AI news archive.
19,861 articles, newest first, from the outlets listed on the sources page.
- arXiv cs.AIresearchRLHarness: Co-evolving Procedural Skills with Reinforcement Learning for Long-horizon Multimodal Reasoning
arXiv:2609.32326v1 Announce Type: new Abstract: Multimodal reasoning requires models to preserve visual evidence through long decision chains while selecting appropriate procedures across diverse scenarios and rules. Whe…
- arXiv cs.AIresearchHyperReCo: Retrieving and Connecting Evidence with Hypergraph Neural Networks for LLM Multi-hop Reasoning
arXiv:2609.32327v1 Announce Type: new Abstract: Large language models (LLMs) have shown strong capabilities, with retrieval-augmented generation (RAG) supporting complex multi-hop reasoning by retrieving evidence distrib…
- arXiv cs.AIagentsEnabling Timely Guidance before Skill Retrieval: Retaining Helpful Warm Tips in Agent Context
arXiv:2609.32339v1 Announce Type: new Abstract: Reusable skills help LLM-based agents solve complex tasks, but the agent must receive guidance before it commits to an ineffective approach. Existing skill mechanisms often…
- arXiv cs.AIresearchALLOT: Budgeted Hybrid-Memory Routing for Knowledge Updates in LLMs
arXiv:2609.32344v1 Announce Type: new Abstract: For large language models (LLMs), parametric adaptation is costly when retrieval already suffices. We introduce ALLOT, a hybrid-memory routing framework that separates lear…
- arXiv cs.AIagentsFrom Trajectories to Grounded Preferences: Process Preference Synthesis via Interaction Element Graphs for Web PRMs
arXiv:2609.32351v1 Announce Type: new Abstract: Comparative Process Reward Models (PRMs) provide critical step-level guidance for autonomous web agents by evaluating state-conditioned preferences between candidate action…
- arXiv cs.AIagentsAuthorityLens: Rethinking LLM-Based Agent Systems Through the Lens of Authority
arXiv:2609.32378v1 Announce Type: new Abstract: LLM-based agents are increasingly deployed with authority over consequential resources and decisions in real systems. These agents often operate alongside human and LLM-bas…
- arXiv cs.AIagentsReward Hacking and Agent Containment Failure: A Monte Carlo Study Based on the 2026 Hugging Face Incident
arXiv:2609.32390v1 Announce Type: new Abstract: The July 2026 intrusion into Hugging Face production infrastructure showed how reward hacking can become an external cybersecurity incident when a capable agent encounters…
- arXiv cs.AIagentsSCLATE: a Substrate for Continual-Learning Agent Training and Evaluation
arXiv:2609.32391v2 Announce Type: replace Abstract: Continual-learning agents are systems of models, harnesses, and memory operating over long multi-session horizons. Evaluating and training them requires interleaving ta…
- arXiv cs.AIagentsBeyond Scripted Search: Sample-Efficient Reward Discovery via Agentic Black-box Optimization
arXiv:2609.32394v1 Announce Type: new Abstract: Designing dense reward functions for low-level reinforcement learning (RL) control remains difficult. Recent work uses large language models (LLMs) to iteratively generate…
- arXiv cs.AIresearchMemory as a cache: Exact context reuse and deletion by construction
arXiv:2609.32395v1 Announce Type: new Abstract: The KV cache of a transformer entangles every token's representation with its entire prefix: a passage encoded once cannot be reused under a different prefix or removed wit…
- arXiv cs.AIresearchFunction Over Form: Distributional Orthogonalization in Mixture-of-Experts with Replica Expert Mechanism
arXiv:2609.32398v1 Announce Type: new Abstract: The scaling of LLMs increasingly relies on MoE architectures to decouple active computation from total parameter count. However, the efficacy of MoE is often constrained by…
- arXiv cs.AIresearchOpening LLM Judges: Recovering Preference Signals Beyond the Final Verdict
arXiv:2609.32407v1 Announce Type: new Abstract: LLM judges are widely used to evaluate model outputs, but their verdicts can be unreliable: a judge may favor the worse answer for its position, length, or other surface fe…
- arXiv cs.AIresearchCarnator: Fast Text-to-Video Generation with Generation-Native Compatibility-Guided Cross-Request Reuse
arXiv:2609.32420v1 Announce Type: new Abstract: Video diffusion transformers produce high-quality videos, yet iterative denoising incurs substantial inference latency, limiting interactive and large-scale serving. Most e…
- arXiv cs.AIresearchMergeHEIR: Mitigating Multimodal Hallucinations as the Tax of Model Merging
arXiv:2609.32422v1 Announce Type: new Abstract: Model merging consolidates task-specialized experts into a single deployable model. However, we show that such capability consolidation incurs a merging tax of increased ha…
- arXiv cs.AIagentsPluginRSI: Recursive Improvement of Agent Harnesses with Reusable Plugins
arXiv:2609.32423v2 Announce Type: replace Abstract: The harness surrounding a language model is a central determinant of agent performance. Recent methods optimize harnesses by searching over complete programs, where ind…
- arXiv cs.AIagentsAuthorization Closure Graph: Minimal Repair for LLM Agents with Evolving User Instructions
arXiv:2609.32428v1 Announce Type: new Abstract: Tool-using large language model (LLM) agents increasingly perform state-changing actions that require user authorization. Yet existing approaches do not provide a principle…
- arXiv cs.AIresearchPrismQuant: Optimal Null-Space Rotations for Grouped Quantizers
arXiv:2609.32429v2 Announce Type: replace Abstract: Smaller activation outliers do not necessarily imply better low-bit quantization: their alignment with the quantizer matters. We introduce PrismQuant, a quantizer-aware…
- arXiv cs.AIagentsMulti-Agent System Search via Active Substructure-aware Policy Optimization
arXiv:2609.32430v1 Announce Type: new Abstract: LLMs enable multi-agent systems (MAS) to tackle complex tasks, but manually designing agent roles, prompts, and communication structures requires substantial expertise and…
- arXiv cs.AIresearchFrom Latents to Wires: Surgical Post-Editing on Large Language Models
arXiv:2609.32434v1 Announce Type: new Abstract: Given a large language model (LLM), can whoever holds the weights name a semantic target (e.g., the model's identity), locate the model components that produce it, and edit…
- arXiv cs.AIresearchControllable GNN Explanations via Multi-Metric Preference Selection
arXiv:2609.32436v2 Announce Type: replace Abstract: Mechanisms for generating GNN explanations are crucial for building trust and mitigating biases in Graph Neural Networks (GNNs), especially in high-stakes scenarios. Mo…
- arXiv cs.AIresearchForkLeft: Entropy-First Rollouts for Prefix-Aligned Autoregressive-to-Diffusion Distillation
arXiv:2609.32448v1 Announce Type: new Abstract: Autoregressive Next-Token Prediction (NTP) has enabled strong reasoning capabilities in language models, while Diffusion Language Models (DLMs) offer flexible token orders…
- arXiv cs.AIresearchPULSE: Identifying Demonstration-Utility Features with Sparse Autoencoders
arXiv:2609.32469v1 Announce Type: new Abstract: In-context learning is highly sensitive to demonstration choice, yet most methods select demonstrations using external query-demonstration similarity. Such criteria can mis…
- arXiv cs.AIresearchVPEvolve: A Self-Evolving Virtual Process Engineer for Computational Lithography
arXiv:2609.32473v1 Announce Type: new Abstract: Optical proximity correction (OPC) recipes grow as engineers add local rules to repair newly discovered lithography hotspots. Each correction can interact with existing rul…
- arXiv cs.AIresearchFrom Outcomes to Strategies: Learning Strategy Utility for Mathematical Reasoning
arXiv:2609.32482v1 Announce Type: new Abstract: Reinforcement learning with verifiable rewards has substantially improved mathematical reasoning. However, terminal correctness alone provides limited insight into the qual…
- arXiv cs.AIresearchSeparating Diagnosis from Disease Representation: Dual-View EEG Learning with Neural-Dynamics-Guided Deformation
arXiv:2609.32483v1 Announce Type: new Abstract: Electroencephalography (EEG)-based closed-loop neuromodulation calls for a subject-specific structured state, as opposed to a single disease probability, specifying which b…
- arXiv cs.AIresearchTowards Scalable Data Diversification for Language Model Pretraining via Leverage Score Sampling
arXiv:2609.32484v1 Announce Type: new Abstract: Data selection for language model pretraining faces a fundamental tension between quality and diversity. While quality filtering is empirically effective, it often induces…
- arXiv cs.AIresearchWhen Helpful Text Hurts: Option-Redirecting Bias in Vision-Language Models
arXiv:2609.32489v1 Announce Type: new Abstract: In tri-modal visual question answering (VQA), auxiliary text is commonly used to complement visual and textual inputs, yet its reliability is often uncontrolled. While prio…
- arXiv cs.AIagentsRepoMAS: Solving Progressively Specified Tasks with Issue-Driven Multi-Agent Systems
arXiv:2609.32490v2 Announce Type: replace Abstract: LLM-based multi-agent systems (MASs) have shown strong potential for solving complex tasks, but most assume that task requirements are sufficiently specified before exe…
- arXiv cs.AIresearchBeyond Prompt or Skill? Attribution-Guided Optimization of Modular LLM Programs
arXiv:2609.32492v1 Announce Type: new Abstract: Large language models can solve increasingly diverse reasoning tasks, yet their performance remains highly sensitive to task prompts, intermediate instructions, and the way…
- arXiv cs.AIagentsDAAF: From Failure Localization to Editable System Assets in LLM Agents
arXiv:2609.32498v1 Announce Type: new Abstract: Deployed LLM agents increasingly rely on persistent, versioned system assets such as routing rules, knowledge segments, prompt instructions, and reusable skills. Failure-lo…
- arXiv cs.AIresearchTreeRef-BFN: Equivariance-Free De Novo Molecule Generation based on 2D Topology and Internal 3D Geometry
arXiv:2609.32502v1 Announce Type: new Abstract: De novo 3D molecular generation jointly models molecular size, topology, and geometry. Most methods pre-sample molecular size and generate Cartesian coordinates, limiting v…
- arXiv cs.AIagentsLearning from Others, Acting for You: Cross-User Memory Sharing for LLM Agents
arXiv:2609.32511v1 Announce Type: new Abstract: Large language model (LLM) agents serving different users often solve related tasks, yet separate user histories can leave reusable experience inaccessible to other agents.…
- arXiv cs.AIagentsFrom Anomalies to Failures: Constructing Causal Error Graphs for Agentic Trace Diagnosis
arXiv:2609.32514v1 Announce Type: new Abstract: LLM-driven agents are increasingly deployed in complex applications, where long agentic traces make failures difficult to diagnose. Existing trace diagnosis methods often c…
- arXiv cs.AIresearchLocalProp: Neuro-Localized Memory-Efficient Backpropagation
arXiv:2609.32517v1 Announce Type: new Abstract: The current deep learning training paradigm employs end-to-end backpropagation, regardless of the training stage, i.e. pre-training or fine-tuning. However, backpropagating…
- arXiv cs.AIresearchSTR: Supervised Transcoder Replacement for Reducing Steering Side Effects
arXiv:2609.32519v1 Announce Type: new Abstract: Model steering can strengthen a target behavior while degrading other useful behaviors. We introduce Supervised Transcoder Replacement (STR) to reduce these side effects fo…
- arXiv cs.AIagentsMemAgent: Learning to Manage Heterogeneous Memory Providers for LLM Agents
arXiv:2609.32521v1 Announce Type: new Abstract: Current agents remain largely stateless across tasks, limiting their ability to continually improve from prior interactions and making memory essential for long-horizon age…
- arXiv cs.AIagentsBeyond Dyadic Memory: Interaction-Aware Multimodal Memory with Adaptive Agentic Retrieval for Multi-Party Spoken Conversations
arXiv:2609.32522v1 Announce Type: new Abstract: Long-term memory enables agents to accumulate information and reason across sessions, yet existing research primarily focuses on dyadic text or image-text conversations, le…
- arXiv cs.AIresearchAmbiModBench: Benchmarking Gene Perturbation Prediction Beyond Shared Responses
arXiv:2609.32527v1 Announce Type: new Abstract: Predicting cellular responses to genetic perturbations helps prioritize experiments in single-cell genomics, where exhaustive measurement is infeasible. While computational…
- arXiv cs.AIagentsFail Loudly: An Auditable Runtime for Agentic Data Analysis
arXiv:2609.32528v1 Announce Type: new Abstract: Large language models (LLMs) have enabled data-science agents to automate multi-step analyses over heterogeneous files. However, incorrect choices regarding data sources, s…
- arXiv cs.AIresearchLLMAdBench: A Human Preference Benchmark for Advertising in LLM Responses
arXiv:2609.32533v1 Announce Type: new Abstract: Inserting advertisements (ads) into consumer-facing LLM output is emerging as a new business model, but there is little shared evidence on how such ad insertion should be e…
- arXiv cs.AIresearchInterpretable Physics Informed WiFi Indoor Localization: Learning an Effective Access Point Geometry and Using It to Prune
arXiv:2609.32539v1 Announce Type: new Abstract: Deep learning models can achieve high accuracy for indoor localization, but their black-box nature limits interpretability and the reuse of learned information. We propose…
- arXiv cs.AIagentsPorimon: An LLM-Based Pok\'emon Battle Agent Enhanced by Long/Short-Term Knowledge Augmented Generation
arXiv:2609.32544v2 Announce Type: replace Abstract: In this paper, we use Pok\'emon Battles as a case study to investigate how to improve the performance of LLM-based agents in tasks that require opponent-aware planning…
- arXiv cs.AIresearchAre Vision-Language-Action Models Robust to One-Step Observation Perturbations?
arXiv:2609.32550v1 Announce Type: new Abstract: Understanding the safety risks of vision-language-action (VLA) models is essential for their deployment in the physical world. Existing safety research has mainly considere…
- arXiv cs.AIresearchArtificial intelligences and human scientists exhibit complementary strengths in theory building
arXiv:2609.32562v1 Announce Type: new Abstract: We investigate the effectiveness of artificial intelligences (AI)-specifically large language models (LLMs)-relative to human scientists at high-level cognitive tasks in so…
- arXiv cs.AIresearchProTTT: Learning to Learn Semantic User Memory with Test-Time Training
arXiv:2609.32564v1 Announce Type: new Abstract: Personalization requires language models to capture user-specific knowledge from a growing user history. Existing context-based approaches incur increasing inference costs…
- arXiv cs.AIagentsCUE-Mem: Benchmarking Long-Term User Memory via Implicit Cues in Multimodal Conversations
arXiv:2609.32574v1 Announce Type: new Abstract: Long-term memory is essential for multimodal agents that interact with users across sustained conversations. However, user memories are not always explicitly stated: they m…
- arXiv cs.AIagentsEMIR$^2$: Evolution-Aware Memory with Intent-Guided Multi-Round Retrieval
arXiv:2609.32584v1 Announce Type: new Abstract: Long-term memory enables large language model (LLM) agents to leverage historical interactions for future tasks. However, existing memory systems struggle to utilize contin…
- arXiv cs.AIagentsMA-FPPO: Multi-Agent Flow-Pretrained Policy Optimization
arXiv:2609.32594v1 Announce Type: new Abstract: Multi-agent flow policies learn cooperative behavior from fixed offline datasets, but often struggle to complete tasks in situations not covered by the offline data. In the…
- arXiv cs.AIagentsCUA-SWE: When Computer-Use Agents Meet Visual Software Engineering
arXiv:2609.32600v1 Announce Type: new Abstract: Software development requires more than editing code: developers repeatedly run software, interact with its interfaces, visually inspect its behavior, and use these observa…
- arXiv cs.AIagents"You're Right, Let Me Fix It": How LLM Agents Damage Correct Work When Falsely Accused
arXiv:2609.32616v1 Announce Type: new Abstract: LLM agents increasingly keep working after a task succeeds as they resume after compaction or take over handoffs. Their finished work keeps receiving follow-up input that s…