source ↗

Aug 10 From cheap fakes to pure synthesis: addressing the new era of T2V fake news videos arXiv cs.AI source ↗

Aug 10 Can MLLMs decode the creative leap? Introducing C4 for cross-concept understanding arXiv cs.AI source ↗

Aug 10 The optimizer is the agent: reasoning-driven search across prompts, programs, and ML workflows arXiv cs.AI source ↗

Aug 10 Theoretical foundations of communication-efficient, robust, and practical distributed and federated optimization arXiv cs.LG source ↗

Aug 10 Measuring the cross-lingual comprehension gap: how the language of the evidence shapes what language models understand arXiv cs.CL source ↗

Aug 10 MolBioKG grounds out-of-graph molecules in biomedical knowledge graphs via multi-resolution structural anchoring. arXiv cs.AI source ↗

Aug 10 NxN E-valuation uses e-value-based hypothesis certification to verify hypotheses without case-specific procedures, given a large dataset; suited for LLM-based exploration systems to mitigate hallucination. arXiv cs.AI source ↗

Aug 10 ADIAS: Automated Design of Interactive Agentic Systems arXiv cs.AI source ↗

Aug 10 Cryptanalytic extraction of isolated bias-free GLU feed-forward blocks by antipodal separation demonstrated for bias-free GLU blocks in language-model components. arXiv cs.LG source ↗

Aug 10 Learning to predict middle-layer attention in MLLMs for visual token pruning arXiv cs.AI source ↗

Aug 10 Sharding prevents LLM oversight failures and adversarial exploitation arXiv cs.LG source ↗

Aug 10 Recovering lesion parameters from aphasic picture naming error profiles in large language models arXiv cs.CL source ↗

Aug 10 Shape your feed: an LLM-based agentic system for conversational recommendation arXiv cs.AI source ↗

Aug 10 FutureBridge ranks joint LLM-SLM token candidates for collaborative decoding beyond local preference arXiv cs.CL source ↗

Aug 10 Selection-aware backdoor attacks in federated learning bypass Krum via Krum-Proxy attack arXiv cs.LG source ↗

Aug 10 Divergent response modes in frontier language models under steering pressure. arXiv cs.AI source ↗

Aug 10 Sparsity-aware pruning preserves differences between outputs in sparsity-sensitive neurons, not just activations, for large language model pruning arXiv cs.LG source ↗

Aug 10 Stockmark-Nemotron-3-Nano-Omni-JapanDocReader enables structured document parsing via capability injection and forgetting control in a Japanese document understanding model built from Nemotron-3-Nano-Omni-30B-A3B-Reasoning-BF16 arXiv cs.CL source ↗

Aug 10 EntropyMoE: entropy-aware sparse expert routing for tokenizer-free LLMs arXiv cs.AI source ↗

Aug 10 WebGrader: Training LLMs for web development with self-evolving programmatic grader arXiv cs.AI source ↗

Aug 10 Interpretable unsupervised community detection with LLM-symbolized structured processes arXiv cs.AI source ↗

Aug 10 Knowledge-Driven AI agents for smart degree pathway planning; arXiv study describes solving curriculum reconstruction and path optimization with prerequisite logic arXiv cs.AI source ↗

Aug 10 Autonomy-of-Heads introduces data-free sparse attention from frozen query-key geometry to reduce long-context inference costs arXiv cs.CL source ↗

Aug 10 CertBind: certifiable composition for frozen multimodal connector graphs arXiv cs.LG source ↗

Aug 10 Beyond “AI Language”: the idiolectal nature of LLM output across six models in 2024 and 2026 corpora arXiv cs.CL source ↗

Aug 10 WebRider: persona-conditioned intent controllers for live-web assistance arXiv cs.AI source ↗

Aug 10 Pre-inference routing for cost-efficient document field extraction improves cost when a cheaper model fails often enough to justify routing, according to arXiv:2608.06607v1 arXiv cs.CL source ↗

Aug 10 TRACE: A multi-layer benchmark for human AI controller coordination under drift and failure arXiv cs.AI source ↗

Aug 10 bioMoR applies mixture-of-recursions to gene- and pathway-level learning for efficient genomic analysis arXiv cs.AI source ↗

Aug 10 Soft-masking accelerates convergence in Masked Diffusion Language Models; the paper argues that linear interpolation in the embedding space is flawed due to near-constant token/mask embedding angles during training. arXiv cs.CL source ↗

Aug 10 MTI-GNN models Beck's cognitive triad for distortion detection in text classification arXiv cs.CL source ↗

Aug 10 The horizon gap: planning, memory, execution, training, and evaluation for long-horizon LLM agents arXiv cs.CL source ↗

Aug 10 Online security learning in cooperative multi-agent systems under hidden Byzantine attacks. arXiv cs.LG source ↗

Aug 10 Bootstrap-conditioned action selection with tabular foundation models arXiv cs.LG source ↗

Aug 10 AgentPatch: coarse-to-fine weak-task repair for merging agentic multimodal large language models arXiv cs.AI source ↗

Aug 10 Flowing through States: Neural ODE regularization for reinforcement learning arXiv cs.LG source ↗

Aug 10 Vehicle routing problem using deep reinforcement learning; a case study on truck planning in industry arXiv cs.AI source ↗

Aug 10 Faithful response-level interpretation of mixture-of-experts reward models via contribution contrast arXiv cs.AI source ↗

Aug 10 TradeVerse: a longitudinal benchmark of political negotiation in international trade arXiv cs.CL source ↗

Aug 10 Factorized hypothesis search for evidence-to-taxonomy retrieval arXiv cs.CL source ↗

Aug 10 GRASP: Reinforcing language model anonymizers with group relative policy optimization arXiv cs.CL source ↗

Aug 10 CrystalGRPO: Target-aligned and coverage-preserving reinforcement learning for flow-based crystal structure prediction arXiv cs.LG source ↗

Aug 10 Latent fact-checking: detecting misinformation through activation engineering arXiv cs.LG source ↗

Aug 10 Weak presupposition verification degrades general QA performance arXiv cs.CL source ↗

Aug 10 Separating decision-rule misalignment from readout-coverage limitations in speech language models arXiv cs.CL source ↗

Aug 10 CellWorld: latent-space predictive pretraining enables foundation models for spatial transcriptomics arXiv cs.AI source ↗

Aug 10 IB-RL: Isolated bilateral reinforcement learning for strategic dialogue agents arXiv cs.AI source ↗

Aug 10 Robust average-reward Markov decision processes: minimax-optimal learning via plug-in reductions arXiv cs.LG source ↗

Aug 10 NTDH: Complex reasoning for comprehensive affective analysis arXiv cs.CL source ↗

Aug 10 Beyond attention: signed integrated gradients attribution in a BiomeGPT-style microbiome transformer arXiv cs.LG source ↗

Aug 10 ConstructCIE: a dataset for extracting causal information from construction accident narratives arXiv cs.CL source ↗

Aug 10 Towards multi-label graph foundation models: from single-vector representation learning to multi-semantic basis learning arXiv cs.AI source ↗

Aug 10 ED-CSP: crystal structure prediction from electron diffraction arXiv cs.LG source ↗

Aug 10 TA-RAG: Tone Awareness as a Design Imperative for Retrieval-Augmented Generation arXiv cs.CL source ↗

Aug 10 SNI-GNN: SmartNIC-assisted full-graph GNN training with in-network embedding prediction arXiv cs.LG source ↗

Aug 10 MiGHT-EHR: a multi-task graph transformer for heterogeneous temporal electronic health records arXiv cs.LG source ↗

Aug 10 Discovering conceptual metaphors across topics and media types arXiv cs.CL source ↗

Aug 10 Do AI personas grow? Analyzing and benchmarking personality evolution in LLM agents after life events arXiv cs.CL source ↗

Aug 10 Progressive content refinement with decaying reward joint LinUCB arXiv cs.CL source ↗

Aug 10 Risk-aware decision policies for agents under noisy perception arXiv cs.LG source ↗

Aug 10 TaskSense: focusing on what matters in world models arXiv cs.AI source ↗

Aug 10 CGMas: a multi-agent framework that automates topology construction, equilibration, mapping, and potential derivation for coarse-grained molecular dynamics of polymers arXiv cs.AI source ↗

Aug 10 Newton-Schulz retraction-based inference enables hidden quantum Markov models to outperform classical HMMs arXiv cs.LG source ↗

Aug 10 Toward reliable context compression for long-horizon agents: an empirical study of execution instability arXiv cs.LG source ↗

Aug 10 Quantization damage is multiplicative, not additive arXiv cs.LG source ↗

Aug 10 Unmasking removal-budget confounding: a matched operating-point evaluation framework for adaptive data cleaning arXiv cs.LG source ↗

Aug 10 Confidence estimation for financial vision-language models in chart and document understanding arXiv cs.CL source ↗

Aug 10 Adversarial causal intervention falsification arXiv cs.LG source ↗

Aug 10 Fixed and adaptive topological DeepONets use functional measurements on Hausdorff locally convex spaces arXiv cs.LG source ↗

Aug 10 Automated item evaluation predicts item acceptance and rejection from item text using historical rejection data and LLM-generated critiques. arXiv cs.AI source ↗

Aug 10 Simple-OPD: Demystifying warm-up for on-policy distillation arXiv cs.CL source ↗

Aug 10 Target-weighted Neyman allocation uses pilot variances to design experiments for heterogeneous treatment effects under population shift arXiv cs.LG source ↗

Aug 7 Graph-Guided Project-Level Planning for Autonomous Research identifies a method to decompose long-horizon research projects into bound tasks with dependency-aware sequencing. arXiv cs.AI source ↗

Aug 7 Matrix Zonotopic Attention: a context-adaptive value projection for set transformers arXiv cs.LG source ↗

Aug 7 ASR adaptation and representation dimensionality reduction for Persian speech emotion recognition using Whisper arXiv cs.CL source ↗

Aug 7 Spectral Distillation enables learning a compact linear state-space model from nonlinear dynamics using an implicit spectral predictor via Observation Spectral Filtering. arXiv cs.LG source ↗

Aug 7 ConWriter: transition-constrained Stateful long-form story generation with lightweight neuro-symbolic consistency control arXiv cs.CL source ↗

Aug 7 Expl explorative modeling (XM) introduces a third pretraining axis by producing multiple outputs per comparison to enhance generative expressivity arXiv cs.LG source ↗

Aug 7 CASCADE: an agentic regulatory network framework predicts downstream transcriptional effects of gene perturbation using ARACNe networks and MCP exposure arXiv cs.AI source ↗

Aug 7 Constraint-First Reasoning: a training-free protocol for exploiting answer-space constraints in mathematical problem solving arXiv cs.CL source ↗

Aug 7 Analysis of numerical localisation in LLM translations arXiv cs.CL source ↗

Aug 7 When privileged guidance misaligns: state-matched routing and contextualized self-distillation for multi-turn agents arXiv cs.AI source ↗

Aug 7 Where privacy risk lives in English-source multilingual RAG: a stage-decomposed audit across five query languages arXiv cs.CL source ↗

Aug 7 Counterfactual analysis via large language models in online lending using GPT-3.5 to assess counterfactual ROI under different interest rate schemes arXiv cs.AI source ↗

Aug 7 Simulator-grounded LLMs for industrial causal reasoning in wastewater treatment use a live simulator oracle, structured parameter injection, and plant-portable retrieval to support decision making. arXiv cs.CL source ↗

Aug 7 LUNAR benchmarks personalized large language models on universal daily-life logs by evaluating responses from longitudinal app interaction histories across domains arXiv cs.AI source ↗

Aug 7 From continuous predictors to clinical thresholds: early evidence on performance trade-offs of guideline-based categorisation for ischaemic stroke outcome prediction arXiv cs.AI source ↗

Aug 7 KV-Skill enables a frozen language model to read external factorized operators via a lightweight interface; introduces a design space for external knowledge in the model’s native language arXiv cs.LG source ↗

Aug 7 PD-GS: Phoneme-driven 3DGS for audio-driven talking heads arXiv cs.AI source ↗

Aug 7 MS-MLB: An Open Machine Learning Benchmark for Blood-Based MS Classification arXiv cs.LG source ↗

Aug 7 The interface of intonation and lexical tone in Mandarin varieties: boundary phenomena between intonation and tone detect and convey sentence type and attitudinal information arXiv cs.CL source ↗

Aug 7 Universal Pathologies, Conditional Consequences: A triple-robustness analysis of RAG for multi-hop traceability arXiv cs.CL source ↗

Aug 7 An emerging retail portfolio management application: personalized, tax-aware reinforcement learning with natural language goals arXiv cs.LG source ↗

Aug 7 EvoHarness-RL: learning a self-evolving runtime harness for long-horizon LLM agents arXiv cs.LG source ↗

Aug 7 SkillTrace: multi-trace provenance auditing for LLM-agent skill reuse arXiv cs.AI source ↗

Aug 7 DG-FedReuse enables cached updates from selected clients based on age-decayed proxy-gradient discrepancy in federated learning, with cache-age limits and fresh-client quotas arXiv cs.LG source ↗

Aug 7 When do corrective features help? An agent for corrective feature discovery on black-box forecasters arXiv cs.LG source ↗

Aug 7 Hybrid probabilistic zonotopes for identifiable and refinable predictive uncertainty arXiv cs.LG source ↗

Aug 7 LLM-based flows modeled as probabilistic programs reveal uncertainty in multi-call LLM pipelines arXiv cs.LG source ↗

Aug 7 The Ignition Index measures global workspace dynamics in transformer language models across 11 models arXiv cs.AI source ↗

Aug 7 Example-guided prompting for document-level text simplification shows improved consistency with retrieved examples arXiv cs.CL source ↗

Aug 7 DREAM: LLM-based dynamic role-playing via event-aware memory graph arXiv cs.CL source ↗

Aug 7 RIG-RoPE: Relation- and instance-gated rotary positional encoding with duration-aware temporal coordinates arXiv cs.CL source ↗

Aug 7 PRISM: Priority-aware rubric internalization via structured multimodal data synthesis arXiv cs.LG source ↗

Aug 7 OrchestraBench evaluates failure, recovery, and decomposition in multi-agent orchestration with a seed-reproducible failure-injection framework over templated enterprise workflows arXiv cs.AI source ↗

Aug 7 Evaluating machine learning models for post-wildfire debris-flow prediction arXiv cs.LG source ↗

Aug 7 C3PO benchmark evaluates cross-modal composition and counterfactual conflict in omnimodal models arXiv cs.AI source ↗

Aug 7 DBLAST: Dependent Block Drafting for Stochastic Speculative Decoding arXiv cs.CL source ↗

Aug 7 CNM-BERT: a drop-in structural embedding for Chinese characters via Ideographic Description Sequences arXiv cs.CL source ↗

Aug 7 Coherence-Oriented Dream Scene Visualisation: a system that turns written dream descriptions into a four-panel, temporally coherent sequence of images using a four-part prompt and text-to-image models. arXiv cs.AI source ↗

Aug 7 Conditional cognitive biases in LLMs: biased user turns modulate in-context reasoning arXiv cs.CL source ↗

Aug 7 Skill distillation beyond knowledge boundaries via rubric-based reinforcement learning arXiv cs.AI source ↗

Aug 7 Disentangling 3D modeling from spatial reasoning. arXiv cs.LG source ↗

Aug 7 Abstract Event Causal Rules: induction and application arXiv cs.AI source ↗

Aug 7 Decoupling perception from description: computation-grounded representation alignment between multivariate time series and language arXiv cs.LG source ↗

Aug 7 Small foundation models of human cognition and behaviour show in-distribution performance insensitive to scale across a 14B-parameter range arXiv cs.AI source ↗

Aug 7 Rectifying geometric misalignment: online source-free adaptation for class-imbalanced EEG arXiv cs.LG source ↗

Aug 7 Mood matters: syntactic sensitivity undermines safety alignment arXiv cs.CL source ↗

Aug 7 CANOE: a multi-agent neuro-symbolic framework for contestable care plan coordination with adaptive team recruitment and role-based modules arXiv cs.AI source ↗

Aug 7 WorldClaw releases a fully agentic, coarse-to-fine framework for open-world 3D scene generation arXiv cs.AI source ↗

Aug 7 AuroSFT for adapter-state multi-task fine-tuning replaces full-model rollback with adapter-state scheduling for multi-task fine-tuning arXiv cs.LG source ↗

Aug 7 Safe evolution with circuit anchors arXiv cs.CL source ↗

Aug 7 Beyond Rotations: AuroOFT enables expressive quantized orthogonal fine-tuning with gated low-rank corrections arXiv cs.LG source ↗

Aug 7 Evidence lock before commitment: a frozen interface degrades LLM-as-judge evaluation arXiv cs.CL source ↗

Aug 7 Align-RAG: Alignment is all you need for TSFM in-context learning arXiv cs.LG source ↗

Aug 7 LC-GRPO: Bridging train-inference gap for flow-based GRPO with Langevin correction arXiv cs.LG source ↗

Aug 7 TriQua: reconciling granularity and context in factuality evaluation arXiv cs.AI source ↗

Aug 7 SearchAuditor: auditing and attributing failures in long-horizon search agents arXiv cs.AI source ↗

Aug 7 Large language models threaten double-blind review by enabling author identification from titles and content. arXiv cs.CL source ↗

Aug 7 Quantum-Structured World Models (QSWMs) for predictive latent dynamics. arXiv cs.LG source ↗

Aug 7 Agentic self-driving microscopy benchmarks support qualification but do not necessarily generalize to unseen tasks arXiv cs.AI source ↗

Aug 7 SemiAdapt-Instruct enables extensible instruction tuning with latent domain-specific adapters and parameter-free routing arXiv cs.CL source ↗

Aug 7 QEvict: recoverable quantized KV eviction for attention-drift-robust long-context decoding arXiv cs.LG source ↗

Aug 7 DoctorAgents: an agentic framework to iteratively refine AutoML pipeline for small clinical temporal data arXiv cs.AI source ↗

Aug 7 Scaffold-mediated post-training: co-evolving model parameters and procedural scaffold graphs arXiv cs.CL source ↗

Aug 7 Otter: a time-aware, history-conditioned human chess AI uses move history and time control signals to predict human move selection arXiv cs.AI source ↗

Aug 7 Perturbation sensitivity at convergence identifies spuriously correlated samples without group annotations arXiv cs.LG source ↗

Aug 7 Cross-architecture steering transfer in language models: a systematic empirical study arXiv cs.CL source ↗

Aug 7 Posture and sustainment optimization under adversarial uncertainty arXiv cs.AI source ↗

Aug 7 Beyond sentiment: comparing traditional NLP and LLM-based multi-dimensional framing analysis for political news evaluation arXiv cs.CL source ↗

Aug 7 Position: it's time to optimize LLMs for self-consistency arXiv cs.CL source ↗

Aug 7 Agentic nesting: a new methodology for existing enterprise application integration and services arXiv cs.AI source ↗

Aug 7 PoolBench: a benchmark for pooling strategies in concept representation evaluation for decoder-only LLMs arXiv cs.CL source ↗

Aug 7 Woodpecker Distillation shows weak models diagnose reasoning bugs in strong models arXiv cs.AI source ↗

Aug 7 IFlowNets: Extending Generative Samplers to Learn Strategies in Incomplete Information Games arXiv cs.LG source ↗

Aug 7 Auditing conditional style leakage in factorized generative models shows marginal matching does not guarantee independence between style and class information arXiv cs.LG source ↗

Aug 6 JudgeArena: a unified framework for reproducible LLM-judge evaluation arXiv cs.CL source ↗

Aug 6 Learning a vector-symbolic model for socio-cultural tasks arXiv cs.CL source ↗

Aug 6 Evaluating OpenAI's Privacy Filter across 42 benchmarks: cross-lingual and cross-domain PII detection arXiv cs.CL source ↗

Aug 6 ARCHead: Activation-Metric Residual Correction for large language model output heads arXiv cs.CL source ↗

Aug 6 Emulate or estimate? The divergent strengths of base and post-trained language models for opinion simulation arXiv cs.CL source ↗

Aug 6 OncoTriad-QA: a patient-level radiology-pathology-genomics benchmark for pan-cancer reasoning arXiv cs.CL source ↗

Aug 6 Crayotter: Learning long-horizon video editing agents via group-relative preference backpropagation arXiv cs.CL source ↗

Aug 6 FLARE: Few-shot learning-based adaptive reflective engine arXiv cs.CL source ↗

Aug 6 Aligned in form, not in meaning: the comprehension–containment decoupling of LLM safety in low-resource Bangla derogatory speech arXiv cs.CL source ↗

Aug 6 TabletCraft enables bidirectional Akkadian NMT and cuneiform rendering to bridge a 4,000-year cultural gap. arXiv cs.CL source ↗

Aug 6 Knowing the form, not the function: automatically auditing answer–authority decoupling in legal benchmarks arXiv cs.CL source ↗

Aug 6 TQLite: multi-LLM jury-guided distillation for real-time MQM translation quality evaluation arXiv cs.CL source ↗

Aug 6 BBOWP-Bench: Evaluating LLMs on Black-Box Optimization Word Problems arXiv cs.CL source ↗

Aug 6 Character iconicity vs. arbitrariness: an Arabic NLP perspective arXiv cs.CL source ↗

Aug 6 Study questions reliability of statistical measures in script decipherment; SIGIL shows 3,000-text corpus with explicit compositional meanings but undeciphered signs arXiv cs.CL source ↗

Aug 6 MemArena: an ego-centric benchmark for on-device agentic personal memory assistants at scale arXiv cs.CL source ↗

Aug 6 Language models encode contextual truth of propositions along linear activation directions across output policies arXiv cs.CL source ↗

Aug 6 Preferred, not safer: pairwise preference is a poor proxy for clinical safety arXiv cs.CL source ↗

Aug 6 Language models reveal urban profiles from anonymized city references using 40 indicators across seven domains arXiv cs.CL source ↗

Aug 6 Stuck on "A": diagnosing and repairing interface injury in attention-to-KDA linearization of a 0.6B language model arXiv cs.CL source ↗

Aug 6 BODHI: do LLMs branch out and discover heterogeneous inferences? arXiv cs.CL source ↗

Aug 6 A long-run persistence theory for AI systems under the redundancy-adjusted artificial age score (AAS) arXiv cs.AI source ↗

Aug 6 Leak-Resistant unlearning: a new benchmark for evaluating multi-hop reasoning consistency and recovery robustness arXiv cs.AI source ↗

Aug 6 Agreement-before-Diversity: a verification-first criterion for heterogeneous language-model coordination arXiv cs.AI source ↗

Aug 6 CAMP: a cycle-aware multi-scale patch mixer for time series forecasting arXiv cs.LG source ↗

Aug 6 Calibrating artificial guilt: neurally grounded reward shaping for prosocial multi-agent reinforcement learning. arXiv cs.AI source ↗

Aug 6 On Hamming-Lipschitz type stability of the subdominant (minmax) ultrametric: theory and simple proofs arXiv cs.LG source ↗

Aug 6 Understanding fault tolerance of adversarially robust pruned models arXiv cs.LG source ↗

Aug 6 Monte Carlo Tree Search for table-to-multimodal report generation arXiv cs.AI source ↗

Aug 6 Trust-region framework for moment estimation constrains update steps in Adam-like optimizers; derives a family of learning-rate mechanisms based on moment constraints arXiv cs.LG source ↗

Aug 6 Interoceptive attention as dynamic homeostatic prioritization in a foraging agent arXiv cs.AI source ↗

Aug 6 C2MOE uses consistency and complementarity-guided mixture of experts to handle incomplete multimodal emotion learning. arXiv cs.LG source ↗

Aug 6 Recurrent Residual Quantization: a PTQ framework for progressive multi-precision LLM representation arXiv cs.LG source ↗

Aug 6 Improving auto-design of neural PDE solvers with a domain-specific language arXiv cs.AI source ↗

Aug 6 Joint UAV flight and opportunistic routing under reinforcement learning for delay-tolerant networks arXiv cs.AI source ↗

Aug 6 From non-convex self-concordant regularization to scalable quasi-Newton training of PINNs arXiv cs.LG source ↗

Aug 6 SpecDrop: parameter-free routing for modular specialization in mixture-of-experts networks arXiv cs.LG source ↗

Aug 6 Tactus: Open-vocabulary object recognition from low-cost pressure arrays arXiv cs.LG source ↗

Aug 6 Lindblad-inspired multi-timescale reservoir computing with separable rotation and dissipation arXiv cs.LG source ↗

Aug 6 NeuMoSync integrates neuron-specific neuromodulation into deep networks to improve plasticity and adaptability in continual learning. arXiv cs.AI source ↗

Aug 6 Random features for Grassmannian kernel approximation with bounded rank-one projections arXiv cs.LG source ↗

Aug 6 Unscented KalmanNet: a hybrid deep learning filter with calibrated posterior covariance for nonlinear state estimation arXiv cs.LG source ↗

Aug 6 AI literacy for legal translation: developing digital resilience arXiv cs.AI source ↗

Aug 6 Comparative study of feature selection methods for EHR diagnosis codes in opioid use disorder prediction arXiv cs.LG source ↗

Aug 6 Attention-Only white-box transformer via LeJEPA-based self-supervised pretraining arXiv cs.LG source ↗

Aug 6 Spatiotemporal graph transformer for traffic intelligence in edge computing arXiv cs.LG source ↗

Aug 6 RAIL Principles for Neurosymbolic AI: Reasoning, Assurances, Interfacing and Learning arXiv cs.AI source ↗

Aug 6 What Is a Skill Worth? Structure-aware Shapley valuation of agent skills arXiv cs.AI source ↗

Aug 6 FinProBench provides a benchmark for professional financial tasks and RGRC derives rubrics from practitioner deliverables to evaluate financial AI agents. arXiv cs.AI source ↗

Aug 6 The LLM proposes, the executive disposes: a self-verifying agent instrument that dissociates commitment drift from binding drift in long-horizon agents. arXiv cs.AI source ↗

Aug 6 CARGO-VL: a group-relative framework for counterfactual arbitration with risk-constrained optimization in vision-language models arXiv cs.AI source ↗

Aug 6 Adversarially robust abductive fusion of pre-trained transformer-based perception models arXiv cs.AI source ↗

Aug 6 Traceable LLM-generated hazard scenarios from ASRS reports for aviation safety analysis arXiv cs.AI source ↗

Aug 6 LiNC: Lightweight Noise Correction via per-sample trust and Gaussian mixture modeling arXiv cs.LG source ↗

Aug 6 Robust and personalized federated learning for aircraft-engine prognostics under benign and adversarial client heterogeneity arXiv cs.LG source ↗

Aug 6 Learning to resolve neutron resonances with fully convolutional neural networks arXiv cs.LG source ↗

Aug 6 MatrAIx: Simulating the world with 8.3 billion persona agents arXiv cs.AI source ↗

Aug 6 MINT: tensor decomposition on stacked recurrence matrices for time series data mining arXiv cs.LG source ↗

Aug 6 LaPrune: controllable differentiable sparsity at million scale. arXiv cs.LG source ↗

Aug 6 BrainBench benchmarks large language models for comprehensive EEG understanding arXiv cs.AI source ↗

Aug 6 SJEPA: learning elegant latent dynamics with hybrid symbolic-neural predictors arXiv cs.LG source ↗

Aug 6 TS2TabPFN: time series classification and extrinsic regression through feature extraction and a tabular foundation model arXiv cs.LG source ↗

Aug 6 Architectural implications of agentic AI workflows; study characterizes fragmentation and heterogeneity in agentic execution across production Azure workflows and open-source frameworks arXiv cs.AI source ↗

Aug 6 A/B Agent: a self-evolving agent for strategy iteration in industrial A/B testing arXiv cs.AI source ↗

Aug 6 An explainable LLM agent layer for open-world anomaly detection in oil wells arXiv cs.LG source ↗

Aug 6 When prompts become pixels: prompt-region grounding for multimodal reasoning arXiv cs.AI source ↗

Aug 6 SafeCommit introduces a risk-controlled layer to certify memory-grounded agents before action under memory uncertainty arXiv cs.AI source ↗

Aug 6 MERaLiON-GR: a speech gender recognition model for English and SEA languages uses LoRA fine-tuning on MERaLiON-SpeechEncoder-2 arXiv cs.CL source ↗

Aug 6 Towards end-to-end multilingual metaphor processing: integrating detection, translation, and evaluation arXiv cs.CL source ↗

Aug 6 Output-budget regimes change the measured multilingual reasoning gap in MGSM for Qwen3-8B and Llama-3.1-8B-Instruct across prompting budgets arXiv cs.CL source ↗

Aug 6 Eliciting intrinsic hallucinations in LLMs via semantically equivalent adversarial attacks arXiv cs.CL source ↗

Aug 6 DataRx: missingness-aware sampling for safer large language model task-specific fine-tuning arXiv cs.CL source ↗

Aug 6 Visualizing graph-to-answer mechanism recovery in materials-science hypothesis generation arXiv cs.CL source ↗

Aug 6 MAP-PO: learning sexism detection by preserving perspectivist annotations on EXIST 2024 tweets arXiv cs.CL source ↗

Aug 6 Large Language Models for low-resource languages: a conceptual framework for an electronic explanatory dictionary of the Tajik language arXiv cs.CL source ↗

Aug 6 Right Reset: Chunking by Prefix Removal arXiv cs.CL source ↗

Aug 6 The Calibration Floor: format repair can masquerade as self-correction at small-to-mid scale arXiv cs.CL source ↗

Aug 6 Patients-like-me: a variational LM–GNN framework for explainable clinical prediction arXiv cs.CL source ↗

Aug 6 Searching for sound-meaning collisions: graph-based affordance retrieval and multi-evaluator ranking for pun translation at CLEF 2026 JOKER Task 2 arXiv cs.CL source ↗

Aug 6 Test whether language models implement an in-context conditional rule via a runtime circuit across models and languages arXiv cs.CL source ↗

Aug 6 Reconstructing persistent worlds from narratives for narrative-grounded interactive experiences arXiv cs.CL source ↗

Aug 6 Hallucinations on the board: tool-augmented evaluation of LLM chess commentary arXiv cs.CL source ↗

Aug 6 FinReportBench benchmarks and improves institution-grade financial report generation using a 35-item rubric arXiv cs.CL source ↗

Aug 6 EdgeLM uses edge demonstrations for improving table understanding in language models arXiv cs.CL source ↗

Aug 6 NOLLI: a procedurally generated English-Korean puzzle benchmark for diagnosing the English-Korean performance gap. arXiv cs.CL source ↗

Aug 6 Equitable system-prompt selection via constrained mixed-strategy groupDRO arXiv cs.CL source ↗

Aug 6 When more becomes less: position-dependent repetition effects in language models arXiv cs.CL source ↗

Aug 6 Social pressure breaks majority voting in LLM safety panels arXiv cs.CL source ↗

Aug 6 The fairness collapse phenomenon: bias amplification in language models trained on synthetic data arXiv cs.CL source ↗

Aug 6 Transfer learning for named entity recognition of classical Latin through LLM prompting arXiv cs.CL source ↗

Aug 6 MIDAS: Multi-LLM Iterative Data-Adaptive Summarization arXiv cs.CL source ↗

Aug 5 Shared organizational memory for enterprise coding agents: system design and deployment snapshot arXiv cs.AI source ↗

Aug 5 SIRIN: a unified toolkit for detecting contextual hallucinations in retrieval-augmented and memory-grounded LLM systems arXiv cs.AI source ↗

Aug 5 Energy efficiency of locally deployed LLMs: a preliminary quantitative GPU power benchmark on consumer hardware arXiv cs.AI source ↗

Aug 5 Personalizing large language model agents with small policy models arXiv cs.AI source ↗

Aug 5 Linguistic context recodes visual representations in vision-language models arXiv cs.AI source ↗

Aug 5 H+ Embedding: harmonizing global and token-level retrieval with context-dependent phrases arXiv cs.AI source ↗

Aug 5 CoT-Core: accelerating LLM evaluation via CoT-aware coreset selection arXiv cs.AI source ↗

Aug 5 TRACE-TS: attribution-grounded and traceable sensor-language reasoning for human activity understanding arXiv cs.AI source ↗

Aug 5 Motif-Mamba: network motif improved mamba for long-range sequence modeling arXiv cs.AI source ↗

Aug 5 Nova: an end-to-end MLIR compiler for deep learning arXiv cs.AI source ↗

Aug 5 Agentic coding in the wild: characterization of GitHub Copilot traces at production scale arXiv cs.AI source ↗

Aug 5 Request-level energy attribution for batched LLM serving arXiv cs.AI source ↗

Aug 5 Geometric self-supervised pre-training improves neural combinatorial optimization generalization to larger TSP instances arXiv cs.AI source ↗

Aug 5 Structural limits persist for homogeneous multi-agent groundedness verification, even with a three-agent panel across six benchmarks. arXiv cs.AI source ↗

Aug 5 Optimization and constraint modeling using LLMs with a retrieval augmented generation process arXiv cs.AI source ↗

Aug 5 AgentStream evaluates self-evolving LLM agents under streaming tasks arXiv cs.AI source ↗

Aug 5 Revisiting classic thought experiments to measure consciousness for artificial intelligence safety arXiv cs.AI source ↗

Aug 5 RF-HOI uses radio frequency signals to recognize human-object interactions (HOI) without vision-based inputs arXiv cs.AI source ↗

Aug 5 WM-Cov: Test adequacy for interactive world-model-style autonomous driving simulation arXiv cs.AI source ↗

Aug 5 RAG-TESTER: automated end-to-end testing of retrieval-augmented large language models arXiv cs.AI source ↗

Aug 5 Enhancing LLMs with context-specific knowledge for mitigating misinformation in SMEs using RAG-based modeling and analysis arXiv cs.AI source ↗

Aug 5 Wiring beats blending: what transfers between transformer sizes — and what doesn’t arXiv cs.LG source ↗

Aug 5 Population-robust feature selection via generalized welfare optimization arXiv cs.LG source ↗

Aug 5 Designing a good virtual node: addressable and cardinality-preserving global memory for message passing architectures arXiv cs.LG source ↗

Aug 5 Geodesic Normalization keeps hidden states on a hypersphere by orthogonalizing layer outputs and updating via the Riemannian exponential map arXiv cs.LG source ↗

Aug 5 Measuring explainer stability via attribution separability arXiv cs.LG source ↗

Aug 5 Topological simplification in predictive coding networks across layers measured with persistent homology on MNIST and synthetic data arXiv cs.LG source ↗

Aug 5 LLMs can annotate attribution graphs arXiv cs.LG source ↗

Aug 5 NANQ: Noise-floor-aware mixed-precision non-uniform quantization for analog compute-in-memory arXiv cs.LG source ↗

Aug 5 Adaptive sampling for automated post-disaster rapid damage assessment via level-set cost-aware Bayesian optimization arXiv cs.LG source ↗

Aug 5 Evaluation blindness: silent measurement failures can hide AI system faults from training to deployment arXiv cs.LG source ↗

Aug 5 Can training logs make model comparisons more precise? arXiv cs.LG source ↗

Aug 5 PatTree: a novel approach for automated creation of multimodal, graph-based patient representations for medical classification tasks arXiv cs.LG source ↗

Aug 5 Verifier-Guided model discovery for physical dynamical systems with pretrained symbolic transformers arXiv cs.LG source ↗

Aug 5 Deep Divide-and-Reduce in Symbolic Regression arXiv cs.LG source ↗

Aug 5 Neural networks with local converging inputs improve efficiency of multi-asset options pricing by locally correcting coarse-mesh solutions with a minimal high-precision input. arXiv cs.LG source ↗

Aug 5 GeoID-PINN models identifiability-aware regional epidemic inference with geographic coupling using a physics-informed neural network for SIRD dynamics arXiv cs.LG source ↗

Aug 5 PhenMol: a structure-preserving framework for phenotypic drug discovery in learning molecular representations from cellular phenotypes arXiv cs.LG source ↗

Aug 5 Contrast-invariant deep ptychography neural networks arXiv cs.LG source ↗

Aug 5 Output-Aware Rotation for INT2 KV-Cache Quantization arXiv cs.LG source ↗

Aug 5 CT-HEG: a bidirectional, timestamp-attributed event graph for ICU in-hospital mortality prediction; an architectural ablation study arXiv cs.LG source ↗

Aug 5 Multimodal auto-regressive transformer surrogate models variable well operations and quantifies uncertainty in geological carbon storage arXiv cs.LG source ↗

Aug 5 Maglev: Sliding Recurrent Memory uses a recurrent Transformer with fixed-size memory to generalize sliding-window attention; introduces a two-model architecture with a prefiller and memory model arXiv cs.LG source ↗

Aug 5 GoT-CD: Graph-of-thoughts causal discovery and the fragility of post-hoc path-specific fairness audits arXiv cs.LG source ↗

Aug 5 NOMADD: numerical optimization of models adapting to data drift arXiv cs.LG source ↗

Aug 5 Globe: Trajectory-aligned gradient matching with structured sparse optimization for coreset selection arXiv cs.LG source ↗

Aug 5 MedPIC-Bench benchmark evaluates counterfactual sensitivity to patient information in medication-safety reasoning arXiv cs.AI source ↗

Aug 5 PULSE: an executable contract language for spatiotemporal knowledge graph engineering arXiv cs.AI source ↗

Aug 5 ISEE: Interactive Semantic Enrichment for Database Fields enhances data sense-making by improving semantic clarity of field descriptions through LLM-based agents. arXiv cs.AI source ↗

Aug 5 Information boundaries for group-robust LLM pruning; a reproducible compression statistic can select the wrong candidate, with a conic law modeling the pooling price for positive linear fixed-candidate damage. arXiv cs.AI source ↗

Aug 5 HyperAgent: planning and acting over tool-schema hypergraphs for tool-use LLM agents arXiv cs.AI source ↗

Aug 5 Beyond the hivemind: escaping LLM homogeneity via meta-persona anchoring and sequential temperature scaling arXiv cs.AI source ↗

Aug 5 Predictive set theory: a generative framework for cognitive architecture with operationalized core mechanisms arXiv cs.AI source ↗

Aug 5 Towards a new paradigm of scientific discovery with socialized artificial intelligence arXiv cs.AI source ↗

Aug 5 BAP-SQL: Budget-aware observation planning for agentic text-to-SQL arXiv cs.AI source ↗

Aug 5 Hypercubes, hyperplanes, and constraint-induced complexity collapse in atomic concept learning arXiv cs.AI source ↗

Aug 5 DiffImaginE introduces a multimodal named entity recognition verifier using diffusion imagery to assess entity-type compatibility arXiv cs.AI source ↗

Aug 5 AI Agent Economics: can autonomous economic behavior emerge among AI agents under minimal external conditions? arXiv cs.AI source ↗

Aug 5 TraceCAD: trace-guided repair for agentic CAD generation arXiv cs.AI source ↗

Aug 5 LoCA: Forward-only LLM tuning after one-shot calibration with local credit assignment arXiv cs.AI source ↗

Aug 5 ProPRL: property-aware prerequisite relation learning in educational knowledge graphs. arXiv cs.AI source ↗

Aug 5 Getting the parameters right: a difficulty-graded benchmark and probe-guided training for LLM tool calls arXiv cs.AI source ↗

Aug 5 Missing data layer in Latin American AI infrastructure; proposes DataHub to address dataset discovery and supply challenges arXiv cs.AI source ↗

Aug 5 Interpreting black-box large language models with sentence-level energy landscapes arXiv cs.AI source ↗

Aug 5 VeriTrace: human-like temporal exploration completes agentic action space arXiv cs.AI source ↗

Aug 5 Explainable AI for the EU Right to Explanation: a systematic review of the law–XAI translation gap arXiv cs.AI source ↗

Aug 5 UrbanAgent: a tool-augmented agent for cross-system urban tasks arXiv cs.AI source ↗

Aug 5 Self-Organising Digital Circuits demonstrate meta-learning-based functional logic generation on graphs for adaptive circuit maintenance arXiv cs.AI source ↗

Aug 4 Machine-learning-guided pipeline enables on-demand solvent candidate generation and testing of electrolyte recipes; SEM images show sodium-metal deposit morphologies for three candidates MIT News (AI) source ↗

Aug 4 Benefits of medical AI assistance vary by user expertise, with non-experts trusting AI advice more than clinicians who recognize AI mistakes. MIT News (AI) source ↗

Aug 4 Dual-penalty evasion framework to fool white-box explainable AI auditors; arXiv:2608.00566v1 reports attacks on post-hoc explainers like LIME, SHAP, and Integrated Gradients arXiv cs.LG source ↗

Aug 4 RubricReviewer: converting direct critique into rubric-driven peer review to make reviews objective and comprehensive arXiv cs.CL source ↗

Aug 4 Response magnitude as a dominant signal for held-out CRISPRi perturbation effect prediction arXiv cs.LG source ↗

Aug 4 Cost-effective automated judging of natural-language mathematical proofs using open-weight models aligns with human grading decisions on IMO-GradingBench instances arXiv cs.CL source ↗

Aug 4 DLLM-TTS: Block discrete diffusion language model for text-to-speech synthesis arXiv cs.CL source ↗

Aug 4 Obshazard-bench: benchmarking multimodal foundation models for real-time disaster intelligence from raw earth observation streams arXiv cs.CL source ↗

Aug 4 SLMs as multi-agent routers: a progressive SFT and reinforcement learning approach arXiv cs.CL source ↗

Aug 4 A Constitution-Grid instrument for data-efficient RL alignment (C-Guard) arXiv cs.CL source ↗

Aug 4 MemoryForge: Synthesize lifelong memory for human-like LLM agents arXiv cs.CL source ↗

Aug 4 Ensemble of unsupervised deep learning for clustering imbalanced tabular data arXiv cs.LG source ↗

Aug 4 Boosting diversity in text diffusion models via entropy-based guidance arXiv cs.CL source ↗

Aug 4 Modeling unknown nonlocal PDE systems via flow-map learning arXiv cs.LG source ↗

Aug 4 Verifier-induced support reshaping in on-policy optimization shows verifiable rewards shaping trajectories and reducing successful behaviors for later objectives in RLVR arXiv cs.LG source ↗

Aug 4 Targeted low-rank adaptation via causal layer selection to bridge English-Arabic medical knowledge gap arXiv cs.CL source ↗

Aug 4 Averaging bias: human faithfulness annotations are not locally faithful arXiv cs.CL source ↗

Aug 4 Agentic Graph Token Reasoning arXiv cs.LG source ↗

Aug 4 Textual descriptors predict early-stage startup exit using a 850-feature text-based framework on 7,419 startups over 20 years. arXiv cs.CL source ↗

Aug 4 CoSynFlow: conformal symplectic neural flows for cross-system prediction of dissipative Hamiltonian dynamics arXiv cs.LG source ↗

Aug 4 Agentic Bayesian Optimization through surrogate-augmented autoresearch arXiv cs.LG source ↗

Aug 4 SeDeM: selective decompression of hidden-state memories for long-context question answering arXiv cs.CL source ↗

Aug 4 Similarity-aware machine unlearning arXiv cs.LG source ↗

Aug 4 Fairness auditing: quantify unavoidable post-audit manipulation under finite resources arXiv cs.LG source ↗

Aug 4 Stabilized best-of-K training for neural combinatorial optimization improves Leader Reward with a stabilized rank signal for POMO on TSP-100; achieves 7.7662 under 100-start, 8-augmentation greedy decoding arXiv cs.LG source ↗

Aug 4 Trustworthiness costs of domain adaptation in small language models; a cross-architecture empirical study arXiv cs.CL source ↗

Aug 4 Neural operator learning for collision-aware trajectory planning of spacecraft swarms arXiv cs.LG source ↗

Aug 4 DSETA: a dual-stage continual learning framework for travel time prediction in dynamic traffic environments arXiv cs.LG source ↗

Aug 4 Inference-time policy alignment for fair reinforcement learning arXiv cs.LG source ↗

Aug 4 Comparing and modeling argumentation in German political communication across arenas arXiv cs.CL source ↗

Aug 4 LLantia: an LLM-based method for automated neural circuit inference and analysis arXiv cs.CL source ↗

Aug 4 What transfers from text to vision? Capability-Driven Multimodal Scaling Law and transfer dynamics for VLMs arXiv cs.CL source ↗

Aug 4 Learning compositional meta-routing for agentic workflows; an executable benchmark arXiv cs.LG source ↗

Aug 4 HP-JEPA: hierarchical partitioning for multi-resolution graph joint-embedding predictive learning arXiv cs.LG source ↗

Aug 4 AutoCause is an open-source Python workflow that automates expert decisions in environmental time-series causal discovery. arXiv cs.LG source ↗

Aug 4 Role steering of language models for social simulations arXiv cs.CL source ↗

Aug 4 Cross-task dissociation in frontier vision-language model theory of mind arXiv cs.CL source ↗

Aug 4 Sixteen language models yield 1.69 distinct formulations on a psychotherapeutic case; ensembles offer multiple perspectives before a decision arXiv cs.CL source ↗

Aug 4 Leak It: a probabilistic approach to training-data extraction from black-box language models arXiv cs.LG source ↗

Aug 4 Rethinking pretraining for specialized design data: evidence from the JONES-19 cultural design dataset arXiv cs.LG source ↗

Aug 4 LLM-OSDA: an optimal-stopping dynamic auction for native advertising in multi-turn LLM conversations arXiv cs.CL source ↗

Aug 4 Progressive^2: a teacher-student progressive co-evolving knowledge distillation method for substantial model compression arXiv cs.LG source ↗

Aug 4 Physics-chemistry-informed neural network (PCINN) enables real-time SALD surface coverage prediction and reliable kinetics inversion arXiv cs.LG source ↗

Aug 4 A Few Neurons Reveal When LLMs Misuse Tools: Sparse Detection and Selective Steering for Reliable Tool Use arXiv cs.CL source ↗

Aug 4 CurveShift: Is Agent Progress Scalar? Separating Level from Shape arXiv cs.CL source ↗

Aug 4 Abstention as an action can kill both the reward gradient and the KL anchor: collapse law and repair for error-penalized reinforcement learning arXiv cs.LG source ↗

Aug 4 DiffusionGemma: an experimental open-weight language model uses discrete diffusion to generate text in parallel by refining blocks of 256 tokens; authors fine-tune mixture-of-e gates? Wait ensure accuracy. We must avoid inventing. The abstract ends with "fine-tuning the mixture-of-e" truncated. We can't infer. So craft headline: "DiffusionGemma introduces an experimental open-weight language model that uses discrete diffusion to generate text in parallel by refining blocks of 256 tokens." That stays within provided. Might be okay. Include arXiv 2608.00146? Not necessary. It's a headline style. Keep sentence case. Under ~200 chars. Let's produce: "DiffusionGemma introduces an experimental open-weight language model that uses discrete diffusion to generate text in parallel by refining blocks of 256 tokens." DiffusionGemma introduces an experimental open-weight language model that uses discrete diffusion to generate text in parallel by refining blocks of 256 tokens. arXiv cs.CL source ↗

Aug 3 Alexander Rakhlin named director of the MIT Statistics and Data Science Center. MIT News (AI) source ↗

Aug 3 Orchard: an open framework for scalable agentic AI Microsoft Research source ↗

Aug 3 Safety, or just capability? A validity audit of agent-safety benchmarks arXiv cs.AI source ↗

Aug 3 SEDR-Seq2P: a lightweight dilated residual sequence-to-point network for multi-task industrial NILM arXiv cs.LG source ↗

Aug 3 Benchmarks are not monolithic: sample-level auditing and orchestration for LLM evaluation show dataset-centric meta-evaluation along five latent dimensions arXiv cs.CL source ↗

Aug 3 Representations from pretrained machine-learning interatomic potentials as coarse coordinates for material generation and evaluation arXiv cs.LG source ↗

Aug 3 Mitigating class-tail undercoverage in medical vision-language models under clinical shift arXiv cs.LG source ↗

Aug 3 MPP-GNN: Subject-Adaptive Community Detection for fMRI-Based Alzheimer's Disease Classification arXiv cs.LG source ↗

Aug 3 Knowledge distillation in small instruction-tuned LMs has asymmetric effects on bias, improving context-following for unambiguous tasks but degrading calibration for ambiguous tasks. arXiv cs.CL source ↗

Aug 3 EarlyDx: an admission-anchored benchmark for open-ended generation of evidence-supported ED-encounter diagnoses arXiv cs.AI source ↗

Aug 3 TextCloak defends against unauthorized LLM exploitation by RL-driven unlearnable text arXiv cs.CL source ↗

Aug 3 Self-Supervised Skill Optimization learns reusable agent skills from unlabeled task instances alone arXiv cs.CL source ↗

Aug 3 Fast rates for swap-agnostic learning of proper losses arXiv cs.LG source ↗

Aug 3 Predicting steel fatigue life from micrographs using physics-informed deep learning arXiv cs.LG source ↗

Aug 3 Reasoning in real world clinical care: why large language models are not yet safe for autonomous clinical decision support; corroborating coverage notes real-world credibility tests of LLM reasoning arXiv cs.AI source ↗

Aug 3 Sensitivity analysis of GRU, LSTM and Transformer encoder in classification of automated driving systems arXiv cs.LG source ↗

Aug 3 PARALLEL: a prefrontal-aligned reinforcement-inspired approach for language-model learning under explicit limits arXiv cs.CL source ↗

Aug 3 Fragility of value under imperfect alignment; model analyzes how optimizing for an imperfect proxy can yield catastrophic outcomes arXiv cs.AI source ↗

Aug 3 Machine learning approaches for enhancing decision-making in complex discrete choice tasks show potential for policy-based preference elicitation over traditional parametric models arXiv cs.LG source ↗

Aug 3 Adaptivity via a parallel architecture for stochastic gradient methods arXiv cs.LG source ↗

Aug 3 Reflection or re-generation? Why LLM revision fails where human revision succeeds arXiv cs.LG source ↗

Aug 3 Topology-aware data movement for disaggregated GPU inference arXiv cs.LG source ↗

Aug 3 Distilling knowledge from large language models into lightweight reinforcement learning agents for autonomous cyber operations arXiv cs.LG source ↗

Aug 3 LARA: Lightweight Additive Residual Adaptation operates in the residual stream for model adaptation without updating base weights; evaluated on a code fine-tuning task and on preference optimization (DPO) arXiv cs.LG source ↗

Aug 3 ViSAGE: constructing self-correcting memories for long-form video understanding arXiv cs.AI source ↗

Aug 3 The Morphological core of Dungan: a two-dialect finite-state model and a multi-genre evaluation arXiv cs.CL source ↗

Aug 3 Identifying informative environments for cognition parameter inference via Bayesian experimental design. arXiv cs.AI source ↗

Aug 3 Ontology-guided deduplication-aware extraction layer for knowledge graph construction from heterogeneous documents arXiv cs.AI source ↗

Aug 3 Gated Q-learning: add off-policy bias to taste arXiv cs.LG source ↗

Aug 3 Model or harness? An interaction-centric taxonomy for localizing agent failures arXiv cs.AI source ↗

Aug 3 Feature interaction modeling for physics-informed neural networks and neural operators arXiv cs.LG source ↗

Aug 3 Can AI evaluate AI scientists? A benchmarking study of autonomous research generation systems using automated multi-model review arXiv cs.AI source ↗

Aug 3 TAPR: a task-aware prompt rewriter improves downstream LLM performance by reformulating user prompts into task-optimized prompts using reinforcement learning with GRPO arXiv cs.AI source ↗

Aug 3 MMFGU: Multimodal Federated Graph Unlearning enables fine-grained deletion in multimodal federated learning; arXiv paper outlines approach arXiv cs.LG source ↗

Aug 3 ZeroR@CHiPSAL 2026: two-stage vision-language adaptation with contrastive learning for Nepali meme classification arXiv cs.CL source ↗

Aug 3 Entropy-based CoT pruning offers no advantage over random pruning; low-entropy token retention is only briefly effective, per tests across models and tasks. arXiv cs.CL source ↗

Aug 3 Mixture-of-Translators: translating KV caches across heterogeneous large language models arXiv cs.CL source ↗

Aug 3 What must be true before AI ships in a regulated firm arXiv cs.CL source ↗

Aug 3 Token-Level diagnosis of sycophancy in LLMs with attribution-guided steering arXiv cs.CL source ↗

Aug 3 NeSyFS: A Neuro-symbolic fast-slow thinking framework for LLM agent under partial observability arXiv cs.AI source ↗

Aug 3 MerchantBench benchmarks LLM agents for long-term coherence in e-commerce operations arXiv cs.AI source ↗

Aug 3 Preference-Optimized LLM counselors trade goal persistence for relational attunement in motivational interviewing arXiv cs.CL source ↗

Aug 3 LLMs tested for predicting item difficulty levels in large-scale tests; study compares prompting strategies and models across LLMs and encoder-only baselines. arXiv cs.CL source ↗

Aug 3 MMShopBench: a real-log benchmark for multimodal, multi-turn shopping agents arXiv cs.AI source ↗

Aug 3 Flow matching with missing data arXiv cs.LG source ↗

Aug 3 TAGTorch: a PyTorch library for geometry, topology, and symmetry-aware machine learning arXiv cs.LG source ↗

Aug 3 Evidence-grounded constraint checking in construction documents uses a pipeline that normalizes facts, applies four-state rules, and retains source spans; evaluated on 160 reference-based tasks from 29 projects. arXiv cs.AI source ↗

Aug 3 Guarantees on dynamical system distinguishability for LLM token generation arXiv cs.LG source ↗

Aug 3 Lawful: Law-aligned witness for faithful use of latents explores interpretability gaps in physics-law representations arXiv cs.LG source ↗

Aug 3 BLADE: Boundary-expanded and layer-adaptive dynamic exit for efficient LLM reasoning arXiv cs.CL source ↗

Aug 3 How hard does it think? Analyzing step-aware reasoning energy in LLM chain-of-thought trajectories arXiv cs.AI source ↗

Aug 3 Empowering cross-domain sequential recommendation with hybrid tokenization and serial-parallel decoding arXiv cs.AI source ↗

Aug 3 Mirror learning: acquiring actionable policies from passive observation through third-person demonstrations arXiv cs.LG source ↗

Aug 3 Stateful knowledge learning enables LLM agents to move from trajectory-level reflection to predictive foresight arXiv cs.CL source ↗

Aug 3 LLM framework for discovering major mathematical conjectures: AI's quest for the next Riemann hypothesis arXiv cs.AI source ↗

Aug 3 Evaluating federated pre-training: reliability of downstream fine-tuning and intrinsic evaluation arXiv cs.CL source ↗

Aug 3 Multi-agent planning with spatio-temporal and topological constraints using STL-GO arXiv cs.AI source ↗

Aug 3 On the generalization of steering vectors for chain-of-thought faithfulness arXiv cs.AI source ↗

Aug 3 Context-preserving organization of inline notes and collected commentaries in classical Chinese texts for exegetical knowledge; NLP framework proposes preserving contextual dependencies while automated compilation arXiv cs.CL source ↗

Aug 3 Chain-of-Models: cross-model auditing for bias-robust LLM judges arXiv cs.CL source ↗

Aug 3 Hypergradient-based bilevel reinforcement learning with improved sample complexity arXiv cs.LG source ↗

Aug 3 Hierarchical Copula-Gumbel-Top-K routing: two-sided dependence control for frozen mixture-of-experts at fixed per-token routing laws arXiv cs.LG source ↗

Aug 3 TELLER: Dual-Path Iterative Preference Optimization for Table Entity Linking arXiv cs.CL source ↗

Aug 3 Imbalanced data clustering via targeted data augmentation using GMM and LLM arXiv cs.CL source ↗

Aug 3 Scaling scientific discovery environments for turn-level agentic RL arXiv cs.AI source ↗

Aug 3 ThinkReset: Learnable intermediate interface construction for bounded-context long-horizon reasoning arXiv cs.AI source ↗

Aug 3 TokenSwap benchmarks and reduces the modality gap in multimodal LLMs arXiv cs.CL source ↗

Aug 3 Technological advances in detecting and managing cognitive impairment in older adults: trends, challenges, and future directions arXiv cs.LG source ↗

Aug 3 OpenClaw and Ollama in agentic AI: toward fully autonomous and scalable AI agent systems arXiv cs.AI source ↗

Aug 3 Best Friends, Not Forever: Evaluating long-horizon persona collapse and behavioral drift in AI companions arXiv cs.AI source ↗

Aug 3 Learning optimal dynamic matching via graph neural networks arXiv cs.LG source ↗

Aug 3 Library reachability in LSR-Synth: anti-memorization design changes measurement of symbolic discovery arXiv cs.AI source ↗

Aug 3 FairFund-Bench evaluates distributive bias in LLM resource allocation. arXiv cs.CL source ↗

Aug 3 SciToolAgent-Evo: an ontology-aware self-evolving agent for open-world scientific tool acquisition arXiv cs.AI source ↗

Aug 3 The Formalism Trap: LLM-as-a-Judge evaluators conflating proceduralism with semantic truth under social load arXiv cs.CL source ↗

2026-07 →