AI news archive.
19,861 articles, newest first, from the outlets listed on the sources page.
- arXiv cs.AIresearchDraftAttention2: Fast Video Diffusion with Low-Resolution-Guided Mixed-Precision Attention
arXiv:2609.32628v1 Announce Type: cross Abstract: Video generation has broad applications in content creation and entertainment. Diffusion transformers have advanced the quality of generated videos, but attention over sp…
- arXiv cs.AIagentsTrust the Brand, Lose Control: How Identity Hijacks LLM Agent Orchestration
arXiv:2609.32635v1 Announce Type: cross Abstract: LLM agents now execute tasks end to end with permission to change real systems and increasingly orchestrate subagents that differ in capability and cost. Prior work treat…
- arXiv cs.AIresearchAsk Without Telling: Local SLMs Consult Cloud LLMs Without Revealing Task Intent
arXiv:2609.32642v1 Announce Type: cross Abstract: As local small language models (SLMs) increasingly collaborate with more capable cloud large language models (LLMs), a natural privacy question arises: Can a local SLM ob…
- arXiv cs.AIresearchInterTab: Interleaved Visual-Structure Alignment for Multi-Modal Table Reasoning
arXiv:2609.32660v1 Announce Type: cross Abstract: Table images preserve structural information that are often lost in text serialization, and reasoning over them requires locating relevant rows, columns, and cells step b…
- arXiv cs.AIagentsAsynCodeBench: Benchmarking Collaboration of Asynchronous Multi-Agent Systems in Software Engineering
arXiv:2609.32662v1 Announce Type: cross Abstract: Multi-agent coding has emerged as an increasingly active direction in software engineering, where complex development tasks are decomposed across multiple specialized age…
- arXiv cs.AIresearchTimestep Weighting: A Hidden Key to Effective ELBO-Based Flow-Matching RL
arXiv:2609.32665v2 Announce Type: replace-cross Abstract: ELBO-based reinforcement learning offers a sampler-agnostic approach to fine-tuning flow matching models with reward feedback. Timestep weighting in ELBO-based RL…
- arXiv cs.AIresearchSTAMP: Predicting Out-of-Distribution Generalization without Target Data
arXiv:2609.32672v1 Announce Type: cross Abstract: Predicting whether a trained model will generalize under distribution shift remains difficult, especially when target-domain data are unavailable. We introduce STAMP (Sem…
- arXiv cs.AIagentsThe GUI Is Not the State: Diagnosing State Aliasing in GUI World Models
arXiv:2609.32679v1 Announce Type: cross Abstract: GUI World Models (GUI-WMs) are increasingly used to predict future states for agent planning and simulation, yet most existing formulations condition only on the current…
- arXiv cs.AIresearchMM-OPD: Towards One More Bottleneck Between Perception and Reasoning
arXiv:2609.32690v1 Announce Type: cross Abstract: Recent multimodal large language models (MLLMs) advance visual reasoning by strengthening both perception and reasoning, implicitly assuming a process that transitions se…
- arXiv cs.AIresearchLearning to Refer: Client-Resolved Generation for Privacy-Aware Language Models
arXiv:2609.32706v1 Announce Type: cross Abstract: Cloud-based large language models (LLMs) require users to disclose plaintext data to service providers, creating privacy risks in sensitive domains. Existing privacy-pres…
- arXiv cs.AIresearchLLM Alignment--Utility Asymmetry under Semantic-Preserving Transformations
arXiv:2609.32717v2 Announce Type: replace-cross Abstract: Large Language Model (LLM) alignment is intended to ensure that models remain helpful and safe, but its stability under input distributional shift is not yet full…
- arXiv cs.AIresearchREALIS: A Curated Dataset for Studying the Challenges of AI Image Detection
arXiv:2609.32734v1 Announce Type: cross Abstract: AI-generated image detectors are often evaluated on benchmarks where real and synthetic images differ in content, quality, or generation artifacts, allowing models to rel…
- arXiv cs.AIresearchGradient-Guided Decoupled Adaptation for Geospatial Vision-Language Models
arXiv:2609.32737v1 Announce Type: cross Abstract: Existing geospatial vision-language models (Geo-VLMs) typically optimize diverse geospatial tasks through a unified multi-task adaptation paradigm without explicitly acco…
- arXiv cs.AIagentsSelf-Evolving Multi-Agent Symbolic Discovery for Financial Fundamental Analysis
arXiv:2609.32746v1 Announce Type: cross Abstract: While symbolic regression (SR) has been successfully used in science to discover new equations, its use in financial valuation is hindered by several limitations. Whereas…
- arXiv cs.AIresearchHow Far Do Persona Effects Generalize in Language Models?
arXiv:2609.32758v1 Announce Type: cross Abstract: Persona prompts ask language models to answer as particular kinds of people. We test whether relationships learned from these effects predict responses to new questions a…
- arXiv cs.AIresearchThe Extender: A Log-Structured Transformer
arXiv:2609.32759v1 Announce Type: cross Abstract: We introduce the Extender, a log-structured variant of the standard Transformer architecture. In a standard Transformer, each layer communicates with subsequent layers ex…
- arXiv cs.AIresearchContinual Learning via Self-Probe Gradients
arXiv:2609.32771v1 Announce Type: cross Abstract: Adapting pretrained models to new data can cause catastrophic forgetting of previously learned behavior. When only a few past samples remain, they give continual learning…
- arXiv cs.AIresearchLearnable Randomization as Commitment Against Adaptive Optimizers
arXiv:2609.32772v1 Announce Type: cross Abstract: A pricing page can walk the posted price up to the last amount a buyer still accepts, a recommender can hold back a better item for a barely acceptable promoted one, and…
- arXiv cs.AIresearchGetting Motif-ated: Controllable AI Compositions from Injected Motif Prompts
arXiv:2609.32788v1 Announce Type: cross Abstract: Deep learning has transformed symbolic music generation by borrowing the training paradigms of large language models, with systems such as NotaGen now producing complete,…
- arXiv cs.AIresearchFinding Emotions Where They Belong: Rethinking Audio Emotion Recognition through Masked Temporal Affective Grounding
arXiv:2609.32804v1 Announce Type: cross Abstract: Audio emotion recognition (AER) typically assigns a single label to an entire recording, leaving the temporal scope of that label ambiguous when multiple speakers and aff…
- arXiv cs.AIresearchMind the Spike: Mechanisms and Brittleness of Visual Massive Activations in Large Vision-Language Models
arXiv:2609.32808v1 Announce Type: cross Abstract: Large vision-language models (LVLMs) inherit massive activations from their text-only bases: spikes where a few fixed hidden channels receive values thousands of times ab…
- arXiv cs.AIresearchRadiomap Blind Prediction under Incomplete Observation: Error Characterization and Correctable Propagation-Prior Learning
arXiv:2609.32836v1 Announce Type: cross Abstract: Radiomap blind prediction aims to infer radiomaps from observable representations of the propagation environment and base station configuration without field measurements…
- arXiv cs.AIresearchScanning While Imagining: A Scene-Graph World Model for Robotic Ultrasound Navigation
arXiv:2609.32837v1 Announce Type: cross Abstract: Ultrasound (US) acquisition depends on the operator's ability to interpret anatomy and anticipate how the view will change with probe motion. Many robotic US navigation m…
- arXiv cs.AIresearchTransfer Learning for Edge Classification on Dynamic Text-Attributed Graphs
arXiv:2609.32849v1 Announce Type: cross Abstract: Learning transferable representations for dynamic text-attributed graphs (DyTAGs) requires models to capture underlying interaction dynamics that persist across domains.…
- arXiv cs.AIresearchLogic Gate Networks and Lookup Table Networks as Lightweight Hardware Classifiers for Inter-patient ECG Arrhythmia Classification
arXiv:2609.32854v1 Announce Type: cross Abstract: Deep Differentiable Logic Gate Networks (LGNs) and Lookup Table Networks (LUTNs) offer a promising approach for very low power inference due to their use of simple binary…
- arXiv cs.AIagentsPolyTopoBench: A Benchmark for Complex Vector Polygon Generation from Remote Sensing Imagery
arXiv:2609.32856v1 Announce Type: cross Abstract: Vector polygon generation converts visual inputs, e.g., remote sensing (RS) images, into vectorized polygonal geometries, supporting applications such as autonomous drivi…
- arXiv cs.AIagentsRoboFoundry: System-as-Policy Evolution for Self-Learning Embodied Agents
arXiv:2609.32862v1 Announce Type: cross Abstract: A foundation model should not act in isolation as an embodied agent. Yet, existing methods often optimize individual components of the agent stack, such as memory, contex…
- arXiv cs.AIagentsASCEND: Personal AI Agents for Autonomous Scientific Computing Across HPC Clusters and GPU Workstations
arXiv:2609.32868v4 Announce Type: replace-cross Abstract: Traditional scientific computing requires researchers to translate intent into environment configuration, resource requests, and executable jobs, then diagnose fa…
- arXiv cs.AIresearchMultimodal LLMs Outperform Pathology Foundation Models in Cross-Domain Histological Similarity
arXiv:2609.32876v2 Announce Type: replace-cross Abstract: State-of-the-art pathology foundation models, trained on millions of histology tiles, can fail to preserve tissue similarity when comparisons cross slide or insti…
- arXiv cs.AIresearchRefreshing Less, Selecting Better: Reusing Stale Gradient Features for Efficient Influence-Based Data Selection
arXiv:2609.32888v1 Announce Type: cross Abstract: Gradient-based data selection methods such as LESS score each candidate by the alignment between its gradient and a target validation gradient, and recomputing per-exampl…
- arXiv cs.AIresearchOptimizing H-Graph Hybridization for Diffusion-Guided RRT
arXiv:2609.32897v1 Announce Type: cross Abstract: Sampling-based motion planners guided by diffusion models produce high-quality trajectories in a single run, yet the stochastic diversity available at inference time is l…
- arXiv cs.AIresearchWhen Less Compute Is More: Adaptive Early Exit Improves Pretrained Outlier Detection
arXiv:2609.32898v1 Announce Type: cross Abstract: Pretrained tabular foundation models process every dataset at a fixed depth, with inference costs growing with dataset size. To address this, we present the first study o…
- arXiv cs.AIresearchProgression- vs Automata-based Anticipatory Monitoring of LTL over Finite Traces (Extended Version)
arXiv:2609.32912v1 Announce Type: cross Abstract: When safety-critical systems are developed from a known internal specification, their correctness can be established by model checking. In the frequent case where such a…
- arXiv cs.AIresearchAllspark: Weak to Strong Transfer via Alternating Chain of Thought
arXiv:2609.32913v1 Announce Type: cross Abstract: Recent progress in frontier models has renewed interest in large-scale reinforcement learning (RL), but the cost of generating large-model rollouts makes even testing RL…
- arXiv cs.AIagentsAgentTell: Behavioural Side-Channel Leakage in Browser-Use Agents
arXiv:2609.32915v1 Announce Type: cross Abstract: Browser-use agents often carry information in their context as they move between websites. While it may be necessary for task completion, it also creates a privacy risk,…
- arXiv cs.AIagentsVision-Language Agents for Active Perception in Optics Laboratories
arXiv:2609.32918v1 Announce Type: cross Abstract: Vision-language models (VLMs) are increasingly being used in scientific workflows, but their ability as agents to directly control laboratory experiments from visual feed…
- arXiv cs.AIresearchThe Geometry of Logic: Stratification Induces Semantic Structure and Robust Reasoning
arXiv:2609.32927v1 Announce Type: cross Abstract: Transformer-based language models perform well on symbolic tasks, yet it remains unclear whether they learn generalizable rules or rely on statistical shortcuts. Mechanis…
- arXiv cs.AIresearchTwinS-GCN: Spectral conjugate for Spectral Graph Convolutional Networks
arXiv:2609.32940v1 Announce Type: cross Abstract: Graph convolutional networks propagate information by repeated local aggregation through a graph shift operator; i.e., a $K$-layer network reaches $K$ hops neighborhood.…
- arXiv cs.AIresearchEfficient Message Passing for Partial Differential Equation Priors
arXiv:2609.32956v1 Announce Type: cross Abstract: Prior information for real-world physical quantities is most elegantly expressed via partial differential equations (PDEs). In this paper, we propose a novel way to solve…
- arXiv cs.AIresearchDynamicDx: Evaluating Evidence Acquisition in Video-Based Diagnosis
arXiv:2609.32957v1 Announce Type: cross Abstract: Diagnosing a patient from video requires more than recognizing the sign: a vision-language model must turn what it sees into hypotheses, questions and tests. DynamicDx ev…
- arXiv cs.AIagentsEnvironmental Impact of Generative and Agentic AI: An in-Depth Analysis and Green Solutions
arXiv:2609.32960v1 Announce Type: cross Abstract: The proliferation of generative and agentic artificial intelligence (AI) systems has introduced computational demands whose environmental consequences are substantial yet…
- arXiv cs.AIagentsBeyond Token Savings: A Systematic Study of Context Compression in LLM Agents
arXiv:2609.32961v1 Announce Type: cross Abstract: As LLM agents tackle longer tasks, they increasingly compress growing histories of reasoning, actions, and tool outputs. Compression can reduce token use, but it also cha…
- arXiv cs.AIresearchCAPEX: Efficiently Distilling Foundation Model Behavior into Deployable Robot Policies through Experience-Adaptive Reasoning
arXiv:2609.33007v1 Announce Type: cross Abstract: Robot learning has largely relied on human-teleoperated demonstrations to acquire effective learnable behaviors. However, human-operated data collection processes can be…
- arXiv cs.AIresearchTCMQA: A 38K-Question Traditional Chinese Medicine Benchmark with a Licensed-Practitioner Reference
arXiv:2609.33014v1 Announce Type: cross Abstract: Medical benchmarks for language models are built almost entirely on Western biomedicine. Traditional Chinese Medicine (TCM) is a separate system, with its own diagnostic…
- arXiv cs.AIresearchRelative Generalization Invariance of LLM Pretraining
arXiv:2609.33016v1 Announce Type: cross Abstract: Large Language Model (LLM) pretraining performance is jointly shaped by three components of the training triplet: the optimizer, model architecture, and training data str…
- arXiv cs.AIresearch\L{}ukasiewicz Neural Networks Extended: Residual Architectures and Crystallization Strategies for Interpretable Rule Extraction
arXiv:2609.33028v1 Announce Type: cross Abstract: A feed-forward neural network whose weights are integers and whose activation is the truncated identity implements, neuron by neuron, the connectives of \L{}ukasiewicz ma…
- arXiv cs.AIresearchImproving the Diversity of LLM Outputs without a Trade-off
arXiv:2609.33038v1 Announce Type: cross Abstract: We propose DAST (Diversifying Arithmetic Sampling with TokenTour), a method that increases the diversity of LLM outputs without any change to the marginal distribution an…
- arXiv cs.AIresearchMore than 83.69% of the zeros of the Riemann zeta function are distinct
arXiv:2609.33043v2 Announce Type: replace-cross Abstract: The lower asymptotic proportion of distinct nontrivial zeros of the Riemann zeta function, relative to the total number counted with multiplicity, is at least $0.…
- arXiv cs.AIresearchZero-Storage Procedural Neural Synthesis via Boundary Dynamics: Formal Verification in Lean 4 and Bare-Metal Gauntlet Validation
arXiv:2609.33066v1 Announce Type: cross Abstract: Contemporary neural inference architectures rely on dense floating-point weight matrices stored in high-bandwidth memory (VRAM), incurring severe memory-wall bottlenecks…
- arXiv cs.AIresearchAlgorithmic Harms Associated with Generative Model-Augmented Recommendation Systems
arXiv:2609.33073v1 Announce Type: cross Abstract: In this work, we consider algorithmic harms that may arise as generative models are incorporated into machine learning platforms. We argue that existing harm taxonomies a…