AI news archive.
10,661 articles filed under Research, newest first, from the outlets listed on the sources page.
- arXiv cs.AIresearchCan Open-Weight Large Language Models (LLMs) Simulate Human Survey Populations? A Cross-Instrument Calibration Study
arXiv:2609.32638v1 Announce Type: new Abstract: Large language models (LLMs) are increasingly used to generate synthetic survey respondents and digital twins of real people, but whether their output preserves real human…
- arXiv cs.AIresearchPrediction Limits and Koopman Closure of Geometry-Induced Soft State Abstractions
arXiv:2609.32652v3 Announce Type: replace Abstract: Soft state representations assign states nonnegative class weights that sum to one, but agreement with reference labels does not guarantee accurate prediction of the ne…
- arXiv cs.AIresearchMixBench-TS: A Multivariate Time Series Forecasting Benchmark Where Channel Mixing Pays Off
arXiv:2609.32656v1 Announce Type: new Abstract: Multivariate Time Series Forecasting (MTSF) models that mix information across channels assume that the past of one channel carries information about the future of another.…
- arXiv cs.AIresearchWorld Models with Predictable Long-Horizon Marginals
arXiv:2609.32657v1 Announce Type: new Abstract: Accurate one-step predictions do not ensure that a world model's rollouts retain the data distribution. We make the model's decoded stationary law explicit by learning a de…
- arXiv cs.AIresearchLearning from a Thoughtful Teacher: Adaptive On-Policy Self-Distillation for Mathematical Reasoning
arXiv:2609.32667v1 Announce Type: new Abstract: On-policy self-distillation (OPSD) trains a question-only student with token-level feedback from a teacher given training-only privileged information (PI). OPSD therefore p…
- arXiv cs.AIresearchRefinement Symmetry in Multimodal Transformers
arXiv:2609.32669v1 Announce Type: new Abstract: Attention weights depend on token counts, which change with the representation of a signal. We study refinement symmetry: splitting a representation while preserving conten…
- arXiv cs.AIresearchWhat Would Falsify It? A Variable Specific Evidence Standard for Mechanistic Claims About Self Explanation
arXiv:2609.32670v1 Announce Type: new Abstract: When a language model explains an answer it has already given, does it reuse the computation that produced the answer or reconstruct a story from the answer alone? Attribut…
- arXiv cs.AIresearchExpected Reasoning-Step Return Unifies On-Policy Learning from Rewards and Teachers
arXiv:2609.32674v1 Announce Type: new Abstract: On-policy reasoning models can learn from task rewards or teacher signals, but these sources differ in form and can favor conflicting updates, leaving unclear which should…
- arXiv cs.AIresearchPINNMorph: Evolving Online Adaptation Policies for Physics-Informed Neural Networks
arXiv:2609.32685v1 Announce Type: new Abstract: Physics-informed neural networks (PINNs) provide a learning-based framework for solving partial differential equations (PDEs), yet their training behavior can change substa…
- arXiv cs.AIresearchFlat-Consensus Diffusion for Robust Data Reshaping under Noisy Evaluator
arXiv:2609.32696v1 Announce Type: new Abstract: Data shape determines how features are structured, how patterns are separated, and how distributions cover the underlying domain. Poor data shape can make models learn nois…
- arXiv cs.AIresearchCoWindow Attention: Full Causal Coverage Is a Collective Property
arXiv:2609.32704v1 Announce Type: new Abstract: FullAttn repeatedly exposes the complete causal history to every attention head, creating substantial redundant computation and memory traffic even with IO-efficient dense…
- arXiv cs.AIresearchMassAlloc Attention: Let Attention Allocate Its Own Compute
arXiv:2609.32712v1 Announce Type: new Abstract: FullAttn often assigns negligible normalized mass to much of the causal score space, yet dense kernels execute the complete post-score path after forming each QK tile. We i…
- arXiv cs.AIresearchRetrospective Distillation Attribution via Normalized Response Similarity
arXiv:2609.32749v1 Announce Type: new Abstract: Model distillation transfers capabilities through supervised fine-tuning (SFT) on teacher responses, often collected from commercial APIs, raising questions of model proven…
- arXiv cs.AIresearchAction Shaping: Policies Absorb What They Can Express
arXiv:2609.32752v2 Announce Type: replace Abstract: Reward shaping has a theorem: a potential-based term can be removed without changing the optimal policy. The same practice on the action channel, an offset added in tra…
- arXiv cs.AIresearchReadout is not Recovery: Dissociating Coordinate Emission from Visual-Corruption Repair in Vision-Language Models
arXiv:2609.32757v1 Announce Type: new Abstract: VLM bounding-box localization is both language generation and spatial commitment. Parseable fields such as bbox_2d make localization easy to score, but dimensions that emit…
- arXiv cs.AIresearchMandela-Bench: Multimodal Models Remember Canonical Images Instead of Seeing Them
arXiv:2609.32763v1 Announce Type: new Abstract: Historical photographs and other canonical images can now be edited seamlessly with a single instruction, often leaving no reliable pixel-level trace. In such cases, the on…
- arXiv cs.AIresearchForecasting Intraday USD/CAD Exchange Rate with News-Derived Monetary-Policy Signals
arXiv:2609.32773v1 Announce Type: new Abstract: Monetary-policy announcements and central-bank communications play a central role in foreign exchange markets, yet their qualitative, unstructured form makes their forecast…
- arXiv cs.AIresearchLearning response-aware patient dynamics for respiratory support
arXiv:2609.32782v3 Announce Type: replace Abstract: Respiratory support can shape the short-term physiological trajectory of critically ill patients, but patients receiving the same intervention may follow different phys…
- arXiv cs.AIresearch$T^5$: Twin-Critic Training for Token-Level Thoughts in Reinforcement Mid-Training
arXiv:2609.32791v2 Announce Type: replace Abstract: Reinforcement mid-training lets language models learn internal thoughts from unlabeled text, but efficient token-level credit assignment remains challenging. Existing g…
- arXiv cs.AIresearchBeyond Accuracy: Counterfactual Fragility and Demographic Bias in Clinical Evaluation of LLMs
arXiv:2609.32807v2 Announce Type: replace Abstract: Clinical LLM evaluation often emphasizes answer accuracy; however, accuracy alone does not test counterfactual consistency or demographic robustness. We evaluated six L…
- arXiv cs.AIresearchOverwhelmed by Choice: Studying LLM Decision Making at Scale
arXiv:2609.32809v1 Announce Type: new Abstract: Multiple-choice and candidate-selection evaluations are widely used to assess LLM reasoning and decision-making, yet most benchmarks contain relatively small candidate sets…
- arXiv cs.AIresearchRight Answer, Wrong Reason: Accuracy, Consistency, and Consensus Are Misleading Indicators of LLM Faithfulness in Clinical Decision Support
arXiv:2609.32817v2 Announce Type: replace Abstract: Clinical Large Language Models (LLMs) achieve strong medical-exam accuracy; however, a correct answer does not guarantee that the explanation names the concepts that ac…
- arXiv cs.AIresearchWhen Can First-Order Models of Fine-Tuning Bound Forgetting?
arXiv:2609.32818v1 Announce Type: new Abstract: Fine-tuning a language model on new data can make it forget facts that it should keep. We ask whether measurements taken at the start of a fine-tuning run can bound, for ea…
- arXiv cs.AIresearchRank Collapse Is Recoverable, Growing $|Q|$ Is Not: Out-of-Sample Early Warning for Value Divergence in High-UTD Soft Actor-Critic
arXiv:2609.32819v1 Announce Type: new Abstract: Raising the update-to-data (UTD) ratio breaks off-policy critics in two ways grouped as "plasticity loss": collapsing representations and growing value magnitude $|Q|$. We…
- arXiv cs.AIresearchRouting Drift Alone Does Not Diagnose Failure in Merged MoE LLMs
arXiv:2609.32821v1 Announce Type: new Abstract: Model merging efficiently combines specialized large language models (LLMs) without joint retraining, but can substantially alter expert routing in Mixture-of-Experts (MoE)…
- arXiv cs.AIresearchThe Decomposition Tax: LLM Pipelines Lose Up to 40 Accuracy Points at Their Own Interfaces
arXiv:2609.32825v1 Announce Type: new Abstract: A four-stage LLM pipeline gives up as much as 40.5 accuracy points at its own interfaces (gemma-3-12B on MATH-500, Holm-corrected p = 1.66e-19; the largest tax in the prima…
- arXiv cs.AIresearchCan LLMs Predict the Future? A Brier Score Analysis of Prediction Markets
arXiv:2609.32885v1 Announce Type: new Abstract: We study whether model upgrades improve probability estimates for prediction-market questions. Our Resolved Market Forecasting (RMF) benchmark contains 3,000 resolved binar…
- arXiv cs.AIresearchStraTune: Adaptive Selection of Revision Operators for Self-Evolving LLM Skills
arXiv:2609.32886v1 Announce Type: new Abstract: Large language models (LLMs) can learn reusable textual skills from execution feedback without updating their parameters, but effectively deciding how to revise these skill…
- arXiv cs.AIresearchConstraints Are Graphs, Not Chains: Exact Decoding for Diffusion Language Models
arXiv:2609.32900v1 Announce Type: new Abstract: Diffusion language models (dLLMs) predict masked positions in arbitrary order, but their exact constrained decoders still encode constraints as sequential languages, whose…
- arXiv cs.AIresearchLogical subspace in LLMs
arXiv:2609.32907v1 Announce Type: new Abstract: Recent work has identified a human brain network specialized for abstract formal reasoning (Kean et al., 2025). Does the same hold true in language models? To answer this q…
- arXiv cs.AIresearchPrecision As You Need: Stochastic Computing Is a Dense Adaptive Quantizer
arXiv:2609.32922v1 Announce Type: new Abstract: Matrix multiplications dominate the inference cost of modern transformer-based vision models, yet existing efficiency techniques such as post-training quantization and mixe…
- arXiv cs.AIresearchTRACE: Learning to Self-Calibrate Wireless Digital Twins from ISAC Measurements
arXiv:2609.32923v2 Announce Type: replace Abstract: Wireless digital twins (DTs) rely on 3D environment models to predict radio propagation and support wireless-network decisions, yet these models are often initialized f…
- arXiv cs.AIresearchDiagnosing Sampled LLM Reasoning in Formal Geometry: Coverage, Realization, and Validity Evidence
arXiv:2609.32924v1 Announce Type: new Abstract: Repeated sampling can reveal a correct numerical answer without yielding either a reliable system output or a supported derivation. We present Coverage, Realization, and Va…
- arXiv cs.AIresearchThe Commit-Abstain Circuit: Why Language Models Hallucinate Instead of Abstaining
arXiv:2609.32964v2 Announce Type: replace Abstract: Language models (LMs) often hallucinate by committing to confident answers rather than abstaining, even when they do not have enough information to answer reliably. A l…
- arXiv cs.AIresearchModel-Aware Data Selection from In-and-Out Information Interplay
arXiv:2609.33010v1 Announce Type: new Abstract: LLMs are effective representations that assimilate vast amounts of knowledge during pretraining, but post-training is necessary for models to reliably access this knowledge…
- arXiv cs.AIresearchBudgetVerify: Budget-Tiered Verification for Financial QA
arXiv:2609.33052v1 Announce Type: new Abstract: Financial question answering often requires precise numerical extraction, unit handling, and arithmetic over tables and text, but applying expensive verification uniformly…
- arXiv cs.AIresearchLarge Language Models Substantially Compress Well-Being Dispersion but Largely Preserve Its Socioeconomic Structure
arXiv:2609.33055v2 Announce Type: replace Abstract: Research using large language models (LLMs) to generate synthetic populations shows that model outputs compress human variation. We ask whether this compression also di…
- arXiv cs.AIresearchLLM sequential decision making under uncertainty in biochemical domains
arXiv:2609.33061v1 Announce Type: new Abstract: Large language models (LLMs) are increasingly used to drive scientific discovery. Understanding how LLMs make decisions from new data and memory of the literature is vital…
- arXiv cs.AIresearchQureRadEmbed: Structuring Radiological Similarity through Attribute and Reasoning Supervision
arXiv:2609.33075v1 Announce Type: new Abstract: Radiological similarity depends on disease relationships and on fine details such as laterality, lobe, severity, size, and certainty. Broad biomedical similarity can overlo…
- arXiv cs.AIresearchStructure-Mapping-Guided Self-Explanation for Learning Mathematical Procedures
arXiv:2609.33079v1 Announce Type: new Abstract: Worked examples are a powerful form of instruction, but learners must infer how the demonstrated steps were produced. A naive simulation of this self-explanation process ca…
- arXiv cs.AIresearchThe Model Knows Another Way: Strategy Switching for Effective RLVR Exploration
arXiv:2609.33085v1 Announce Type: new Abstract: Reinforcement learning with verifiable rewards (RLVR) is often limited by insufficient exploration: difficult problems can yield uniformly incorrect rollout groups and ther…
- arXiv cs.AIresearchModular Discovery of General Game-Playing Algorithms with Large Language Models
arXiv:2609.33115v1 Announce Type: new Abstract: General Game Playing across arbitrary games from rules alone remains challenging due to differing algorithmic requirements across game classes and strict decision-time cons…
- arXiv cs.AIresearchCeiling of a Task: When Can a Transformer Succeed Without Its Chain of Thought?
arXiv:2609.33134v1 Announce Type: new Abstract: Reasoning models generate long chains of thought before they answer, yet it is debated whether the content of these chains does real computational work or is largely decora…
- arXiv cs.AIresearchNot Too Hard, Not Too Easy: Learning from Intermediate States for LLM Structured Reasoning
arXiv:2609.33149v2 Announce Type: replace Abstract: A common principle of effective learning is to practice material that is neither already mastered nor too difficult to permit progress. We ask how to apply this princip…
- arXiv cs.AIresearchSeOPD: Self-Evolving LLMs via Online Policy Distillation from Self-Generated Chain-of-Thought
arXiv:2609.33181v2 Announce Type: replace Abstract: Recent advances in online policy self-distillation (OPSD) have demonstrated that large language models (LLMs) can improve their capabilities by leveraging external priv…
- arXiv cs.AIresearchUnlocking Latent Personalization in LLMs
arXiv:2609.33182v1 Announce Type: new Abstract: Large language models (LLMs) are increasingly expected to adapt to individual users, yet effective personalization remains challenging when only limited user-specific sampl…
- arXiv cs.AIresearchAre Benchmarks Reliable? Toward Structural Diagnosis via Item-Level Capability Boundaries
arXiv:2609.33196v2 Announce Type: replace Abstract: Evaluating large language models (LLMs) relies heavily on benchmark scores, yet aggregate metrics can obscure whether individual items serve their intended measurement…
- arXiv cs.AIresearchStructured Sparse Memory for Recurrent Reasoning
arXiv:2609.33270v1 Announce Type: new Abstract: Recurrent models trained from scratch have recently become competitive on ARC-style reasoning tasks, but the usual framing around small recurrent backbones overlooks two im…
- arXiv cs.AIresearchNext Thoughts Are Distributions: Generative Autoregressive Reasoning in the Latent Space
arXiv:2609.33271v1 Announce Type: new Abstract: Reasoning problems often admit multiple valid ways to proceed. Continuous reasoning promises to move computation beyond language tokens into a more compact latent space, bu…
- arXiv cs.AIresearchChronoFlow: Hierarchical Flow Matching for Irregular Time Series Generation
arXiv:2609.33276v1 Announce Type: new Abstract: Recent advances in generative modeling have substantially improved time series generation, yet most existing methods either assume a regular temporal grid or focus on featu…