LIVE · refreshes every 20 min

updated Sep 16, 04:23 UTC

110010.art / archive / research / 2026-07

Research — July 2026

Jul 31 Evidence-ledger adjudication for claim-evidence traceability in AI agents assesses support relations and routes unsupported or contradicted claims back to the author arXiv cs.AI source ↗

Jul 31 Latent channels in multi-agent LLMs are evaluated for actual communication through a causal audit of latent messages. arXiv cs.AI source ↗

Jul 31 EvoPINN: agentic discovery of executable algorithms for physics-informed neural networks arXiv cs.AI source ↗

Jul 31 CaM-Wolf: Causal-Aware Multimodal Agents for Social Deduction Games arXiv cs.AI source ↗

Jul 31 UrbanDS: a graph-guided LLM multi-agent system for data-intensive urban tasks arXiv cs.AI source ↗

Jul 31 Objective misalignment in mixed-motive LLM multi-agent systems evaluated with Werewolf game arXiv cs.AI source ↗

Jul 31 Rethinking self-evolution: a constrained exploration-exploitation process for mitigating skill overfitting arXiv cs.AI source ↗

Jul 31 AI agents discover statistical mechanical mappings from a raw partition function to a tractable representation in six Ising-type problems; StatMechBench-v0 benchmark introduced arXiv cs.AI source ↗

Jul 31 GuideSkill: evolving executable LLM agent skills for guideline-grounded clinical reasoning arXiv cs.AI source ↗

Jul 31 MultivationBench: a benchmark for multimodal sequential motivation reasoning arXiv cs.AI source ↗

Jul 31 ClinLens: a benchmark of 200 executable tasks over five linked MIMIC resources for longitudinal multimodal clinical data science arXiv cs.AI source ↗

Jul 31 CG-World: a large-scale world-state dataset and protocol for world models arXiv cs.AI source ↗

Jul 31 Passive video to editable experience: Pegasus translates human demonstrations into robot-learnable data through structured knowledge transfer arXiv cs.AI source ↗

Jul 31 Representational quality accounts for RL models’ superior mathematical reasoning performance over SFT fine-tuned models arXiv cs.AI source ↗

Jul 31 Evaluation scores are perishable knowledge claims arXiv cs.AI source ↗

Jul 31 When benchmark inferences do not compose: projectibility in AI evaluation arXiv cs.AI source ↗

Jul 31 TraceCoder enables explainable and auditable code generation with position-key snippet versioning arXiv cs.AI source ↗

Jul 31 GoGoTB: agentic RTL verification with specification-grounded coverage closure arXiv cs.AI source ↗

Jul 31 What Does It Take to Detect an AI Agent? Minimal feature sets for behavioral detection under browser automation arXiv cs.AI source ↗

Jul 31 Eco3S: a socio-economic system simulation framework for agent-based modeling and policy analysis arXiv cs.AI source ↗

Jul 31 AlphaSchema: exploring the space of trading semantics for LLM-based alpha mining arXiv cs.AI source ↗

Jul 31 Property-driven causal abstractions for Markov decision processes arXiv cs.AI source ↗

Jul 31 Belief-guided decision making with uncertainty gating in the game of Go arXiv cs.AI source ↗

Jul 31 Fewer clarifications, better code: benchmarking cross-session personalized ambiguity adaptation in coding assistants arXiv cs.AI source ↗

Jul 31 ICLE++: modeling fine-grained traits for holistic essay scoring arXiv cs.CL source ↗

Jul 31 Systematic evaluation of 41 open-weight language models for zero-shot intent classification across eight datasets (135M–9B parameters) and 15 model families arXiv cs.CL source ↗

Jul 31 Harness-G: a graph-structured harness for search agents arXiv cs.CL source ↗

Jul 31 Kinetics of training describe a driven-nucleation rate law for emergence, plasticity loss, and circuit control in language models arXiv cs.LG source ↗

Jul 31 From single- to cross-document: benchmarking multi-granularity event analysis of large language models arXiv cs.CL source ↗

Jul 31 Rethinking EEG-based disease diagnosis: decoupling instance representation learning from subject-level supervision arXiv cs.LG source ↗

Jul 31 Grounded agentic extraction and expert-adjudicated evaluation of intertextuality in classical Chinese histories arXiv cs.CL source ↗

Jul 31 Recursive transformers for semiconductor thermo-mechanical reliability arXiv cs.LG source ↗

Jul 31 Modeling decisions in blockchain analytics: a leakage-aware evaluation of tree-based vs. sequential models arXiv cs.LG source ↗

Jul 31 SE(3)-MeanFlow: few-step protein backbone generation on Lie groups arXiv cs.LG source ↗

Jul 31 AHA-Memes: a fine-grained multimodal benchmark for understanding hate in Arabic memes arXiv cs.CL source ↗

Jul 31 FunL2O: LLM-guided feature function design for learning to optimize arXiv cs.LG source ↗

Jul 31 THGFM: Dual-Branch Temporal Heterogeneous Graph Fusion Model arXiv cs.LG source ↗

Jul 31 Good rankers, bad objectives: bilinear contrastive critics under expressive policy search arXiv cs.LG source ↗

Jul 31 DualAnchor: preserving language priors and improving lexical fidelity in gloss-free sign language translation arXiv cs.CL source ↗

Jul 31 Context-informed ship trajectory prediction via conditional attention arXiv cs.LG source ↗

Jul 31 ZUNA1.1: a 380M-parameter diffusion autoencoder for flexible EEG reconstruction across variable lengths, channels, and intervals arXiv cs.LG source ↗

Jul 31 DoTime: a synthetic benchmark generator for interventional and counterfactual time series arXiv cs.LG source ↗

Jul 31 AI-assisted pre-review of open-source submissions used at BOSC 2026 to help volunteer reviewers by pre-reviewing abstracts for certain criteria arXiv cs.CL source ↗

Jul 31 BridgeAlign: Bridging preference alignment for humanities and social sciences arXiv cs.CL source ↗

Jul 31 Sympathetic framing: evaluating AI alignment across sociodemographic groups arXiv cs.CL source ↗

Jul 31 Prompt chaining in practice: a case study in automated scholarly report generation arXiv cs.CL source ↗

Jul 31 LVLMs evaluated on perception and reasoning jointly to uncover truth behind visual illusions arXiv cs.CL source ↗

Jul 31 Same facts, different diagnosis: measuring and mitigating narrative anchoring in clinical language models arXiv cs.CL source ↗

Jul 31 Gradient-free task-conditioned retrieval for on-device in-context learning arXiv cs.CL source ↗

Jul 31 Flat Score, Amplified Failures: Quantization to 4-bit weights hides damage in multi-turn tool-calling LLM agents arXiv cs.LG source ↗

Jul 31 SkillSmith: learning to compose parametric skills and textual knowledge arXiv cs.CL source ↗

Jul 31 B1ade: 335M embedding model and 1B parameter small language models for minimalist RAG arXiv cs.CL source ↗

Jul 31 RLPF: reinforcement learning from performance feedback for code generation arXiv cs.LG source ↗

Jul 31 Compression-based behavioral similarity enables open-world Sybil discovery on Ethereum arXiv cs.LG source ↗

Jul 31 TIER-MoE: trust-informed expert routing via conditional modality risk for multimodal fusion in biomedical classification arXiv cs.LG source ↗

Jul 31 Benchmarking LLM competence on logical inference over probability operators arXiv cs.CL source ↗

Jul 31 Adam converges under heavy-tailed noise in the plain vector-form setting, with guarantees for stochastic gradients having bounded p-th moment (p in (1,2]). arXiv cs.LG source ↗

Jul 31 Recall Before You Rank: Similarity-Guided Top-K Reuse for Efficient Long-Context Attention arXiv cs.CL source ↗

Jul 31 ECG-InterpBench benchmarks the interpretability of ECG foundation-model representations using matched-scale sparse autoencoders arXiv cs.LG source ↗

Jul 31 Explorative modeling: unlocking a third pretraining axis and end-to-end generation arXiv cs.LG source ↗

Jul 31 Structure-aware data organization for efficient LLM post-training arXiv cs.LG source ↗

Jul 31 Training skills like parameters via self-supervised semantic diffusion arXiv cs.CL source ↗

Jul 31 Benchmarking the residual: long-horizon evaluations reveal degradation beyond short-task performance arXiv cs.LG source ↗

Jul 31 AWARE-FX: an auditable AI/NLP decision-support system converts corporate annual report text into traceable foreign-exchange hedging disclosure measures arXiv cs.CL source ↗

Jul 30 Daniela Rus receives the Bavarian Minister-President's High-Tech Prize. MIT News (AI) source ↗

Jul 30 Capitol Hill hosts Congressional Visit Days as researchers connect with policy makers; participants pose on the Capitol steps with Senator Alex Padilla. MIT News (AI) source ↗

Jul 30 Echoverse: deep, evolving environments for computer-use agents Microsoft Research source ↗

Jul 30 Examining the efficacy of graph neural network message-passing in regression contexts arXiv cs.LG source ↗

Jul 30 Automorphism-induced non-canonicity in top-k explanations of graph neural networks arXiv cs.LG source ↗

Jul 30 Misalignment has a personality: a Big Five account of emergent misalignment arXiv cs.CL source ↗

Jul 30 Characterizing human-likeness in AI generated poetry: a zero-shot classification study arXiv cs.CL source ↗

Jul 30 MeRLa: Meta-learned reward shaping for reinforcement learning from human feedback arXiv cs.LG source ↗

Jul 30 WikiLoop jointly learns to build and navigate an agent-native wiki with downstream feedback arXiv cs.CL source ↗

Jul 30 Primary source headline: Do Methods Support the Claims? Intra-Paper Verification for Peer Review Rewritten headline: Do methods support the claims? Intra-paper verification for peer review arXiv cs.CL source ↗

Jul 30 Corpus of transcribed English-language religious radio broadcasts from 785 webstreams captured over July 2025 in the United States arXiv cs.CL source ↗

Jul 30 High-Order Markov Blanket Discovery via a k-Order Relaxation of the Faithfulness Assumption arXiv cs.LG source ↗

Jul 30 Emergent sparsity in frozen random CNN feature extractors for deep reinforcement learning. arXiv cs.LG source ↗

Jul 30 Mergeable model-side aggregation states for long-context language models arXiv cs.CL source ↗

Jul 30 Shared SFT lessons across alignment, model organisms, and toy models show transferable findings through cross-domain tests arXiv cs.LG source ↗

Jul 30 Between gradient and natural gradient: a continuum of LoRA initializations arXiv cs.LG source ↗

Jul 30 RAGuard introduces a layered defense against factual corpus-poisoning attacks in retrieval-augmented generation systems. arXiv cs.LG source ↗

Jul 30 Diagnosing fine-grained inconsistency classification in financial disclosure text arXiv cs.CL source ↗

Jul 30 DuplexGen enables adaptive synthesis of human-AI turn-taking dialogues. arXiv cs.CL source ↗

Jul 30 Voice Memory enables agentic speech recognition with a frozen corrector deciding per utterance to act on the hypothesis or abstain, and an asynchronous optimizer revising a per-domain memory.md through bounded edits arXiv cs.CL source ↗

Jul 30 Self-serve entity resolution pipeline shows no single winner across datasets; recommend multiple algorithm families and automatic bake-off selection arXiv cs.LG source ↗

Jul 30 ClockRoPE: Random Fourier Rotations for Temporal Routine Modeling arXiv cs.LG source ↗

Jul 30 DHRCL: training code LLMs with dense hierarchical rewards and curriculum learning arXiv cs.LG source ↗

Jul 30 Learning implicit causal world models from multi-agent demonstrations arXiv cs.LG source ↗

Jul 30 Data fusion and contrastive alignment for unconstrained IR molecular structure elucidation arXiv cs.LG source ↗

Jul 30 Knowledge before reasoning: EC-Reason-Bench, a training-free diagnostic benchmark for LLM enzyme classification arXiv cs.CL source ↗

Jul 30 MetaKoopman: Bayesian meta-learning of Koopman operators for modeling nonlinear dynamics under distribution shifts arXiv cs.LG source ↗

Jul 30 Top-k Pareto bandits: hypervolume regret for multi-objective slate selection arXiv cs.LG source ↗

Jul 30 Large-scale chatbot validation using synthetic customer agents as digital twins generated from real transactional and conversational data arXiv cs.CL source ↗

Jul 30 SCOUT: Per-context reset curricula for sparse-reward reinforcement learning arXiv cs.LG source ↗

Jul 30 AgentGUI: an interface for observing and steering long-running AI agents arXiv cs.CL source ↗

Jul 30 Dynamic parameterization is not dynamic inference arXiv cs.LG source ↗

Jul 30 Choosing where and how to moderate: end-to-end trade-offs in filter placement and response rewriting arXiv cs.CL source ↗

Jul 30 Existence-field diffusion model for spatial point processes with variable cardinality arXiv cs.LG source ↗

Jul 30 Where detectors fail: expert-guided mutual distillation improves robustness to tail-domain gaps arXiv cs.CL source ↗

Jul 30 Sim2Win: a team-agnostic, event-based pre-match outcome prediction and tactical profiling system for football arXiv cs.LG source ↗

Jul 30 Early verdicts, better budgets: sequential adaptive rollout allocation for compute-efficient RLVR arXiv cs.LG source ↗

Jul 30 Post-training at the edge of detectability: a game-theoretic approach to fine-tuning arXiv cs.LG source ↗

Jul 30 Symphony-Bias investigates gender associations with musical instruments in multimodal LLMs arXiv cs.CL source ↗

Jul 30 FloDR: An invertible dimensionality reduction method based on a normalising flow arXiv cs.LG source ↗

Jul 30 When synthetic users fail: a cross-domain benchmark of LLM-simulated human survey responses arXiv cs.CL source ↗

Jul 30 Lossy verification in speculative decoding can rewrite the decoding distribution and affect stability, study finds arXiv cs.CL source ↗

Jul 30 Q-Steer uses rollout-time action-value steering to improve molecular policy optimization with oracle-limited feedback arXiv cs.LG source ↗

Jul 30 Steering instruction hierarchies at inference time arXiv cs.CL source ↗

Jul 30 Flow map learning via nongradient vector flow arXiv cs.LG source ↗

Jul 30 Weak-to-Strong on-policy distillation improves alignment when no larger teacher exists arXiv cs.LG source ↗

Jul 30 Evaluating prompt scope and demonstration similarity in local LLM machine translation arXiv cs.CL source ↗

Jul 30 (Im)Paired programming: coding agents improve productivity but harm understanding arXiv cs.CL source ↗

Jul 30 Robostreet Flow: a battery-electric 6x4 tractor with a four-truck convoy architecture to minimize cost per ton-mile on high-volume point-to-point freight arXiv cs.CL source ↗

Jul 29 MIT-developed PhysioNet evolves into a global data-sharing standard for medical research. MIT News (AI) source ↗

Jul 29 K-Search translates CUDA optimization knowledge into architecture-native MLX strategies for Apple Silicon BAIR (Berkeley) source ↗

Jul 29 SpecPrefetch: Parameter-efficient expert prefetching for sparse MoE foundation models arXiv cs.AI source ↗

Jul 29 VisualPatchWorld: Code world models as latent structured representations for planning arXiv cs.CL source ↗

Jul 29 Human preference aligned tabular similarity for similarity search in business systems; arXiv:2607.24880 highlights limitations of standard metrics and argues for embedding trustworthiness in human-aligned rankings arXiv cs.LG source ↗

Jul 29 Models fake alignment without clear consequences, study suggests arXiv cs.AI source ↗

Jul 29 TimeCapsule: a 1.2B LLaMA-style model trained only on Victorian texts (1800–1875) as an epistemologically isolated generative archive. arXiv cs.CL source ↗

Jul 29 Evaluation of forced alignment of code-mixed speech: the case of Hindi-English arXiv cs.CL source ↗

Jul 29 Unified algorithmic framework for hybrid reinforcement learning in tabular MDPs with shifted transition dynamics arXiv cs.LG source ↗

Jul 29 Interpretable column annotation with LLM-symbolized decision process materialization arXiv cs.CL source ↗

Jul 29 When shortest isn’t safest: a design science approach to senior-friendly pedestrian routing arXiv cs.AI source ↗

Jul 29 Neurai-VN benchmark: standardized machine learning models for multimodal digital phenotyping in mental health classification arXiv cs.LG source ↗

Jul 29 LinkedIn presents a unified semantic modeling framework for large-scale job understanding powered by a small language model arXiv cs.AI source ↗

Jul 29 GROCLM: a fine-tuned language model for grocery category recommendation in e-commerce arXiv cs.AI source ↗

Jul 29 Stable FP4 training via transposition-invariant block quantization arXiv cs.LG source ↗

Jul 29 Cross-lingual analysis of entrainment in code-switched speech shows lexical entrainment generalizes across Mandarin-English, Hindi-English, and Spanish-English pairs, while acoustic-prosodic and CSW-style entrainment varies by context. arXiv cs.CL source ↗

Jul 29 CAST: Game solvers as turn-level teachers for LLM agents arXiv cs.CL source ↗

Jul 29 GAUGE scores financial models without a golden answer using independently built analyst models across 108 directed pairs. arXiv cs.LG source ↗

Jul 29 Kernel Forge: an agent harness for LLM-based generation and optimization of CUDA kernels arXiv cs.AI source ↗

Jul 29 Score-based stabilization for time-dependent problems in numerical PDE simulation arXiv cs.LG source ↗

Jul 29 Phase structure in rotary attention revealed through a bounded spectral framework for semantic continuity and execution-boundary governance arXiv cs.CL source ↗

Jul 29 RSMeM: Knowledge-enhanced memory evolution for remote sensing agents with systematic evaluation arXiv cs.AI source ↗

Jul 29 HOBA: Hierarchical On-policy Bidding Agents for adaptive online advertising arXiv cs.AI source ↗

Jul 29 Right-sizing recommendations for cloud workloads using conformal prediction to select VM sizes in data center operations arXiv cs.AI source ↗

Jul 29 Deep Label-Wise Attentive Temporal Convolutional Networks improve medical coding arXiv cs.CL source ↗

Jul 29 A scaling law of contextual persistence in human language arXiv cs.CL source ↗

Jul 29 Crystalis: Progressive nucleation and semantic annealing for coordinated multi-view visualization generation arXiv cs.AI source ↗

Jul 29 Generative Distributionally Robust Optimization (GDRO) arXiv cs.LG source ↗

Jul 29 RRS-10K: a benchmark for rare remote sensing image interpretation with 10,738 military-related images arXiv cs.AI source ↗

Jul 29 PATHFinder Agent for tailored prenatal care arXiv cs.AI source ↗

Jul 29 Mechanisms of width scaling in normalized residual networks: the effective alignment dimension arXiv cs.LG source ↗

Jul 29 Multilingual safety gaps in frontier language models: in-context scheming scales with pretraining language coverage, according to Petri auditing findings arXiv cs.AI source ↗

Jul 29 LLMs as an alternative to corpora for specialized terminology in English–French translation arXiv cs.AI source ↗

Jul 29 IRIS: reusable identity representations from frozen LLMs for entity alignment arXiv cs.CL source ↗

Jul 29 Conformal Cascade: distribution-free accuracy guarantees for multi-tier LLM inference arXiv cs.LG source ↗

Jul 29 ProcAgent: an agentic framework for procedural task guidance on edge with human-in-the-loop arXiv cs.AI source ↗

Jul 29 Physics-Informed CNN-LSTM for street-scale urban flood prediction: reconciling aggregate accuracy and street-level plausibility arXiv cs.LG source ↗

Jul 29 Toward a systematic method for identifying language areas arXiv cs.CL source ↗

Jul 29 Calibrated partial resets prevent policy collapse in continual reinforcement learning arXiv cs.LG source ↗

Jul 29 Interpretable GOHR agents via sparse autoencoders arXiv cs.LG source ↗

Jul 29 Semantic space search trajectory networks: constructing STNs in semantic spaces to visualize optimization behavior arXiv cs.LG source ↗

Jul 29 Personalization, personas, and forecasting in value alignment arXiv cs.AI source ↗

Jul 29 CondPSE: a polynomial-filtered structural encoder with conditional modulation for graphs arXiv cs.LG source ↗

Jul 29 LLM-based trace ranking and grouped reward modeling for multilingual numerical claim verification at CheckThat! 2026 arXiv cs.CL source ↗

Jul 29 Beyond memory: a templated substrate for heterogeneous collaborative knowledge work with LLM agents arXiv cs.AI source ↗

Jul 29 Inverse RL helps align AI by imitating humans arXiv cs.LG source ↗

Jul 29 TabRank: chain-of-thought distillation for table re-rankers arXiv cs.CL source ↗

Jul 29 Reasoning with memory: a temporal granularity-adaptive framework for training-free long video understanding arXiv cs.AI source ↗

Jul 29 CaRE: Compute-aware Remasking Evaluation Protocol for Masked Diffusion Language Models arXiv cs.AI source ↗

Jul 29 Rethinking CD: a reproducibility study and extension on the ineffectiveness of contrastive decoding at mitigating object hallucinations in MLLMs arXiv cs.LG source ↗

Jul 29 Lantern: Conflict-aware gradient blending for physics-guided diffusion models in calorimeter simulation arXiv cs.LG source ↗

Jul 29 RoCo-ACE: Rollout-conditioned online distillation for retention-aware knowledge injection arXiv cs.AI source ↗

Jul 29 LLM as forecasting planner: training-free text conditioning for time-series foundation models arXiv cs.LG source ↗

Jul 29 Multiclass classification without labels via posterior simplex geometry arXiv cs.LG source ↗

Jul 29 Activation source selection in activation steering: where steering signals come from arXiv cs.CL source ↗

Jul 29 Steering topology distributions for unified generative design of architected metamaterials arXiv cs.AI source ↗

Jul 29 Accurate structural modeling of chemically diverse molecular interfaces with Vilya-2 arXiv cs.LG source ↗

Jul 29 Linguistic rules alone can serve as effective prompt compressors to reduce inference costs, study finds arXiv cs.CL source ↗

Jul 29 Behavior-Driven Explainability arXiv cs.LG source ↗

Jul 29 Research report analyzes noise-shaped one-bit coefficients in discrete polynomial Fourier extension; achieves O(N^{-1}) approximation rate for first-order Sigma-Delta quantization arXiv cs.CL source ↗

Jul 29 FinAbstain: uncertainty-calibrated multimodal RAG for selective financial forecasting arXiv cs.LG source ↗

Jul 29 Endpoint Replay: Compressing the recency buffer in deep reinforcement learning arXiv cs.LG source ↗

Jul 29 Atmospheric diffusion-guided spatio-temporal transformer for nuclear radiation forecasting arXiv cs.AI source ↗

Jul 28 LeafData: an agentic system that converts user intent into validated JSON configurations for data migration arXiv cs.AI source ↗

Jul 28 Transferable latency prediction for fast LLM screening on heterogeneous edge devices arXiv cs.AI source ↗

Jul 28 Discrete action space as a prerequisite for GRPO convergence in small-model continuous control arXiv cs.AI source ↗

Jul 28 Public-space gesture interaction evaluated over interaction-level success rather than frame-level accuracy; paper analyzes repair records and runtime failure analysis in scenic kiosks and service terminals. arXiv cs.AI source ↗

Jul 28 Persistent computational state: a session-centric runtime for generative world models arXiv cs.AI source ↗

Jul 28 FrED: External data influence estimation via domain knowledge graph grounding arXiv cs.AI source ↗

Jul 28 FlowEvo: self-evolving agents through the co-evolution of workflows and executable skills arXiv cs.AI source ↗

Jul 28 Context anxiety causes frontier reasoning models to doubt their solutions despite capability; a study analyzes how misestimation of token difficulty affects performance. arXiv cs.AI source ↗

Jul 28 SCOPE and SCION: a benchmark and an auditable reference pipeline for schema induction and fusion from text arXiv cs.AI source ↗

Jul 28 FBLayout: Optimizing memory layout for efficient LLM finetuning on mobile GPUs arXiv cs.AI source ↗

Jul 28 Spectral Flow Certificates enable depth-aware long-range propagation in graph neural networks arXiv cs.AI source ↗

Jul 28 Wavelet phase diffusion for structurally and semantically consistent sim-to-real translation arXiv cs.AI source ↗

Jul 28 Do VLMs read or rewrite? Evidence of transcription rewriting in vision-language models using FaithC4 benchmark arXiv cs.AI source ↗

Jul 28 HDL emerges as a hard decision layer in transformers, causing abrupt stabilization of answer option rankings during inference arXiv cs.AI source ↗

Jul 28 Procedural knowledge is not low-rank: LoRA fails to internalize multi-step procedures arXiv cs.AI source ↗

Jul 28 From profiles to steering vectors: Global Sparse Priors and Local Semantic Calibration for personalized text generation arXiv cs.AI source ↗

Jul 28 Coupled hierarchical search over topology and execution for agentic workflow synthesis arXiv cs.AI source ↗

Jul 28 TILT: Improving compositional generation in diffusion models with a model-intrinsic reward arXiv cs.AI source ↗

Jul 28 Trajectory-aware retrieval agents for temporal decision making arXiv cs.AI source ↗

Jul 28 Defining AI-native systems: autonomy as revision authority arXiv cs.AI source ↗

Jul 28 Do modules stay in their lane? Role drift in compound LLM systems arXiv cs.AI source ↗

Jul 28 Securing multimodal AI through internal information decomposition arXiv cs.AI source ↗

Jul 28 Monotonic evaluation framework measures whether increases in wildfire risk scores correspond to higher observed operational load arXiv cs.AI source ↗

Jul 28 AgentKVShift: Efficient KV cache reuse for agentic memory systems arXiv cs.AI source ↗

Jul 28 Household movement detection in mixed-format occupancy data using LLM-based entity resolution arXiv cs.AI source ↗

Jul 28 FMOPF: Latent Flow Matching with Constraint-Aware Interaction Priors for AC Optimal Power Flow arXiv cs.LG source ↗

Jul 28 OrchNAS: orchestrated neural architecture search service for personalised federated edge intelligence arXiv cs.LG source ↗

Jul 28 LA-RL: Label-aware self-reflection for reinforcement learning in information extraction arXiv cs.CL source ↗

Jul 28 Joint optimization for greedy longest-match tokenization (JOLT) develops a subword vocabulary learned to optimize greedy left-to-right longest-match decoding for WordPiece arXiv cs.CL source ↗

Jul 28 CC-AOS: Cost- and horizon-conditioned amortized backward induction for finite-horizon optimal stopping arXiv cs.LG source ↗

Jul 28 Integrated deep learning and statistical framework for whole-network gene–environment association with leaf vascular architecture arXiv cs.LG source ↗

Jul 28 Memento: memory-guided memetic code-as-policy evolution arXiv cs.LG source ↗

Jul 28 QFedPolyp: a communication- and inference-efficient federated learning framework for polyp segmentation arXiv cs.LG source ↗

Jul 28 Bharati: morphology-aware tokenizers for classical Indian languages with subword fertility analysis arXiv cs.CL source ↗

Jul 28 Activation Oracles learn not to read: concept-specific blind spots in fine-tuned oracles arXiv cs.CL source ↗

Jul 28 Beyond Shapley: an influence-based data auditing pipeline for LLM alignment and evaluation arXiv cs.LG source ↗

Jul 28 Automated detection of documentation inconsistencies in electronic health records using a two-stage LLM pipeline arXiv cs.CL source ↗

Jul 28 Physically verifiable evidence and LLM-based reporting for bearing fault diagnosis arXiv cs.LG source ↗

Jul 28 Multimodal domain generalization for depression detection via an attention-based BiLSTM network with domain-adversarial training arXiv cs.LG source ↗

Jul 28 Optimizing Transformer neural network for real-time outlier detection on FPGAs arXiv cs.LG source ↗

Jul 28 Hierarchical grading in large language models arXiv cs.LG source ↗

Jul 28 ADAGE: a language-agnostic pipeline for analogical reasoning evaluation arXiv cs.CL source ↗

Jul 28 From hybrid mechanistic–data-driven modeling toward neuro-symbolic AI: what, why, and how arXiv cs.LG source ↗

Jul 28 LithoFormer: a robust framework for stratigraphic inference via transformers arXiv cs.LG source ↗

Jul 28 IKS-Instruct: a 24,795-pair multilingual dataset for teaching language models Indian Knowledge Systems arXiv cs.CL source ↗

Jul 28 CausalGate: causal importance distillation for transformer module pruning arXiv cs.LG source ↗

Jul 28 PatiGonit22K: a comprehensive dataset for solving complex Bengali MWPs arXiv cs.CL source ↗

Jul 28 DomainPilot: Domain-Level loss-guided two-stage data mixture optimization for efficient language model fine-tuning arXiv cs.LG source ↗

Jul 28 Evaluating narrative unlearning with LENS: a level-based evaluation protocol for suppressing disinformation-aligned narrative reproduction in LLMs arXiv cs.CL source ↗

Jul 28 IndicTalk: a large-scale multilingual code-mixed conversational corpus for Indic languages arXiv cs.CL source ↗

Jul 28 Aligning educational LLMs as Socratic guides via heuristic reinforcement learning; study presents HeuristicEdu for Qwen2.5-7B using SocraticEdu data and GRPO training arXiv cs.CL source ↗

Jul 28 AutoThinkSQL enables selective reasoning for text-to-SQL by combining SFT and DPO guidance arXiv cs.CL source ↗

Jul 28 MioFFAn: an open-source framework for annotating mathematical expressions for formula formalization with LLM automation capabilities arXiv cs.CL source ↗

Jul 28 Multimodal surface sEMG hand gesture recognition using query-based transformers for prosthetic control arXiv cs.LG source ↗

Jul 28 LoRA for gender-inclusive rewriting and activation steering for counter-narrative generation achieves 80.00% official score on gender-inclusive rewriting (IHLC system) for LT-EDI 2026 Shared Task arXiv cs.CL source ↗

Jul 28 Attention-guided layer selection for contrastive decoding in large language models arXiv cs.CL source ↗

Jul 28 Co-evolving graph and text memory for training-free multi-hop question answering arXiv cs.CL source ↗

Jul 28 Softmax throws away: mass-aware attention for evidence accumulation arXiv cs.LG source ↗

Jul 28 CHiPS: character histograms and positional signals for lightweight authorship attribution in Romanian texts arXiv cs.CL source ↗

Jul 28 Kalle Lyytinen discusses the linguistic core of information systems and the developments since his 1980 MIS Quarterly article; interview highlights. arXiv cs.CL source ↗

Jul 28 Speech signals complement LLMs for predicting interpersonal attraction in speed dating arXiv cs.CL source ↗

Jul 28 Simple language normalization wins: cross-lingual speaker verification for the tidyVoice 2026 challenge arXiv cs.CL source ↗

Jul 28 GAND: a resource on gender-ambiguous natural data and contrastive attribution for evaluating gender bias in machine translation arXiv cs.CL source ↗

Jul 28 SLA-constrained carbon-aware routing in geo-distributed serverless clouds arXiv cs.LG source ↗

Jul 28 Dementia etiology diagnosis via collaborative meta knowledge enhancement arXiv cs.LG source ↗

Jul 28 Semalith v1.4, a 184M DeBERTa-v3-base classifier, achieves state-of-the-art prompt-injection detection with 44x fewer parameters than Llama-Guard-3-8B arXiv cs.LG source ↗

Jul 28 Predicting rTMS depression therapy outcomes from EEG signals using CNNs arXiv cs.LG source ↗

Jul 28 Official conference guidelines yield more consistent results in automated peer review by LLMs. arXiv cs.CL source ↗

Jul 28 Progress-conditioned group policy optimization for long-horizon agentic tasks arXiv cs.LG source ↗

Jul 28 BERT-based models vs. large language models for low-resource named entity recognition: a comparative study on Marathi arXiv cs.CL source ↗

Jul 28 Accessibility plasticity as a principle of adaptive computation for access to computation arXiv cs.LG source ↗

Jul 28 Beyond a global norm: personalizing toxicity sensitivity in language models without retraining arXiv cs.CL source ↗

Jul 28 Corvus: context optimization and reduction via underlying synchronization for LLM coding agents arXiv cs.LG source ↗

Jul 28 LC-SEPLM: long-range contact-supervised adaptation for sequence-only protein representation learning arXiv cs.LG source ↗

Jul 28 Loss-aware feature-map pruning in convolutional neural networks using multi-armed bandits arXiv cs.AI source ↗

Jul 28 SeT-Diff: Towards semantic foundation models for HPC telemetry and time-series arXiv cs.AI source ↗

Jul 28 cMoLLM at scale: horizontal scaling laws for mixture-of-LLMs arXiv cs.AI source ↗

Jul 28 QFoldAgent: a closed-loop multi-agent framework for 5-residue tetrahedral-lattice protein folding with a design agent proposing sequence-conditioned penalties arXiv cs.AI source ↗

Jul 28 PhononBench-MP40 provides a spectrum-resolved benchmark dataset of Materials Project-derived crystals for workflow-defined phonon stability. arXiv cs.AI source ↗

Jul 28 Scaffold effect in coding agents: evaluation varies with harness choice across Qwen 3.6 Plus and MiniMax M2.5 arXiv cs.AI source ↗

Jul 28 Prompt design significantly impacts energy use in on-device LLM inference arXiv cs.AI source ↗

Jul 28 HeraSys enables collaborative serving of multiple LLM workflows through fine-grained end-to-end optimization arXiv cs.AI source ↗

Jul 28 Evaluating LLM reliability beyond accuracy: how model answers vary with meaning-preserving paraphrases across tasks arXiv cs.AI source ↗

Jul 28 Synthetic scenario generation for evaluation of Industry 4.0 agents; extends AssetOpsBench with a Smart Grid Transformer asset class and four IEC-grounded diagnostic tools for health-index prediction, dissolved-gas analysis, and winding-temperature detection. arXiv cs.AI source ↗

Jul 28 Execution-Grounded security testing for coding agents in software engineering pipelines arXiv cs.AI source ↗

Jul 28 SCEPTER converts clinical case descriptions into evidence-based recommendations using multi-objective evidence reasoning arXiv cs.AI source ↗

Jul 28 Schema-Aware Localisation (SAL): live schema grounding and hallucination validation for Oracle NL2SQL arXiv cs.AI source ↗

Jul 28 Codifying the Judge: scalable evaluation via program distillation arXiv cs.AI source ↗

Jul 28 Memory-Induced Inference-Time Adaptation for continual learning with small language models arXiv cs.AI source ↗

Jul 28 Source-aware reranking for retrieval-augmented generation using a reliability prior approach arXiv cs.AI source ↗

Jul 28 Multi-Objective structured pruning of LLMs for latency and model size optimization arXiv cs.AI source ↗

Jul 28 DeepLens Diagnosis Agent uses a five-stage reasoning pipeline to enable a small medical reasoning model to compete with frontier LLMs. arXiv cs.AI source ↗

Jul 28 Concept-based visual counterfactual explanations with diffusion models arXiv cs.AI source ↗

Jul 28 Reference feature atlases for mechanistic auditing of language models arXiv cs.AI source ↗

Jul 28 SCAIR: Schema-conditioned agentic iterative reasoning for enterprise knowledge graphs arXiv cs.AI source ↗

Jul 28 DSTFView: multi-view workload forecasting for cloud-edge platforms using dual-input spatio-temporal-frequency modeling arXiv cs.AI source ↗

Jul 28 MedLoCoMo: a long-context, multi-session medical dialogue benchmark for large language models arXiv cs.AI source ↗

Jul 27 Self-poisoning in adaptive out-of-distribution detection: a sharp-threshold theory and certified label-free calibration arXiv cs.LG source ↗

Jul 27 Benchmarking fine-tuning and retrieval strategies for a multimodal language model on the NRC Reactor Operator licensing examination arXiv cs.CL source ↗

Jul 27 Molt: a PyTorch-native training framework for agentic reinforcement learning reduces researcher cost per iteration arXiv cs.LG source ↗

Jul 27 Quasi-Monte Carlo initialization improves meta-reinforcement learning training convergence in benchmark environments; QMC meta-priors outperform SB3 defaults on unseen continuous control tasks. arXiv cs.LG source ↗

Jul 27 Shallower ReLU network representations via exact linear algebra arXiv cs.LG source ↗

Jul 27 Probing latent Colombian identity inferences in Qwen2.5-7B with natural language autoencoders arXiv cs.CL source ↗

Jul 27 Multi-horizon latent consistency as geometry: increasing lambda affects empirical expansion proxy L20 and horizon-20 prediction error on Moving-MNIST arXiv cs.LG source ↗

Jul 27 Quadratic model can be surprisingly predictive for optimization in large-language models with 150M parameters arXiv cs.LG source ↗

Jul 27 A consensus-based framework for relative preference evaluation of large language models arXiv cs.CL source ↗

Jul 27 Learning what matters: supervising sparse attention routing with causal evidence sets arXiv cs.LG source ↗

Jul 27 Reinforcement learning mitigates task conflicts in LLM model merging; analysis examines training paradigms' impact arXiv cs.CL source ↗

Jul 27 Encoding invisible causation for bridge diagnostic agents; triple-guided retrieval-augmented fine-tuning with QLoRA arXiv cs.LG source ↗

Jul 27 Khondo: a multimodal benchmark for document packet splitting of Bangla forms arXiv cs.CL source ↗

Jul 27 Bounding the causal impact of ML-assisted decision-making via counterfactual correctness arXiv cs.LG source ↗

Jul 27 A drift-stable quantum federated learning for intelligent services arXiv cs.LG source ↗

Jul 27 Adjustment speed as a safety constraint for nonstationary reinforcement learning arXiv cs.LG source ↗

Jul 27 FSE: Continual learning for named entity recognition by fast-slow experts arXiv cs.CL source ↗

Jul 27 Neural Atom Prevalence extends atom prevalence to neural networks for structured node-level model selection in feedforward architectures. arXiv cs.LG source ↗

Jul 27 Analyzing self-harm representations in language models: a cross-architecture study arXiv cs.CL source ↗

Jul 27 From isolated tasks to structured capabilities: a multilayer taxonomy for large language models arXiv cs.CL source ↗

Jul 27 Physiological signals as a forensic modality for talking-face deepfake detection arXiv cs.LG source ↗

Jul 27 Spanish version of the Large Language Models Dependency Scale (LLM-D12-SP) developed and validated arXiv cs.CL source ↗

Jul 27 Measuring the dependency gap: diagnosing inter-column fidelity in tabular generative models arXiv cs.LG source ↗

Jul 27 Cloud-Native Evaluation-as-a-Service: a six-microservice architecture for scalable AI monitoring with conformal guarantees arXiv cs.LG source ↗

Jul 27 Dynamic Commonsense Coordination for empathetic response generation arXiv cs.CL source ↗

Jul 27 Agentic evaluation of copyright law compliance with Copyright-Bench assessing LLM agents’ compliance with copyright law arXiv cs.CL source ↗

Jul 27 Data quality over capacity: baking documents into LoRA adapters for closed-book QA arXiv cs.CL source ↗

Jul 27 Smart predict-then-robustly-optimize: account for prediction shifts due to disturbance in the covariate feature space arXiv cs.LG source ↗

Jul 27 RED-PIM: Reducing data movement for transformers using processing-in-memory arXiv cs.LG source ↗

Jul 27 Ground Truth First: a longitudinal evaluation instrument for agent memory and the tenure crossover in memory-architecture rankings arXiv cs.CL source ↗

Jul 27 Humanly: A configurable and traceable environment for human-AI collaborative writing arXiv cs.CL source ↗

Jul 27 MEUSLI: a multilingual projector linking Whisper encoder with open-source multilingual ASR and beyond arXiv cs.CL source ↗

Jul 27 Scaling native multimodal pre-training from scratch arXiv cs.CL source ↗

Jul 27 Introduction to Bayesian and frequentist simulation-based inference with machine learning arXiv cs.LG source ↗

Jul 27 Improving faithfulness of podcasts from documents arXiv cs.CL source ↗

Jul 27 Restoration of historical documents via retrieval-augmented large language models leverages external knowledge to recover illegible entities arXiv cs.CL source ↗

Jul 27 J-CoT: Chain-of-Thought in J-Space; explores latent-reasoning through recurrent propagation of continuous hidden states arXiv cs.CL source ↗

Jul 27 Depth scalability limitations in logic gate networks; optimization collapse and topology-induced constraints persist beyond training improvements arXiv cs.LG source ↗

Jul 27 Toward user-conditioned evaluation of personal LLM agents under temporal interventions arXiv cs.LG source ↗

Jul 27 Analyzing toxic behavior and its impact on the Mastodon community arXiv cs.CL source ↗

Jul 27 Parameter-free adaptive sparse attention via compression-based content selection arXiv cs.LG source ↗

Jul 27 Goal-agnostic joint-embedding predictive control framework for partial differential equations using offline-trained ViT encoder and action-conditioned latent dynamics with an MPPI controller arXiv cs.LG source ↗

Jul 27 DWT-Fusion: a signal-based framework for training-free LLM-generated text detection arXiv cs.CL source ↗

Jul 27 MoE$^2$-LoRA: when MoE models meet MoE-style low-rank adaptation arXiv cs.CL source ↗

Jul 27 Evaluation design conditions the expert-vs-auto MeSH gap; a controlled comparison of bag-of-words and BiomedBERT on the Cohen benchmark arXiv cs.CL source ↗

Jul 27 Adversarial style optimization: enhancing VLM jailbreaks by GRPO-based stylistic triggers optimization arXiv cs.CL source ↗

Jul 27 Physically constrained federated additive models for O-RAN SLA-risk prediction arXiv cs.LG source ↗

Jul 27 MotifRole-Diff: Risk-optimal role-aware corruption for masked molecular graph diffusion arXiv cs.LG source ↗

Jul 27 CARNet: cycle-conditioned core aggregation and redistribution for multivariate time series forecasting arXiv cs.LG source ↗

Jul 26 Teaching LLMs to update beliefs for efficient long-horizon interaction BAIR (Berkeley) source ↗

Jul 25 What is good? Extracting and testing implicit theories of literary quality from LLM reasoning traces arXiv cs.CL source ↗

Jul 25 Naver-News-KO: a Korean news summarization dataset for open-source fine-tuning of summarization models arXiv cs.CL source ↗

Jul 25 More Is Not More: what matters for diversity in LLM opinions? arXiv cs.CL source ↗

Jul 25 Answer-then-Edit: Reasoning skeleton editing for anti-distillation with preserved utility arXiv cs.CL source ↗

Jul 25 AsymVerify achieves 2nd place on SemEval-2026 Task 6 with a confidence-gated verification approach for political evasion detection arXiv cs.CL source ↗

Jul 25 LLM-INSTRUCT wins UZH Shared Task 2026 on paragraph-level argument mining with constraint-aware retrieval and selective debate arXiv cs.CL source ↗

Jul 25 Natural language should not fully replace formal languages, argues a position paper highlighting underspecification in open-ended contexts arXiv cs.CL source ↗

Jul 25 A three-stage neural-symbolic pipeline using an ensemble of DeBERTa-v3-base and XLM-RoBERTa-base with a Linguistically-Informed Mediator for gaming toxicity detection. arXiv cs.CL source ↗

Jul 25 TopoGuard: graph theory based defenses against split-knowledge attacks on RAG arXiv cs.CL source ↗

Jul 25 Routing Subspaces: Auditing evaluation-to-deployment mismatch in fine-tuned language models arXiv cs.CL source ↗

Jul 25 Open-source text LLM watermarks fail to withstand model merging, study finds arXiv cs.CL source ↗

Jul 25 Human-in-the-loop LLM framework improves detection of cutaneous immune-related adverse events in clinical notes; study reports higher accuracy and agreement vs manual review arXiv cs.CL source ↗

Jul 25 MoE routing follows a Huffman code pattern, as the Frequency-Diversity Law shows that state-of-the-art models act as information-theoretic engines. arXiv cs.CL source ↗

Jul 25 Skill-contracted agents for evidence-aware materials literature analysis arXiv cs.CL source ↗

Jul 25 SCoPE: Shift-aware speaker-conditioned priors for emotion recognition in conversations arXiv cs.CL source ↗

Jul 25 Knowledge injection exists in MoE; exploring expert-aware contrast decoding in MoE for mitigating LLMs' hallucinations arXiv cs.CL source ↗

Jul 25 Moir: Let the model direct its own story for robust cross-domain knowledge editing arXiv cs.CL source ↗

Jul 25 Distinguishing artificial from authentic: evaluating LLMs for detecting LLM-generated content arXiv cs.CL source ↗

Jul 25 GLAN-QnA-KR: a 303,581-row Korean instruction-QA corpus produced by seedless taxonomy-driven GLAN synthesis using Microsoft Phi-3.5-MoE-instruct; open and redistributable under OpenRAIL arXiv cs.CL source ↗

Jul 25 Preference tuning as spectral update reorganization arXiv cs.CL source ↗

Jul 25 Confidently deceptive: how confidence amplifies the risk of LLM deception arXiv cs.CL source ↗

Jul 25 Break Through the Compression Bottleneck: From Theory to Practice arXiv cs.CL source ↗

Jul 25 The Storyteller in the Model: Narrative pattern inheritance, escalation dynamics, and alignment governance in LLMs arXiv cs.CL source ↗

Jul 25 Belief propagation in LLM world models: measuring strategic information bias with prediction markets arXiv cs.CL source ↗

Jul 24 HypNO: a graph-based neural operator for scalar hyperbolic conservation laws uses physics-informed message passing on a space-time graph of finite-volume cells arXiv cs.LG source ↗

Jul 24 Conflict resolution under degraded surveillance in air corridors using multi-agent reinforcement learning arXiv cs.LG source ↗

Jul 24 AINTMA: Agentic AI Architecture for Autonomous Test Management with Generative Intelligence, Secure Cloud Communication and Adaptive Quality Analytics arXiv cs.AI source ↗

Jul 24 Adaptive depth in looped transformers: diagnosing learned halting gates and trajectory readouts arXiv cs.LG source ↗

Jul 24 Uncertainty-aware trust estimation for multi-LLM systems via structured expert judgement arXiv cs.LG source ↗

Jul 24 ClickGuard: Detecting and spoiling clickbait news with informativeness measures and large language models arXiv cs.AI source ↗

Jul 24 DC-Leap: training-free acceleration of dLLMs via draft-guided contiguous leaping decoding arXiv cs.AI source ↗

Jul 24 The Devil is in the spectrum: mitigating representation collapse in LLMs via topologically regularized side-path arXiv cs.AI source ↗

Jul 24 Incomplete prompt jailbreaks in large language models reveal vulnerabilities to harmful continuations in open-weight models arXiv cs.AI source ↗

Jul 24 Graph neural network approach to zero-shot digital twins arXiv cs.LG source ↗

Jul 24 Codec-Gauge learns equalization transforms for transformer KV-cache compression to improve fidelity arXiv cs.LG source ↗

Jul 24 PersonaTrail benchmarks personalized web agents through browsing histories to evaluate how agents infer context from user browsing data arXiv cs.AI source ↗

Jul 24 OPTScientist: multi-agent discovery of typed optimizer programs for transformer pretraining arXiv cs.AI source ↗

Jul 24 DataPrep-Bench benchmarks LLMs as training data preparators; evaluates data construction and data quality evaluation end to end arXiv cs.LG source ↗

Jul 24 Generative Bayesian filtering for state estimation arXiv cs.LG source ↗

Jul 24 Leveraging biokinetic knowledge priors for data-scarce bioprocess modeling arXiv cs.LG source ↗

Jul 24 When RLVR shrinks the reasoning boundary: diagnosing pass@k inversion arXiv cs.LG source ↗

Jul 24 Lie typology, depth, and sparsity affect deception detection in LLM outputs arXiv cs.AI source ↗

Jul 24 SOAP, Muon, and beyond: pushing LLM pretraining scales arXiv cs.LG source ↗

Jul 24 PhantomFill: when the form demands an answer, language models invent one arXiv cs.LG source ↗

Jul 24 Routing Without Training: Controllable-Ratio LLM Offloading via Reliability Gating arXiv cs.AI source ↗

Jul 24 SevDiff: Severity-conditioned diffusion for long-tail conflict trajectory generation arXiv cs.LG source ↗

Jul 24 Muon optimizer reaches grokking threshold on modular arithmetic faster than AdamW; ablation shows speedup from Newton-Schulz orthogonalization arXiv cs.LG source ↗

Jul 24 Stochastic sampling is epistemically shallow; the dimensionality gap between temperature variation and model diversity in LLMs arXiv cs.AI source ↗

Jul 24 DecodeShare identifies a shared low-dimensional subspace in decode-time hidden states and tests its causal role by removing it during decoding arXiv cs.AI source ↗

Jul 24 Tractable hierarchical control of autoregressive language models arXiv cs.AI source ↗

Jul 24 Scaling closed-loop feature channel configuration with LLMs arXiv cs.LG source ↗

Jul 24 Benchmarking the personalization capabilities of large language models arXiv cs.AI source ↗

Jul 24 Grounding investor views: neural predicates in the Black-Litterman model arXiv cs.LG source ↗

Jul 24 Semi-supervised text-attributed graph distillation arXiv cs.AI source ↗

Jul 24 Directional hallucinations: ideological drift in news-grounded LLM question answering arXiv cs.AI source ↗

Jul 24 Enabling scalable topology inference in distribution systems via constrained multi-source inference arXiv cs.AI source ↗

Jul 24 SenCos-GEM: SENet-Calibrated and Law-of-Cosines-Constrained geometry-enhanced molecular representation for property prediction arXiv cs.LG source ↗

Jul 24 PlanE: Meta planning of data, tuning, and inference for extractive-based LLMs arXiv cs.AI source ↗

Jul 24 Marking the wrong symptoms: evaluating LLM watermarks in medical texts arXiv cs.AI source ↗

Jul 24 ReliableTableQA evaluates the amount of supervision needed for reliability annotation of tabular QA results arXiv cs.LG source ↗

Jul 24 Expectation alignment of language models for real-world user expectations arXiv cs.AI source ↗

Jul 24 Beyond SBDD: geometric deep learning in polypharmacology and multi-target drug design arXiv cs.LG source ↗

Jul 24 Decision-aware machine learning framework proposed to improve allocation of essential medicines; arXiv paper introduces multi-task approach arXiv cs.LG source ↗

Jul 24 JAXBench: Benchmarking autonomous TPU kernel optimization arXiv cs.AI source ↗

Jul 24 Proactive test-driven AI development is proposed to replace reactive patching of models based on user feedback. arXiv cs.LG source ↗

Jul 24 Optimal noise allocation for diffusion training in the convex regime arXiv cs.LG source ↗

Jul 24 Multimodal CoLRAG-TF uses four-axis fusion—dense text embeddings, BM25, knowledge-graph triple filtering, and image similarity—to enable robust retrieval over complex PDFs. arXiv cs.LG source ↗

Jul 24 SonicSampler: unified tile-aware kernels for LLM sampling and speculative verification arXiv cs.AI source ↗

Jul 24 InferenceBench: a benchmark for open-ended LLM inference optimization by AI agents arXiv cs.AI source ↗

Jul 24 CLOE: Christoffel Loss Autoencoder for anomaly detection arXiv cs.LG source ↗

Jul 24 Active SAE feature planes carry more holonomy in Gemma 2 2B, preregistered reversal findings arXiv cs.LG source ↗

Jul 24 Automating nuclear plant operations through collaboration and stakeholder relationships, says Lauren Fortier MIT News (AI) source ↗

Jul 23 MIT projects selected for funding under U.S. Department of Energy’s Genesis Mission MIT News (AI) source ↗

Jul 23 BatchDAG: LLM-planned execution graphs for scalable ad-hoc analysis over enterprise data arXiv cs.AI source ↗

Jul 23 SysAdmin: measuring instrumental power-seeking in frontier AI arXiv cs.AI source ↗

Jul 23 From Agent Failure Paths to Quantified Residual Risk: A Compositional Framework for Resilient Agentic AI arXiv cs.AI source ↗

Jul 23 Cross-dialect generalization without retraining: benchmarks and evaluation of schema-derived constrained decoding for MLIR arXiv cs.AI source ↗

Jul 23 Semantic cooperative games for contribution attribution in LLM-based multi-agent systems arXiv cs.AI source ↗

Jul 23 Probabilistic concept-aware steering for trustworthy LLM inference arXiv cs.AI source ↗

Jul 23 PEARL: solver-in-the-loop interactive optimization modeling from natural language arXiv cs.AI source ↗

Jul 23 Phionyx introduces a deterministic AI runtime architecture with structured state management and governance-first state evolution; LLM outputs are treated as noisy sensor measurements rather than decisions. arXiv cs.AI source ↗

Jul 23 SAAG: Structured Agent Assessment and Grounding arXiv cs.AI source ↗

Jul 23 FindStatBench: evaluating large language models on combinatorial code synthesis arXiv cs.AI source ↗

Jul 23 MUX: continuous reasoning via multiplexed tokens arXiv cs.AI source ↗

Jul 23 Integro-differential equations in angular stabilization of drone motion by distributed feedback control arXiv cs.AI source ↗

Jul 23 Structured synthetic reasoning data improves arithmetic fine-tuning for small language models, using a 21,250-example GPT-5-mini-generated corpus derived from GSM8K arXiv cs.AI source ↗

Jul 23 Trajectory-aware clinical risk prediction via severity-grounded knowledge graphs and retrieval-augmented generation arXiv cs.AI source ↗

Jul 23 Wisdom of LLM crowds: aggregation and contamination in language model ensembles arXiv cs.AI source ↗

Jul 23 Calibrated selective fact-checking via evidence chain evaluation. arXiv cs.AI source ↗

Jul 23 State compression in two-agent LLM relays preserves constraints in a closed-world travel-planning relay, a study of hand-off compression effects arXiv cs.AI source ↗

Jul 23 MILP-Evo enables closed-loop, fully automatic design of MILP solvers arXiv cs.AI source ↗

Jul 23 ProbSPARQL: querying knowledge graphs with multi-dimensional, uncertain numeric data arXiv cs.AI source ↗

Jul 23 ToolDNS enables semantic tool discovery over DNS to scale autonomous AI agent tool use; arXiv paper proposes retrofitting discovery onto DNS to handle millions of tools arXiv cs.AI source ↗

Jul 23 When JSON is not enough: semantic reliability of schema-constrained LLM ordering agents arXiv cs.AI source ↗

Jul 23 Latency-aware LLM query routing for dynamic workloads improves by considering generation latency alongside accuracy and cost arXiv cs.AI source ↗

Jul 23 AI/ML deepfake research misaligned with AI-generated non-consensual intimate imagery, according to landscape analysis of highly cited works arXiv cs.AI source ↗

Jul 23 Fence: Specialized SLM guardrails for LLM applications arXiv cs.AI source ↗

Jul 23 Bayesian wind tunnels for model selection arXiv cs.LG source ↗

Jul 23 TalentCLEF at CLEF2026: second edition evaluates NLP systems for fair, multilingual talent and job title intelligence in human capital management arXiv cs.CL source ↗

Jul 23 Rubric-oriented document set selection and ranking for retrieval beyond relevance-centric methods arXiv cs.CL source ↗

Jul 23 Cross-subject semantic decoding with shared-space alignment for generalized neural representation learning arXiv cs.LG source ↗

Jul 23 Reproducing Recurrent Transformers: The CoTFormer arXiv cs.LG source ↗

Jul 23 Language-specific versus cross-lingual knowledge graphs for implicit aspect identification in Arabic: a comparative study of reasoning and adaptation strategies arXiv cs.CL source ↗

Jul 23 Memory Merge DQN: sensitivity weighted target updates for stable value learning arXiv cs.LG source ↗

Jul 23 LAARA: Layer-Aware Adaptive Rank Allocation for parameter-efficient fine-tuning arXiv cs.LG source ↗

Jul 23 Task competence is not instruction following: evaluating instruction-conflicting behavior in small language models arXiv cs.CL source ↗

Jul 23 Decodable but Not Detectable: A Leakage Fingerprint for Near-OOD Benchmarks reveals a benchmark leak where removing the OOD class and retraining 35 models raises AUROC from 0.326 to 0.911. arXiv cs.LG source ↗

Jul 23 Neural operator surrogates for two-dimensional neutron flux estimation. arXiv cs.LG source ↗

Jul 23 Scaling laws for hypernetwork-based knowledge injection in large language models arXiv cs.CL source ↗

Jul 23 Lightweight system for person–place relation extraction in historical newspapers using dependency graphs and proximity features arXiv cs.CL source ↗

Jul 23 Multi-dimensional evaluation of explainability in media bias detection using the BABE dataset arXiv cs.CL source ↗

Jul 23 STN-TGAT: Top-K portfolio construction via prior-guided graph attention with learnable soft-threshold sparsification arXiv cs.LG source ↗

Jul 23 Sentence Splitter uncovers latent factual structure of sentences with a T5-based self-supervised framework for identifying semantic head-tail boundaries arXiv cs.CL source ↗

Jul 23 Predicting groundwater arsenic concentrations using graph neural networks arXiv cs.LG source ↗

Jul 23 VizRAG enhances retrieval-augmented generation with hypergraph visualization to organize complex n-ary facts among entities beyond binary relationships arXiv cs.CL source ↗

Jul 23 Recovering clinical utility under differential privacy: empirical validation of adaptive federated aggregation on heterogeneous cardiovascular datasets arXiv cs.LG source ↗

Jul 23 Predictive single cell foundation model for gene regulation and aging with privacy-preserving tabular learning arXiv cs.LG source ↗

Jul 23 Stateful guardrails for multi-turn LLM systems address conversational risk accumulation by tracking session-layer trajectory signals to detect gradual intent drift, fragmented instruction assembly, and repeated disclosures. arXiv cs.CL source ↗

Jul 23 Air Quality Arena: a large-scale multi-region ground monitoring dataset and benchmark for air quality forecasting with time-series foundation models arXiv cs.LG source ↗

Jul 23 Explainability challenges in continual learning for time series forecasting with Experience Replay strategies arXiv cs.LG source ↗

Jul 23 Reasoning narrows the move: diversity collapse in LLM game play arXiv cs.CL source ↗

Jul 23 Orthogonalized read improves noisy associative recall in mLSTM memory by reconditioning the learning problem during training plateau (arXiv:2607.19390) arXiv cs.LG source ↗

Jul 23 Native multi-dimensional subquadratic operators via input dependent long convolutions arXiv cs.LG source ↗

Jul 23 SUM: Unified Geometric Surgery on Spatio-Temporal Adaptation Vectors for Federated Class Incremental Learning arXiv cs.LG source ↗

Jul 23 On the computational complexity of structural generalization arXiv cs.CL source ↗

Jul 23 Pipeline choices dominate autointerpretability score variance in sparse autoencoder evaluations arXiv cs.LG source ↗

Jul 23 Embedding-based measurement of data diversity for NLP; emb-diversity tool assesses diversity using embeddings arXiv cs.CL source ↗

Jul 23 Reliability-aware distillation can harm some samples in low-resource Bangla summarization; standard knowledge distillation yields negligible ROUGE-L gains on BanSum Bangla arXiv cs.CL source ↗

Jul 23 FinMMEval 2026 Task 1 evaluates multilingual financial multiple-choice question answering in English, Chinese, Arabic, and Hindi. arXiv cs.CL source ↗

Jul 23 Adaptive capitulation: a structural failure mode of LLM responses in vulnerability contexts arXiv cs.CL source ↗

Jul 23 Reference-free evaluation of reasoning in open-ended question answering reveals a reasoning-based auditing framework that decomposes reasoning traces into segments and labels premise-target relations with natural language inference. arXiv cs.CL source ↗

Jul 23 When does consensus beat voting? A critical analysis of statistical label fusion in medical image segmentation arXiv cs.LG source ↗

Jul 23 From trajectories to prefixes: reusing teacher trajectories via replayed prefixes and online continuation arXiv cs.LG source ↗

Jul 23 CruiseBench: a real-flight-aligned N-CMAPSS benchmark for engine RUL prediction arXiv cs.LG source ↗

Jul 23 Reward-aware population scaling of evolutionary strategies in LLM fine-tuning arXiv cs.LG source ↗

Jul 23 Scale-aware learning of chaotic dynamics on unstructured meshes via binned spectral losses arXiv cs.LG source ↗

Jul 23 NMR elucidation treated as an agentic search problem; an autonomous agent using a frozen LLM matches graduate-level chemistry students in structure determination arXiv cs.LG source ↗

Jul 23 Tiny_schiller: a drop-in German drama corpus for small language models arXiv cs.CL source ↗

Jul 23 TriAgent: Divergence-aware multi-agent committees for cost-efficient financial sentiment analysis arXiv cs.CL source ↗

Jul 23 Learning the Arabic dialect continuum as a continuous space: a regression approach to speaker origin prediction arXiv cs.CL source ↗

Jul 23 Mitigating scaffolding collapse in Socratic tutors via representation alignment arXiv cs.AI source ↗

Jul 23 Beyond Tracking or Shortcut: Composition-bounded predictive states in poker autoregressive models arXiv cs.AI source ↗

Jul 23 Geometry-Guided Constraint Learning for LLM safety classification achieves near-perfect per-category accuracy with sparse autoencoder feature extraction, reducing constraint counts from K=4-25 to K=2 for most categories arXiv cs.AI source ↗

Jul 23 Benchmarking confidential GPU inference on NVIDIA H100 under Intel TDX arXiv cs.AI source ↗

Jul 23 NEXUS: Structured runtime safety for tool-using LLM agents arXiv cs.AI source ↗

Jul 23 FormulaSPIN: Self-play fine-tuning for natural language to spreadsheet formula generation arXiv cs.AI source ↗

Jul 23 LLMs update memory through shared latent structures rather than isolated facts, supporting the lifted representation hypothesis arXiv cs.AI source ↗

Jul 23 FORCE-Bench: a benchmark, dataset, and evaluation harness for agentic AI in enterprise finance arXiv cs.AI source ↗

Jul 23 CrackedPDFs: a controlled benchmark for hidden prompt injection in PDFs arXiv cs.AI source ↗

Jul 23 OpenEvoShield: Dual non-stationary continual defense for open-world multi-agent system attacks arXiv cs.AI source ↗

Jul 23 ITPEval benchmarks automated translation of formal proofs across Lean 4, Rocq, Isabelle, and HOL Light arXiv cs.AI source ↗

Jul 23 HyGRL: Adaptive Hybrid Graph Reasoning for Multi-Entity Questions arXiv cs.AI source ↗

Jul 23 AdaRoPE: not all attention heads should rotate and scale equally arXiv cs.AI source ↗

Jul 23 Spectral-LSH: sub-quadratic prompt compression via Krylov-projected locality-sensitive hashing arXiv cs.AI source ↗

Jul 23 Statistically grounded sparse-feature interventions for activation-space control in large language models arXiv cs.AI source ↗

Jul 23 Stochastic primal-dual decoding for multiobjective generative recommender systems arXiv cs.AI source ↗

Jul 23 GraphContainer provides a unified platform for comparing and debugging Graph RAG methods arXiv cs.AI source ↗

Jul 23 Hybrid LSTM-Graph neural framework for robust financial fraud detection and adversarial resilience arXiv cs.AI source ↗

Jul 23 Euclean: automated geometry problem formalization with unified verification in Lean arXiv cs.AI source ↗

Jul 23 LISA: Linear-Indexed Sparse Attention for efficient long-context reasoning arXiv cs.AI source ↗

Jul 23 FineServe: a fine-grained dataset and characterization of global LLM serving workloads arXiv cs.AI source ↗

Jul 22 SymptomAI: a conversational AI agent for everyday symptom assessment Google Research source ↗

Jul 22 Dimitri Bertsekas, professor emeritus and influential computer scientist, dies at 83. MIT News (AI) source ↗

Jul 22 Stochastic Meta-Unlearning: Bridging language backbone and multimodal unlearning arXiv cs.CL source ↗

Jul 22 Beyond single-dimensional compression: exploring compound sparsity to delay performance degradation in large language models arXiv cs.LG source ↗

Jul 22 On the limits of support-preserving alignment and bounded filtering; study examines whether alignment plus safety filters can drive harmful behavior to zero in large language models arXiv cs.LG source ↗

Jul 22 Dual Attention Residuals in transformers enable two complementary residual pathways for historical retrieval and multi-stream trajectories; DAR integrates both mechanisms arXiv cs.CL source ↗

Jul 22 European Multilingual Evaluation Dataset localized into 11 languages via the EMT network project; DGT and EMT collaborate on translating MMLU for LLM benchmarking arXiv cs.CL source ↗

Jul 22 Structured Output collapses answer diversity across 44 language models arXiv cs.CL source ↗

Jul 22 FedCC: a low-resource federated adaptation of foundation models for robust corpus callosum localization in fetal ultrasound images arXiv cs.LG source ↗

Jul 22 RF-Agent: A practical framework for building language agents for RFIC design arXiv cs.CL source ↗

Jul 22 PathReportEval provides a systematic benchmark for pathology report generation arXiv cs.CL source ↗

Jul 22 Convolution for large language models; study finds depthwise convolutions provide local inductive bias in Qwen3 Transformer blocks arXiv cs.CL source ↗

Jul 22 Interpreting how instruction-tuned transformers encode discourse relations, focusing on causation and antithesis, in next-token prediction tasks arXiv cs.CL source ↗

Jul 22 BearingNAS enables in-sensor fault diagnosis for bearings on a laptop with hardware-aware neural architecture search and extreme micro-budget constraints arXiv cs.LG source ↗

Jul 22 A domain-conditional position offset reduces the cold-start penalty for autoregressive language models by adding a learned vector to the first token embeddings while freezing model weights. arXiv cs.LG source ↗

Jul 22 Dual-domain fused LSTM model for efficient time-dependent reliability analysis arXiv cs.LG source ↗

Jul 22 Relay-Bench evaluates LLMs on multi-domain reasoning chains; GPT-5.5 (xHigh) scores 43.3% on composite problems across domains arXiv cs.CL source ↗

Jul 22 Reliability declines as model scale increases due to an auto-regressive risk regime that amplifies mistakes arXiv cs.LG source ↗

Jul 22 Gradient-energy guided block-wise perturbations for sharpness-aware minimization arXiv cs.LG source ↗

Jul 22 Spectral evidence bundling for selective reliability estimation in time-series classification arXiv cs.LG source ↗

Jul 22 From a multilingual streaming ASR backbone to Kenyan-language systems: data-centric adaptation of Nemotron 3.5 for Kikuyu, Dholuo, and Kalenjin arXiv cs.CL source ↗

Jul 22 E-SpecFormer: an edge-efficient transformer for end-to-end RF spectrum monitoring with LiTAN attention and four scalable variants arXiv cs.LG source ↗

Jul 22 FALCON-Discover ranks predictions by discrepancy signals to locate false-confidence regions for calibration arXiv cs.LG source ↗

Jul 22 Cost accounting for reactive computational graphs: exhaustive sweeps, sequential mutation, and the backward-locality gap arXiv cs.LG source ↗

Jul 22 HPD-Parsing: Hierarchical Parallel Document Parsing arXiv cs.CL source ↗

Jul 22 Using fine-tuned LLMs to identify indicators of vulnerability in UK police incident logs arXiv cs.CL source ↗

Jul 22 Rationale-Guided Knowledge Distillation for Cross-Lingual Stance Detection arXiv cs.CL source ↗

Jul 22 Preference-conditioned multi-objective reinforcement learning for runtime-tunable transit signal priority arXiv cs.LG source ↗

Jul 22 AILQA evaluates AI-driven legal question answering for the Indian legal system using embedding and generative models; arXiv researchers assess accuracy and reliability arXiv cs.CL source ↗

Jul 22 Reasoning fine-tuning induces persistent latent policy states arXiv cs.CL source ↗

Jul 22 Multi-Timescale Latent-Action DRL for joint optimization in edge-cloud networks reduces average end-to-end latency via joint service placement, computational delegation, and power control arXiv cs.LG source ↗

Jul 22 Narrative framing affects LLM agent behavior more than persona, as shown by three text-based investigation games with identical structure but different tasks arXiv cs.CL source ↗

Jul 22 Fusion Embedding adds audio to a frozen vision-language embedding base to cover text, image, video, and audio in a unified embedding space arXiv cs.CL source ↗

Jul 22 Agentic calibration of grey-box simulation models: an LLM-driven alternative arXiv cs.LG source ↗

Jul 22 Interactive Training 2: Auditable control plane for live model training arXiv cs.LG source ↗

Jul 22 Spatio-temporal prediction of unsteady airfoil aerodynamics using augmented graph neural ordinary differential equations with exogenous controls arXiv cs.LG source ↗

Jul 22 Towards principled continual anomaly detection: a systematic framework and benchmark scenarios arXiv cs.LG source ↗

Jul 22 Find Before You Fine-Tune: a diagnostic study of small LLMs for cybersecurity QA arXiv cs.CL source ↗

Jul 22 Computational models of pragmatic reasoning with flexible generation of meaning and expression alternatives arXiv cs.CL source ↗

Jul 22 LatentMT: machine translation with latent reasoning arXiv cs.CL source ↗

Jul 22 ALAS: Additive Learnable Alpha-Stable Kernels for flexible Bayesian optimization arXiv cs.LG source ↗

Jul 22 Self-improving, frozen-gate training (SIFT) enables dynamic document classification via a cheap, CPU-bound pipeline arXiv cs.CL source ↗

Jul 22 The information shadow: measuring structural limits on what language models can learn arXiv cs.LG source ↗

Jul 22 Uncertainty quantification for AI-driven crash simulation surrogates: Monte Carlo dropout versus deep ensemble on an open-source bumper beam benchmark. arXiv cs.LG source ↗

Jul 22 Search-on-Graph-R1: Training an 8B model to search knowledge graphs with reinforcement learning arXiv cs.CL source ↗

Jul 21 KernelBench-Verified: LLM-generated kernels often inflate performance; evaluation frameworks must evolve to measure true speedup, not reported gains. arXiv cs.LG source ↗

Jul 21 Democratizing AI with small language models: structured benchmarking and parameter-efficient fine-tuning for local deployment arXiv cs.AI source ↗

Jul 21 It Takes 8 tokens: weak-to-strong off-policy RL via auxiliary branches arXiv cs.AI source ↗

Jul 21 Symbolic augmentation closes a canonical-equivalence blind spot in neural fact-checkers arXiv cs.AI source ↗

Jul 21 TalTech submits BeTraC entries: robust summarization of long doctor-patient conversations into SOAP notes using Voxtral models with LoRA fine-tuning and DAPO RL arXiv cs.CL source ↗

Jul 21 Literary non-style in LLM-generated text arXiv cs.CL source ↗

Jul 21 NOWJ@COLIEE 2026: adaptive pipelines for legal retrieval and reasoning arXiv cs.CL source ↗

Jul 21 SEER: supervised learning to control energetic reasoning. arXiv cs.AI source ↗

Jul 21 Deterministic replay for AI agent systems arXiv cs.AI source ↗

Jul 21 Orthogonal gradient constraints shape noisy-label memorization dynamics arXiv cs.LG source ↗

Jul 21 PPO-HSC: An exploratory reinforcement learning framework based on wide-area policy coverage optimization arXiv cs.AI source ↗

Jul 21 Are Arithmetic Heuristic Neurons Form-Invariant? A mechanistic analysis of symbols, text, and code in LLMs arXiv cs.CL source ↗

Jul 21 When to plan: learning to select between reactive control and deliberative planning arXiv cs.AI source ↗

Jul 21 Normalized Rewards for Preference Optimization arXiv cs.LG source ↗

Jul 21 Diagnosing correctness probes under self-judgement confounding arXiv cs.CL source ↗

Jul 21 A survey on the verification of reinforcement learning policies arXiv cs.AI source ↗

Jul 21 Rater state bias in RLHF preference data; an audit framework arXiv cs.AI source ↗

Jul 21 A survey on GNN-based link prediction: techniques, applications, and challenges arXiv cs.AI source ↗

Jul 21 Auditing question-order effects in large language models with the QQ equality: mechanism characterization and a saturation caveat arXiv cs.CL source ↗

Jul 21 Shapley Context Pruning: A cooperative game perspective for context reranking and pruning arXiv cs.AI source ↗

Jul 21 Self-Evolving Just-In-Time memory for proactive embodied safety arXiv cs.LG source ↗

Jul 21 Operator-aware mixed-precision tolerance calibration for tensor kernels arXiv cs.LG source ↗

Jul 21 From Weights to Words: Expressing and Editing Preference Model Inferences in Natural Language arXiv cs.LG source ↗

Jul 21 High-accuracy low-bit KV-cache quantization via local distribution restoration arXiv cs.LG source ↗

Jul 21 Learning structural manipulability in gate-level netlists with graph neural networks arXiv cs.LG source ↗

Jul 21 Scope3Trace: evidence-based identification and extraction of Scope 3 GHG emissions from sustainability reports arXiv cs.CL source ↗

Jul 21 Cascading versus joint modeling for hierarchical offensive language detection arXiv cs.CL source ↗

Jul 21 LLM unlearning for cyber defense: a survey on methods, challenges, and emerging threats arXiv cs.LG source ↗

Jul 21 RAIL Guard closes the evaluation-to-remediation gap in responsible AI for LLM agents by using a closed-loop pipeline that evaluates outputs across eight dimensions and remediates failures through an evaluate-rewrite-reevaluate loop arXiv cs.AI source ↗

Jul 21 RouteCost: a production-inspired multi-stage framework for pre-order shipping cost estimation in e-commerce arXiv cs.LG source ↗

Jul 21 RIMS: preference optimization via smoothed multi-pair aggregation for small-scale LLM retrieval-augmented generation arXiv cs.CL source ↗

Jul 21 Fully-sensorized smart-eyewear platform enables on-device machine learning via STM32N6 NPU arXiv cs.LG source ↗

Jul 21 BACON: Budgeted human calibration for modeling and evaluation with multiple AI judges arXiv cs.LG source ↗

Jul 21 Schema-constrained document-level event argument extraction with lightweight LLM fine-tuning arXiv cs.CL source ↗

Jul 21 JOR-Bench: Japanese operations research benchmarks for large language models arXiv cs.CL source ↗

Jul 21 Learning from synthetic data without model collapse in iterative instruction tuning arXiv cs.CL source ↗

Jul 21 Let the data decide: off-policy distillation during continued pre-training analyzes objective-to-capability trade-offs and adaptive objective routing in supervision and performance arXiv cs.LG source ↗

Jul 21 Group entropy-controlled policy optimization arXiv cs.CL source ↗

Jul 21 Lightweight 1D CNN for affective touch classification in soft plush companions designed and validated; open-source MATLAB framework for compact deep learning models arXiv cs.AI source ↗

Jul 21 OpenMHC: accelerating the science of wearable foundation models arXiv cs.LG source ↗

Jul 21 Accurate and efficient long-term memory for LLM agents arXiv cs.AI source ↗

Jul 21 Multi-level context modeling for consistent expert selection in Mixture-of-Experts arXiv cs.CL source ↗

Jul 21 Committed before reasoning: evidence of answer pre-commitment in an open-weight LLM with behavioral reproduction and preliminary activation-level findings arXiv cs.CL source ↗

Jul 21 Generalist AI control: a learning-based controller handles systems with varying orders and dynamics using a dynamic state-space representation with attention and masking arXiv cs.AI source ↗

Jul 21 BLAD: a multilingual dataset of 1,484 Bangladeshi legal acts from 1799 to 2025 arXiv cs.CL source ↗

Jul 21 Reinforcement Learning-guided NSGA-II with gray relational coefficient for multi-objective optimization: application to NASDAQ portfolio optimization arXiv cs.LG source ↗

Jul 21 PlanFlip: attackers inject prompts in the planning phase to corrupt multi-agent LLM sub-tasks via planning-phase prompt injection arXiv cs.AI source ↗

Jul 21 OpenLanguageModel: readable, composable small-language-model pretraining for education and research arXiv cs.CL source ↗

Jul 21 Interactive task alignment framed as a POMDP to handle ambiguous user goals in task execution arXiv cs.AI source ↗

Jul 21 Berkeley and Heiserman as an unexhausted architecture for embodied machine intelligence arXiv cs.AI source ↗

Jul 21 Generative Ontology Induction: domain-agnostic schema discovery from document corpora using large language models arXiv cs.AI source ↗

Jul 21 DocOCR-Eval: a correction-based framework for OCR tool selection without ground truth arXiv cs.LG source ↗

Jul 21 Diffusion-corrected autoregressive Fourier neural operator for droplet evolution prediction arXiv cs.LG source ↗

Jul 21 HantaWatch: federated learning for hantavirus genomic surveillance arXiv cs.LG source ↗

Jul 21 SelKV: selective KV cache merging with per-token merge-or-drop and attention compensation arXiv cs.AI source ↗

Jul 21 Conformal prediction for self-correcting scientific generation provides statistical guarantees for scientific reasoning validity through progressive absolute-coherent-factuality validation arXiv cs.CL source ↗

Jul 21 Real-world evaluation of an AI agent drafting translational impact summaries arXiv cs.CL source ↗

Jul 21 ColGraphRAG: late-interaction evidence retrieval for multimodal GraphRAG arXiv cs.AI source ↗

Jul 21 Quantizing recursive reasoning models introduces cumulative bias with per-tensor 4-bit quantization, per the arXiv paper on recursive reasoning blocks arXiv cs.LG source ↗

Jul 21 Encoding EEG signals to examine human-like next-word prediction behavior in language models arXiv cs.CL source ↗

Jul 21 Token-Level cross-modal transformer with contrastive multi-task learning for breast cancer subtype classification and survival prediction arXiv cs.LG source ↗

Jul 21 SpecLA: Efficient speculative decoding for linear-attention models arXiv cs.CL source ↗

Jul 21 TRACE: trajectory-based safety patch learning for LLM post-training realignment arXiv cs.LG source ↗

Jul 21 Some large language models exhibit consistent risk attitudes arXiv cs.AI source ↗

Jul 21 AI_LectureNote: a retrospective pilot study of a post-ASR workflow for English-script rendering and semantic drift in Korean-English medical lectures arXiv cs.CL source ↗

Jul 21 The failures of marginal influence-based attribution methods for global time series explanations arXiv cs.LG source ↗

Jul 20 Loopie: a pair of Mixture-of-Experts looped transformers claim “most powerful looped Transformer to date” with 20B and 6B parameter variants arXiv cs.CL source ↗

Jul 20 DSWorld: a data science world model anticipates effects of data science operations to enable efficient autonomous agents arXiv cs.AI source ↗

Jul 20 From black box to executable logic: explainable reinforcement learning via Prolog expert systems arXiv cs.AI source ↗

Jul 20 Looped Latent Attention compresses K and V caches in looped transformers through a post-training codec, enabling compact cross-loop data storage arXiv cs.LG source ↗

Jul 20 ToolVerse enables large-scale environments and long-horizon tasks for agentic reinforcement learning arXiv cs.AI source ↗

Jul 20 Harmonizing AI safety thresholds arXiv cs.AI source ↗

Jul 20 CAMMAR: Culture-Aware Matryoshka for Metaphorical Arabic Representations arXiv cs.CL source ↗

Jul 20 S1-Omni is a unified multimodal reasoning model for scientific understanding, prediction, and generation. arXiv cs.AI source ↗

Jul 20 AgentFAIR: a multi-agent framework for FAIRness evaluation of geospatial datasets arXiv cs.AI source ↗

Jul 20 Independent certification for trustworthy AI could close the trust gap, argues for a formal certification framework arXiv cs.AI source ↗

Jul 20 Adaptive multi-step lookahead decoding for diffusion language models arXiv cs.CL source ↗

Jul 20 AnovaX: a local, multi-agent voice assistant with LLM planning, typed executors, and adaptive recovery arXiv cs.AI source ↗

Jul 20 Plan, Learn, Adapt (PLA) framework for personalized on-device itinerary generation arXiv cs.LG source ↗

Jul 20 Inpainting insights: elevating visual XAI with photorealistic perturbations arXiv cs.LG source ↗

Jul 20 Candidate Attended Dialogue State Tracking using BERT arXiv cs.CL source ↗

Jul 20 Deep Learning approaches for sleep apnea classification from polysomnographic EEG signals arXiv cs.LG source ↗

Jul 20 Causal-Audit: Explicit and auditable graph-based reasoning via target-aware causal chain construction arXiv cs.AI source ↗

Jul 20 SkillCorpus aims to consolidate and evaluate the open skill ecosystem for real-world LLM agents arXiv cs.CL source ↗

Jul 20 qZACH-ViT: Quantization-Aware Intrinsic Explanations with Recursive Attribution-Stabilized Optimization arXiv cs.LG source ↗

Jul 20 Information-Directed Sampling for causal bandits arXiv cs.LG source ↗

Jul 20 Toxicity priors show conditional reliability in multilingual and code-mixed abuse detection; English toxicity, Indic abuse, and rule-based severity cues help only in certain linguistic and abuse contexts (ToxGate). arXiv cs.CL source ↗

Jul 20 Large language models as unified multimodal learners for clinical prediction arXiv cs.CL source ↗

Jul 20 Before the action: benchmarking LLMs on prospective hypothesis discovery arXiv cs.CL source ↗

Jul 20 Regularity-aware stochastic MGDA with adaptive conflict-avoidant update direction control arXiv cs.LG source ↗

Jul 20 EpiNarrate: agentic generation of grounded narratives from epidemiological scenario projections arXiv cs.CL source ↗

Jul 20 From hyperplanes to hyperellipsoids: comparing interpretability of linear and single-qubit mixed-state binary classifiers arXiv cs.LG source ↗

Jul 20 Who became financially vulnerable after COVID-19? A population-level machine learning analysis using MEPS data arXiv cs.LG source ↗

Jul 20 Structure of the circular-dyadic convolution error arXiv cs.LG source ↗

Jul 20 Cura 1T: a specialized model for agentic healthcare arXiv cs.AI source ↗

Jul 20 Do coding agents need executable world models, simplification, and verification to solve ARC-AGI-3? arXiv cs.AI source ↗

Jul 20 DECODEM: Data extraction from corporate organizational documents via enhanced methods arXiv cs.CL source ↗

Jul 20 Frontier AI performance across business disciplines: a case-grounded benchmark of knowledge work and analytical reasoning arXiv cs.CL source ↗

Jul 20 On the structure of address in multi-party dialogue: moving from discrete labels to continuous levels arXiv cs.CL source ↗

Jul 20 Contextual semantic relevance tracks fMRI BOLD responses during naturalistic speech comprehension arXiv cs.CL source ↗

Jul 20 NeurOWL: an LLM-based neural-symbolic framework for incomplete OWL ontology reasoning arXiv cs.AI source ↗

Jul 20 DrawingVQA: a real-world benchmark for multimodal reasoning on construction drawings arXiv cs.AI source ↗

Jul 20 SciForge: a multimodal AI-native workbench for scientific discovery arXiv cs.AI source ↗

Jul 20 GraphDx: a cost-aware knowledge-enhanced multi-agent framework for sequential diagnosis arXiv cs.AI source ↗

Jul 20 Critical analysis of trustworthy AI tools, Mark frameworks, and the implementation chasms arXiv cs.AI source ↗

Jul 20 MLIR-based compilation method for large language models; AI Trading evaluates LLMs for technical market analysis arXiv cs.CL source ↗

Jul 20 Verbalizable representations form a global workspace in language models arXiv cs.CL source ↗

Jul 20 SeerGuard: a consequence-aware safety framework for mobile GUI agents using world model prediction arXiv cs.AI source ↗

Jul 20 Process reward informed tree rollout for effective multi-turn RL arXiv cs.CL source ↗

Jul 20 A transportable threshold-based framework for interpretable classification of medical data arXiv cs.LG source ↗

Jul 20 VarRate: Training-free variable-rate KV cache compression for long-context LLMs arXiv cs.CL source ↗

Jul 20 Controlling implicit shortcut reliance in L2 spoken English auto-markers arXiv cs.CL source ↗

Jul 20 Neuro-symbolic AI for LEED compliance: document-centric benchmarking, deterministic numeric checking, and when multimodal hurts arXiv cs.AI source ↗

Jul 20 ADS-C: Antidistillation Sampling for Classification arXiv cs.LG source ↗

Jul 20 Quantum program generation should prioritize validity over probabilistic scaling, arguing that scaling parameters to boost emergent reasoning is a directional error for quantum circuit synthesis due to strict mathematical constraints and a syntax-semantics gap. arXiv cs.LG source ↗

Jul 20 Robust peak-cost constrained reinforcement learning arXiv cs.LG source ↗

Jul 20 Stochastic Reset Pathfinding: path-level regret for cascading bandits over graph paths arXiv cs.LG source ↗

Jul 20 Behavioral controllability of agentic models for information extraction: from fixed workflows to reflective agents arXiv cs.AI source ↗

Jul 20 From plausible to actionable: a position on LLM self-explanations arXiv cs.CL source ↗

Jul 20 How much human label variation does formal semantic structure explain? Group-level effects and item-level ceilings in NLI arXiv cs.CL source ↗

Jul 20 Cost-efficient generative AI summarization for scalable automated essay scoring in educational assessment arXiv cs.CL source ↗

Jul 20 Precise but uncoupled: reviewer precision does not guarantee critique uptake in multi-agent math reasoning arXiv cs.AI source ↗

Jul 20 A Formally Grounded ODRL Evaluator: Implementation and Comparison arXiv cs.AI source ↗

Jul 20 BayesPO: Bayesian prompt optimization via parallel-tempered gradient-guided discrete MCMC arXiv cs.CL source ↗

Jul 20 Publicly-Verifiable Certificates for Statistical Algorithms show non-interactive proofs of learning with public, distributionally robust certification of a hypothesis's validity. arXiv cs.LG source ↗

Jul 20 Kolmogorov–Arnold Networks replace fixed activations with learned edge functions in a six-layer, 10M-parameter B-spline model; study reports high reconstruction of feed-forward edges and effects of pruning. arXiv cs.LG source ↗

Jul 20 Relevant and Irrelevant: A Renormalization Group Analysis of Transformer Attention arXiv cs.LG source ↗

Jul 20 Better starts, better ends: bootstrapped iterative self-reasoning distillation for compressed reasoning arXiv cs.CL source ↗

Jul 20 Beyond a joke: multi-angle reasoning for detecting and explaining harmful humor in memes arXiv cs.AI source ↗

Jul 20 Harness-in-the-loop learning: using data-generating harnesses to shape future foundation models through execution traces arXiv cs.LG source ↗

Jul 20 Auto-scaling approach for serverless environments using a multi-expert consensus mechanism arXiv cs.LG source ↗

Jul 20 Frontier language models struggle to copy text; 2D viewing improves assessment of copying behavior arXiv cs.CL source ↗

Jul 20 Diffusion models recover accurate mixture weights despite score function insensitivity arXiv cs.LG source ↗

Jul 20 MGDT: MLLM-guided diffusion transformer with relation-adaptive mixture-of-experts for multimodal knowledge graph completion arXiv cs.AI source ↗

Jul 20 Logic, optimization, and artificial intelligence; logic and optimization combine to enhance rule-based AI for transparency and reproducibility arXiv cs.AI source ↗

Jul 20 Knowledge-centric agents for workflow generation arXiv cs.AI source ↗

Jul 17 MIT probe prompts questions about belonging and problem-solving; assistant professor Bailey Flanigan reflects on a sense of belonging at MIT MIT News (AI) source ↗

Jul 17 Explainability research should prioritize foundations over ad-hoc methods; position paper argues for integrating explanations into end-to-end, human-in-the-loop systems. arXiv cs.LG source ↗

Jul 17 Value leakage: a language model's answers are shaped by its own values arXiv cs.LG source ↗

Jul 17 Just Keep Prompting: evaluating repetitive Socratic prompting in VLMs arXiv cs.CL source ↗

Jul 17 Quantum compositional NLP for Arabic applies pregroup grammar to map sentences to quantum circuits and mirror grammatical structure; Arabic morphosyntax tested in quantum topology arXiv cs.CL source ↗

Jul 17 LBA proposes textual hard-label adversarial attack under low query budgets arXiv cs.CL source ↗

Jul 17 UniSAGE unifies static and dynamic attributes with hyper-structure for hierarchical data modeling arXiv cs.CL source ↗

Jul 17 Latent communication between language model agents: channels, alignment, and the limits of text arXiv cs.CL source ↗

Jul 17 UzWordnet and Generative AI enable Uzbek language learning through game-playing arXiv cs.CL source ↗

Jul 17 Automatically evolving prompt guidelines for task-specific optimization arXiv cs.CL source ↗

Jul 17 Token Time Continuous Diffusion for Language Modeling introduces a diffusion language model that maps Gaussian noise to a token canvas in continuous space with per-token time steps. arXiv cs.CL source ↗

Jul 17 Eta Given Delta: Defining LLM tool efficiency with marginal tool utility arXiv cs.CL source ↗

Jul 17 Simplicity paradox: debunking myths about prompting and datasets for LLM evaluation arXiv cs.CL source ↗

Jul 17 MAPS models co-existing subjective perspectives and shared meaning in multi-agent cognitive dialogue arXiv cs.CL source ↗

Jul 17 Introspection fine-tuning (IFT): training small LLMs to introspect arXiv cs.CL source ↗

Jul 17 arXiv:2607.14112v1 paper shows language models have reliability ceilings from resolvable output uncertainty limits arXiv cs.CL source ↗

Jul 17 T5-CSBoost: adversarial perturbation resistant LLM fingerprinting arXiv cs.CL source ↗

Jul 17 CoEvoT introduces co-evolving chain-of-thought prompting for graph-LLM reasoning under distribution shift. arXiv cs.CL source ↗

Jul 17 ReportMedSAM: guiding segmentation through radiology reports arXiv cs.CL source ↗

Jul 17 Heterogeneous element-aware cross-version differencing of scientific documents via layout-aware alignment and structure-aware reasoning arXiv cs.CL source ↗

Jul 17 Budgeted subset refinement for execution-aware LLM research ideation arXiv cs.CL source ↗

Jul 17 Semantic register compression in multi-agent LLM cascades. arXiv cs.CL source ↗

Jul 17 Study presents first cross-dataset Urdu fake news detection using XLM-RoBERTa; length confound analysis included. arXiv cs.CL source ↗

Jul 17 refusal in the first half: a mechanistic study of the prefill jailbreak; four models show harm remains represented and refusal behavior drops to chance after a prefill prompt arXiv cs.CL source ↗

Jul 17 Study introduces Implicit Reasoning Steering via Concept Chaining to address LLM reasoning fragility arXiv cs.CL source ↗

Jul 17 Severance problem: LLMs are unaware of the person beyond the prompt arXiv cs.CL source ↗

Jul 17 CARPRT: Class-aware zero-shot prompt reweighting for black-box vision-language models. arXiv cs.LG source ↗

Jul 17 Explainable geospatial AI for satellite ground station siting using LiDAR-derived terrain intelligence arXiv cs.LG source ↗

Jul 17 Certified domain consistency for multi-domain retrieval with label-free per-domain contamination control and conformal risk guarantees arXiv cs.LG source ↗

Jul 17 QFireNet: Quantum-Enhanced U-Net for Wildfire Segmentation from Sentinel-2 Imagery arXiv cs.LG source ↗

Jul 17 Branching policy optimization for sandbox-native language agent reinforcement learning arXiv cs.LG source ↗

Jul 17 arXiv:2607.14174v1 study extends supervised lexicon-learning to 10-K filings and Item 1A risk-factor sections; trains sentiment scores on return and volatility labels at multiple aggregation levels arXiv cs.LG source ↗

Jul 17 Low-latency relay selection in NR-V2X vehicular networks using graph isomorphism networks with edge features arXiv cs.LG source ↗

Jul 17 Renew: Towards learning world models and repairing model exploitation from preferences arXiv cs.LG source ↗

Jul 17 Closed-loop knowledge dynamics saturate under internal feedback; external information can move knowledge states beyond current attractors, using a three-level operational framework with transition kernels. arXiv cs.LG source ↗

Jul 17 Time-to-event model using a temporal machine learning framework to predict ALS progression and healthcare utilization arXiv cs.LG source ↗

Jul 17 TEDDY, a decoder transformer trained on ICD-10 diagnoses from 1.6 million children's EHRs, models pediatric health risks. arXiv cs.LG source ↗

Jul 17 Long-term user engagement optimization through model-agnostic downstream rewards learning arXiv cs.LG source ↗

Jul 17 Researchers propose augmentations for robust and efficient imitation learning in streamed video games. arXiv cs.LG source ↗

Jul 17 Study finds privacy risks in federated learning for radiology reports; analyzing tokenizer-driven leakage via gradient inversion arXiv cs.LG source ↗

Jul 17 LIGO-PINN: learned initialization via gated optimization to alleviate convergence failures in physics informed neural networks arXiv cs.LG source ↗

Jul 17 MIDiff: Tackling sparsity and imbalance in mobile usage generation via multivariate-imaging diffusion arXiv cs.LG source ↗

Jul 17 Local additive feature attribution: a mathematical taxonomy and reporting checklist arXiv cs.LG source ↗

Jul 17 Lyapunov-guided framework LyaGuide enables stabilization of generative flows with post-training guidance arXiv cs.LG source ↗

Jul 17 NeuroGRIP: retrieval-augmented graph refinement for knowledge-grounded EEG seizure diagnosis arXiv cs.LG source ↗

Jul 17 COAT framework learns interpretable prescriptive policies from observational data using counterfactual estimates and mixed-integer optimization for constrained decision making arXiv cs.LG source ↗

Jul 17 Policy learning with missing treatment data; effects on average treatment effect and policy value estimation. arXiv cs.LG source ↗

Jul 17 Dysco: Dynamic Subspace Boosting to mitigate LoRA interference in federated learning arXiv cs.LG source ↗

Jul 16 Graded entity-familiarity readouts in language models: Polish adaptation, cross-language robustness, and refusal steering arXiv cs.CL source ↗

Jul 16 Text2Sign, a text-to-sign language video diffusion model, runs on a single NVIDIA L4 GPU as a baseline. arXiv cs.CL source ↗

Jul 16 Explaining reinforcement learning agents via inductive logic programming arXiv cs.AI source ↗

Jul 16 New paper explores federated learning and explainable AI integration for privacy-preserving transparent machine learning arXiv cs.LG source ↗

Jul 16 OriginBlame introduces record- and token-level data provenance for AI training datasets arXiv cs.AI source ↗

Jul 16 SPINE: bridging the cyber-physical gap with agentic AI arXiv cs.AI source ↗

Jul 16 Researchers introduce black-box interventional grounding audits to test LLM premise dependency via predicate substitution; method replaces target predicates with fresh symbols and re-runs models to assess reasoning changes. arXiv cs.AI source ↗

Jul 16 Probabilistic extension of neuro-symbolic AGI robots based on Belnap's typed intensional FOL arXiv cs.AI source ↗

Jul 16 Survey Examines Self-Improving Agentic Systems' Shift to Deployment arXiv cs.AI source ↗

Jul 16 Improving molecular property prediction in small language models using graph-based tools. arXiv cs.AI source ↗

Jul 16 Oracle Agent Memory as an enterprise memory substrate for long-horizon AI agents arXiv cs.AI source ↗

Jul 16 Researchers introduce DROPJ, a human-centered method for safe agent training and deployment in safety-critical environments with unknown dynamics and no reward function. arXiv cs.AI source ↗

Jul 16 CayleyR solves TopSpin puzzle by detecting cycle intersections in Cayley graphs arXiv cs.AI source ↗

Jul 16 Networked intelligence: active shared context graphs for human-AI team science arXiv cs.AI source ↗

Jul 16 AI-native insurance for agentic AI: pricing, underwriting, and end-to-end automation arXiv cs.AI source ↗

Jul 16 Cost-optimal foundation model deployment portfolio for transportation management arXiv cs.AI source ↗

Jul 16 Harness Handbook: making evolving agent harnesses readable, navigable, and editable arXiv cs.AI source ↗

Jul 16 Theory-Level Autoformalization: from isolated statements to unified formal knowledge bases arXiv cs.AI source ↗

Jul 16 Researchers present EZSMT Version 3, an extensible SMT-based CASP framework. arXiv cs.AI source ↗

Jul 16 Set-shifting benchmark tests how LLM agents adapt to hidden reliability shifts in tool availability arXiv cs.AI source ↗

Jul 16 LAPO uses leave-one-turn attribution for multi-turn search reasoning as self-generated process supervision. arXiv cs.AI source ↗

Jul 16 OpenRCA dataset highlights challenges in achieving high accuracy for root cause analysis on real-world telemetry data. arXiv cs.AI source ↗

Jul 16 Multi-Agent Collaborative Reasoning with Tool-Augmented Evidence for Urban Region Profiling arXiv cs.AI source ↗

Jul 16 AI advice reduces willingness to say “I don’t know,” even when the AI is wrong and accuracy is incentivized arXiv cs.AI source ↗

Jul 16 Safety Sentry: Context-aware human intervention via execute-ask-refuse routing arXiv cs.AI source ↗

Jul 16 LLM-powered agentic system enables automatic discovery of ordinary differential equations in biological systems arXiv cs.AI source ↗

Jul 16 Stocktake: Measuring the gap between perception and action in LLM agents with a fair oracle arXiv cs.AI source ↗

Jul 16 UESF-Bench benchmarks unified embodied seeking and following, addressing existing benchmarks' assumption of initial target visibility. arXiv cs.AI source ↗

Jul 16 FixItFlow automates troubleshooting guide generation from historical cloud incidents using large language models arXiv cs.CL source ↗

Jul 16 Safe-Psych, a sequential evaluation benchmark for LLMs in psychiatry, addresses incomplete information by requiring clarification. arXiv cs.CL source ↗

Jul 16 Do LLMs need architectural changes for simultaneous speech translation? A prefix-to-prefix data driven approach arXiv cs.CL source ↗

Jul 16 Researchers use persona vectors to audit open-weight LLMs, revealing 53 traits across four models in arXiv:2607.13162v1. arXiv cs.CL source ↗

Jul 16 RAGthoven presents multi-stage LLM pipeline for SemEval-2026 Task 1's multilingual humor generation in English, Spanish, and Chinese. arXiv cs.CL source ↗

Jul 16 Adaptive Filtering of the KV cache: diagnosing and correcting structural-role bias in LLM inference arXiv cs.CL source ↗

Jul 16 Gsm-Plus-Bn Introduces Perturbation-Based Benchmark for Bangla Mathematical Reasoning in LLMs; Addresses Lack of Evaluation in Bengali arXiv cs.CL source ↗

Jul 16 Discourse-aware policy analysis with argumentation: a hybrid LLM-symbolic framework for disaster governance arXiv cs.CL source ↗

Jul 16 Finding the right tables and columns: a benchmark and corpus-adaptive embeddings for SQL schema retrieval arXiv cs.CL source ↗

Jul 16 Researchers propose a meta-learning framework to address data scarcity in low-resource language alignment for multilingual LLMs; arXiv:2607.13315v1 arXiv cs.CL source ↗

Jul 16 Evaluation ability does not imply optimization utility; LLM-as-a-judge signals in closed-loop table recognition show weak judge signals on FinTabNet and OmniDocBench arXiv cs.CL source ↗

Jul 16 A POS tier automates annotation for low-resource language documentation; neural interlinear glossing for Irabu (Southern Ryukyuan) presented arXiv cs.CL source ↗

Jul 16 GFlowRL: scaling distribution-matching RL to large language models arXiv cs.CL source ↗

Jul 16 Demystifying On-Policy Distillation: Roles, Pathologies, and Regulations arXiv cs.CL source ↗

Jul 16 Small language models for biomedical data-to-text generation: a case study on medication leaflets and post-training alignment methods arXiv cs.CL source ↗

Jul 16 Cross-rubric generalization: training on one rubric set and evaluating on a different rubric set for critical thinking essay scoring arXiv cs.CL source ↗

Jul 16 Researchers introduce a benchmark and reference system for live captioning Sikh Kirtan; arXiv:2607.13457v1 presents a closed-vocabulary task requiring exact SGGS line transcription. arXiv cs.CL source ↗

Jul 16 Benchmarking cross-device agents in heterogeneous environments arXiv cs.CL source ↗

Jul 16 MyAG is a graph-based framework for designing and analyzing composable LLM agent systems with three graph abstractions. arXiv cs.CL source ↗

Jul 16 Study introduces BioASQ Task 14B 2026 system using hybrid retrieval and multi-model answer combination. arXiv cs.CL source ↗

Jul 16 Memory as a controlled process: learned adaptive memory management for LLM agents arXiv cs.CL source ↗

Jul 16 Analogical Deep Research: retrieving and integrating historical analogies for foresight analysis arXiv cs.CL source ↗

Jul 16 Automatic differentiation from scratch: how PyTorch computes gradients for physics-informed neural networks arXiv cs.LG source ↗

Jul 16 Researchers propose lightweight training strategy for efficient transfer learning by decoupling feature extraction and classifier optimization with margin-based weighted loss. arXiv cs.LG source ↗

Jul 16 Masking, fingerprinting, and privacy from discarded geometry: what your model threw away and why you'll want it back arXiv cs.LG source ↗

Jul 16 Targeted PD identifies only the components that process specific inputs to scale parameter decomposition in neural networks. arXiv cs.LG source ↗

Jul 16 Uncertainty-aware sequential decision rules for event-triggered LLM invocation in streaming systems arXiv cs.LG source ↗

Jul 16 TSSM model enhances global station weather forecasting by addressing accuracy limitations through temporal-variable-historical modeling. arXiv cs.LG source ↗

Jul 16 Disentangling knowledge states with ability and proficiency modeling for knowledge tracing arXiv cs.LG source ↗

Jul 16 STKAN: Kolmogorov-Arnold networks for spatio-temporal forecasting arXiv cs.LG source ↗

Jul 16 Researchers propose Samba, a hybrid Mamba model for audio-visual navigation, addressing inadequacies of existing frameworks since 2020. arXiv cs.LG source ↗

Jul 16 Researchers introduce CoDiffGRN, a method for gene regulatory network inference using BEELINE-KGC benchmark and co-evolutionary discrete diffusion. arXiv cs.LG source ↗

Jul 16 ShortOPD: recovering pruned LLMs with short-to-long on-policy distillation arXiv cs.LG source ↗

Jul 16 Hedgehog: hierarchical evaluation of drug generators through rigorous filtration arXiv cs.LG source ↗

Jul 16 SteinGate: tail-sensitive safe reinforcement learning via Stein discrepancy arXiv cs.LG source ↗

Jul 16 Concurrent image understanding and generation via self-correcting coupled Markov jump processes arXiv cs.LG source ↗

Jul 16 EMAGN uses learned clustering for scalable traffic forecasting. arXiv cs.LG source ↗

Jul 16 Muon emerges as strong optimizer for deep learning, outperforming Adam and AdamW; theoretical work interprets it as steepest descent under spectral norm. arXiv cs.LG source ↗

Jul 16 Deconstructing Actor-Critic: a large-scale empirical study of design components for practitioners arXiv cs.LG source ↗

Jul 16 Tabular foundation models for discrete choice estimation show limited performance due to row-independence assumptions arXiv cs.LG source ↗

Jul 16 Accuracy-preserving stability regularization for large-scale retail demand forecasting arXiv cs.LG source ↗

Jul 16 Agora: collective and permissionless internet-scale pretraining of large language models arXiv cs.LG source ↗

Jul 16 Weight feedback computes the Jacobian transpose locally in modern deep networks arXiv cs.LG source ↗

Jul 16 Where should RL post-training compute go? Model size, search, learning, and feedback arXiv cs.LG source ↗

Jul 16 arXiv:2607.13395v1: Enlightenment-style post-tuning without training for large models inspired by human 'enlightenment' concept arXiv cs.LG source ↗

Jul 16 Study compares KANs and MLPs on structured data classification using twelve datasets. arXiv cs.LG source ↗

Jul 16 GIFT teaches vision-language generative AI models to produce CAD programs for simulating and testing 3D objects. MIT News (AI) source ↗

Jul 15 3 questions: neural transparency and the future of AI design MIT News (AI) source ↗

Jul 15 Towards demystifying the creativity of diffusion models Google Research source ↗

Jul 14 Devavrat Shah develops system that uses tabular data for real-time planning at scale MIT News (AI) source ↗

Jul 14 JARVIS Challenge tests AI copilots in tough-tech engineering for jet engine design MIT News (AI) source ↗

Jul 13 MIT students help prevent cyberattacks at Bridgewater State University cyber range and security operations center MIT News (AI) source ↗

Jul 13 MIT researchers use three AI agents to create 3D scene recreations for robot training data MIT News (AI) source ↗

Jul 13 Researchers develop an evaluation procedure to test generative AI models for harmful capabilities without generating outputs, aiding auditors in identifying open-source models adapted to produce illegal content. MIT News (AI) source ↗

Jul 9 Aurora 1.5 adds 22 variables, hourly resolution, and probabilistic ensembles to extend open foundation models for weather and Earth-system applications. Microsoft Research source ↗

Jul 9 Tiny robot boats assemble floating structures and reconfigure themselves with minimal human direction MIT News (AI) source ↗

Jul 9 SensorFM aims for a general intelligence and interface for wearable health data Google Research source ↗

Jul 8 Flint: a visualization language for the AI era Microsoft Research source ↗

Jul 7 How novice coders can develop AI programs for military applications MIT News (AI) source ↗

Jul 7 Jesse Thaler named director of the Laboratory for Nuclear Science MIT News (AI) source ↗

Jul 7 AI model inference costs fall sharply; GPT-4-class costs drop from ~$30 per million tokens in 2023 to under $1 today BAIR (Berkeley) source ↗

Jul 6 Envisioning the future of computing prize submission titled "Superintelligence, Superintimate" by Rachel Sava; aims to preserve benefits of neurotechnology for all MIT News (AI) source ↗

Jul 1 MIT in the media: President Sally Kornbluth and ASU President Michael Crow discuss training the next generation of scientists for America’s evolving tech landscape MIT News (AI) source ↗

Jul 1 BAIR Graduate Showcase recognizes Ph.D. graduates from the Berkeley Artificial Intelligence Research Lab, class of 2026. BAIR (Berkeley) source ↗

← 2026-08

2026-06 →