AI news archive.
10,661 articles filed under Research, newest first, from the outlets listed on the sources page.
- arXiv cs.AIresearchTsubame: Tree Replay for Diffusion-Based Speculative Decoding
arXiv:2609.33652v1 Announce Type: cross Abstract: Context-aware dynamic trees allocate the speculative decoding budget according to draft path probabilities, adapting their depth and branching to the current context. Und…
- arXiv cs.AIresearchDemonstration-Free Success-Probability Reward Learning for Generalist Robot Policies
arXiv:2609.33653v1 Announce Type: cross Abstract: Reinforcement learning (RL) enables generalist robot policies to improve through trial-and-error interaction, yet its effectiveness is fundamentally constrained by sparse…
- arXiv cs.AIresearchWhen Noise Meets Long-Tail: Feature-Threshold Dual Calibration for Robust Pseudo-Labeling
arXiv:2609.33668v1 Announce Type: cross Abstract: Pseudo-labeling has become a cornerstone of learning from unlabeled data in semantic segmentation. Yet its effectiveness drops sharply in real-world scenarios where stron…
- arXiv cs.AIresearchDynamic Kuramoto-Hodge Operators for PDEs on Complex Geometries and Topologies
arXiv:2609.33693v1 Announce Type: cross Abstract: Learning PDE operators on complex domains requires capturing interactions among fields on vertices, edges, and faces, alongside global responses shaped by topology. Exist…
- arXiv cs.AIresearchUnderstanding Confabulation and Rethinking Reconstruction in Activation Explanations
arXiv:2609.33702v1 Announce Type: cross Abstract: Natural Language Autoencoders (NLAs) produce unsupervised text explanations of a model's activations: a verbalizer describes an activation and a reconstructor learns to r…
- arXiv cs.AIresearchFrom Granular Revision Operations to Meaningful Revision Units: Evaluating LLMs for Revision Boundary Detection
arXiv:2609.33720v1 Announce Type: cross Abstract: Revision traces provide valuable evidence about students' writing processes, but their usefulness for learning analytics depends on how individual revisions are represent…
- arXiv cs.AIresearchBOReFT: Manifold Steering of Language Models for Black-box Optimization
arXiv:2609.33722v1 Announce Type: cross Abstract: Language models are increasingly used as proposal models for black-box search, from program optimization to molecular design. Existing approaches typically improve propos…
- arXiv cs.AIresearchCollaborative Synthetic Data for Privacy-Preserving Financial Fraud Detection Across Organizational Silos
arXiv:2609.33754v2 Announce Type: replace-cross Abstract: Organizations seek analytical value from AI, yet relevant data are often fragmented across organizations and constrained by privacy. This is acute in financial fr…
- arXiv cs.AIresearchTransformer-based Neural Beamforming for Real-Time Speech Enhancement on Smart Low-Power Hearable Devices
arXiv:2609.33755v1 Announce Type: cross Abstract: Accurate, efficient, and low-latency spatial beamforming is a key component in emerging smart hearable devices, enhancing speech while suppressing noise and interference.…
- arXiv cs.AIresearchBeyond Fixed Features: Architecture-Dependent Sensitivity to Node Representations under Heterophily
arXiv:2609.33764v1 Announce Type: cross Abstract: Graph Neural Networks (GNNs) perform well on homophilic graphs but struggle in heterophilic settings, where connected nodes often carry dissimilar labels. Existing evalua…
- arXiv cs.AIresearchSelecting Diverse SFT Traces Improves Post-RL Generalization
arXiv:2609.33780v1 Announce Type: cross Abstract: Verified solutions are not equally useful for preparing reasoning models for reinforcement learning (RL). We present a comprehensive study of route diversity, the variati…
- arXiv cs.AIresearchSurprising Success, Repeated Failure: Entropy-Guided Credit Assignment for Exploration in LLM Reasoning
arXiv:2609.33781v1 Announce Type: cross Abstract: Reinforcement learning with verifiable rewards (RLVR) enhances reasoning in large language models (LLMs) through outcome-level feedback, yet recent approaches to finer-gr…
- arXiv cs.AIresearchDo We Really Need KL Divergence for On-Policy Distillation of Large Language Models?
arXiv:2609.33791v1 Announce Type: cross Abstract: Since the advent of knowledge distillation, KL divergence has been the standard loss in distillation. Recently, on-policy distillation (OPD) has emerged as an efficient p…
- arXiv cs.AIresearchDiffusion Reward Models
arXiv:2609.33803v2 Announce Type: replace-cross Abstract: Reward models underpin the alignment of large language models, yet the dominant designs reduce each prompt--response pair to a point estimate or to a distribution…
- arXiv cs.AIresearchMinkowskiPE: Minkowski Positional Encoding for Spatiotemporal Perception
arXiv:2609.33804v1 Announce Type: cross Abstract: Modeling spatiotemporal coupling is a key challenge in building physical intelligence across scales, from microscopic to macroscopic. Existing models capture such structu…
- arXiv cs.AIresearchAugmenting Visual Anomaly Detection with Automated Interpretability
arXiv:2609.33818v1 Announce Type: cross Abstract: Visual anomaly detectors identify deviations from known-normal data, but their anomaly signals may mix evidence of actual anomalies with benign visual variation. We inves…
- arXiv cs.AIresearchAchieve What You Imagined: Learning to Align Actions with Visual Plans
arXiv:2609.33832v1 Announce Type: cross Abstract: World-action models can jointly predict future visual observations and robot actions. However, discrepancies may exist between their visual predictions and the consequenc…
- arXiv cs.AIresearchAn Active-Bottleneck Mechanism for Weak-to-Strong Generalization
arXiv:2609.33835v1 Announce Type: cross Abstract: Weak-to-strong generalization (W2SG) occurs when a student trained on a teacher's predictions outperforms that teacher. We study when this happens under fully converged,…
- arXiv cs.AIresearchProgram-Verified Self-Evolution for Vision-Language Models
arXiv:2609.33855v1 Announce Type: cross Abstract: Self-evolving vision-language models train on questions they generate from unlabeled images. Since these questions have no gold answers, prior methods label them by major…
- arXiv cs.AIresearchPI-NOMT: Physics-Informed Neural Optimal Mass Transport for Brain Fluid Dynamics
arXiv:2609.33857v1 Announce Type: cross Abstract: Recovering hidden transport mechanisms from sparse spatiotemporal observations is a fundamental inverse problem in scientific machine learning. In brain tracer imaging, d…
- arXiv cs.AIresearchRobot-GST: geometry-aware spatial-temporal robot policy representation and evaluation
arXiv:2609.33872v1 Announce Type: cross Abstract: Robotic manipulation policies are advancing rapidly with increasing reliance on vision-language models for end-to-end decision making. However, reliable deployment remain…
- arXiv cs.AIresearchFinite Probes Suffice: Identifiability and Universality for Weight-Space Learning
arXiv:2609.33901v2 Announce Type: replace-cross Abstract: Learning properties of neural networks has recently attracted growing interest, with existing approaches operating either directly on network parameters or throug…
- arXiv cs.AIresearchJIVE: Jacobian-Informed Volume Expansion for Diverse Generative Sampling
arXiv:2609.33906v1 Announce Type: cross Abstract: Generative models often suffer from mode collapse and limited sample diversity. While prior works attempt to mitigate this by jointly generating a batch of samples and re…
- arXiv cs.AIresearchValidating Memory-Optimal Transformer Kernels on Real Hardware: From Formal Derivation to Measured Performance Across Two HPC Clusters
arXiv:2609.33916v1 Announce Type: cross Abstract: We validate memory-optimal cost functions for transformer kernels derived via the Mathematics of Arrays (MoA). Companion Papers I-IV formally derive kernels for attention…
- arXiv cs.AIresearchQuantization Error Is Spectrally Flat: A Single Random Probe Is a Calibrated, Data-Free Sensitivity Estimator, with Application to Budget-Targeted Mixed-Precision Quantization
arXiv:2609.33923v2 Announce Type: replace-cross Abstract: A single random Gaussian probe gives an unbiased estimate of the squared Frobenius norm of a layer's quantization error. The estimator is well-behaved because rou…
- arXiv cs.AIresearchDiffusion-Based Rollouts as a Stabilization Mechanism for Long-Horizon Environmental Forecasting
arXiv:2609.33930v1 Announce Type: cross Abstract: Extending forecast lead times while maintaining predictive skill remains a major challenge in environmental forecasting. We investigate diffusion-based rollouts as a stab…
- arXiv cs.AIresearchUnifying Video Tasks via Spatiotemporal Analogy
arXiv:2609.33935v2 Announce Type: replace-cross Abstract: Adapting video models to new tasks typically requires dedicated data curation and fine-tuning. While visual analogy provides a training-free alternative by specif…
- arXiv cs.AIresearchEEG-Fusion: Failure-Informed Source-Free Expert Routing for Robust Motor Imagery EEG Decoding
arXiv:2609.33962v1 Announce Type: cross Abstract: Subject-independent motor-imagery (MI) EEG decoding can exhibit subject-level failures even when average performance appears acceptable: under subject shift, a decoder ca…
- arXiv cs.AIresearchThinkNet: Evolutionary Deep Ensembles for Efficient Motor-Imagery EEG Classification
arXiv:2609.33967v2 Announce Type: replace-cross Abstract: Practical assistive and rehabilitative brain--computer interfaces require subject-independent motor-imagery EEG (MI-EEG) decoders that generalize to new users und…
- arXiv cs.AIresearchOn the Token Value Inequality in Efficient Reasoning
arXiv:2609.33970v1 Announce Type: cross Abstract: Chain-of-Thought reasoning has enabled large language models to achieve substantial performance gains on complex tasks. However, these gains come at the cost of dramatica…
- arXiv cs.AIresearchHARMONIA: Interpretable Graph Learning through Mixtures of Neural Bases
arXiv:2609.33972v1 Announce Type: cross Abstract: Existing interpretable graph additive models still face limitations in either computational scalability or modeling flexibility. In terms of structural modeling, previous…
- arXiv cs.AIresearchGroupMask: Layer-Adaptive Group-wise Sparsity for Semi-Structured LLM Pruning
arXiv:2609.33977v1 Announce Type: cross Abstract: Semi-structured pruning compresses large language models (LLMs) while keeping a regular sparse structure, but the prevailing N:M pattern fixes the same local sparsity rat…
- arXiv cs.AIresearchA packet-level digital hardware twin for commissioning megahertz diagnostic edge AI and plasma control system integration in tokamaks
arXiv:2609.33994v1 Announce Type: cross Abstract: High-bandwidth plasma diagnostics increasingly provide inputs to machine learning and signal-processing algorithms intended for real-time tokamak control, but the complet…
- arXiv cs.AIresearchWhen Known Physics Helps Neural PDE Models: Residual Constraints Out-Regularize Generic Priors for Nonlinear Dynamics
arXiv:2609.34012v1 Announce Type: cross Abstract: Neural PDE surrogates increasingly incorporate structural priors, yet it is often unclear whether their gains arise from physics-specific information or simply from regul…
- arXiv cs.AIresearchSR4-Fit: A Unified Interpretable Rule-Based Machine Learning Framework for Informative and Trustworthy Decision-Making
arXiv:2609.34019v1 Announce Type: cross Abstract: In many high-stakes applications, machine learning is dominated by black-box models that require post hoc explanations to justify their predictions. These explanations ar…
- arXiv cs.AIresearchUncovering shortcut learning in audio classifiers by discovering recurring concepts in temporal explanations
arXiv:2609.34030v1 Announce Type: cross Abstract: Correlations between events in machine learning datasets may result in shortcut learning, where models learn to predict the target event based on the presence of a correl…
- arXiv cs.AIresearchADPTNet: Adaptive with Prescriptive Timescales Non-Linear SSM for Sequence Modelling
arXiv:2609.34034v1 Announce Type: cross Abstract: A central aim of neuromorphic computing is to provide a viable alternative to highly energy-intensive Transformer-based AI. However, efficient alternatives struggle to ca…
- arXiv cs.AIresearchARCH-B: Architectural Representation, Comprehension and Hierarchy Benchmark
arXiv:2609.34047v1 Announce Type: cross Abstract: Multimodal models increasingly interpret visual environments, but their ability to recognize the same building across photographs, floor plans, elevations, sections, and…
- arXiv cs.AIresearchDo World Models Learn Global Understanding?
arXiv:2609.34058v1 Announce Type: cross Abstract: AI systems often feel brittle and fragmented. A large language model (LLM) may correctly explain a concept but fail to apply it, or follow safety instructions in one cont…
- arXiv cs.AIresearchWho Gets a Token, and What Does It Carry? Unequal Name Support and Concept Access in Large Language Models
arXiv:2609.34065v1 Announce Type: cross Abstract: Names are personal identifiers, but they also carry social meaning and are widely used to evaluate how language models treat different people. Such evaluations typically…
- arXiv cs.AIresearchFLARE: Flow Matching with Local Axis-Angle Representations for Stochastic Micromagnetic Evolution
arXiv:2609.34070v1 Announce Type: cross Abstract: Long-horizon micromagnetic simulation remains expensive because conventional and learned solvers typically propagate Landau--Lifshitz--Gilbert (LLG) dynamics step by step…
- arXiv cs.AIresearchMaskCoFT: Masked Co-Adaptive Fine-Tuning for Memory-Efficient MoE Inference
arXiv:2609.34077v2 Announce Type: replace-cross Abstract: Mixture-of-experts (MoE) language models often exceed the memory of a single GPU. Expert offloading keeps most experts in host memory and loads them on demand, so…
- arXiv cs.AIresearchWhiteCon: Semi-Supervised Domain Adaptation Regression Through Whitening Transform and Dual Consistency
arXiv:2609.34078v1 Announce Type: cross Abstract: Domain adaptation is crucial for addressing distributional shifts that degrade model performance across domains. While most existing research has centered on classificati…
- arXiv cs.AIresearchTRACE: Expert-Aligned ECG Representation Learning with Rigorous Benchmarking and Real-World Validation in Acute Cardiac Care
arXiv:2609.34088v1 Announce Type: cross Abstract: TRACE (Text-Reinforced Analysis of Cardio ECGs) is a multimodal electrocardiogram (ECG) representation model that learns clinically grounded signal embeddings for downstr…
- arXiv cs.AIresearchThe Devil is in the Spectrum Bias: Spectrum-Balanced Feature Matching for Robust Representation Distillation
arXiv:2609.34106v1 Announce Type: cross Abstract: Large visual foundation models have demonstrated remarkable transferability across a wide range of downstream tasks. To deploy such models efficiently, feature matching h…
- arXiv cs.AIresearchUnknown is not normal: separating language-model extraction from rule-based decision logic for clinical risk scores
arXiv:2609.34112v1 Announce Type: cross Abstract: Large language models (LLMs) are increasingly used to compute clinical risk scores from free-text notes. Notes are often incomplete, and treating undocumented findings as…
- arXiv cs.AIresearchSlimWise: Decoupling Expert Pruning Across Prefill and Decode for Efficient MoE Serving
arXiv:2609.34117v1 Announce Type: cross Abstract: Mixture-of-experts (MoE) models activate few experts per token, yet batched decoding can access nearly the entire expert pool, making expert-weight traffic a major bottle…
- arXiv cs.AIresearchProbabilistic electrical power demand forecasting with uncertainty quantification
arXiv:2609.34120v1 Announce Type: cross Abstract: The majority of research on electricity consumption forecasting has focused on deterministic approaches, which generate a single point estimate for each time step in the…
- arXiv cs.AIresearchSTITCH-RAG: Spatio-Temporal Influence Tracing over Topic Hypergraphs for Multi-Hop Retrieval-Augmented Generation
arXiv:2609.34127v1 Announce Type: cross Abstract: Multi-hop retrieval-augmented generation requires a retriever to connect evidence distributed across documents while preserving a concise, faithful generation context. Ex…
- arXiv cs.AIresearchPrefLUT: Reusable and Refinable Personalized Color Editing from Pairwise Preferences
arXiv:2609.34133v1 Announce Type: cross Abstract: Photographic color editing is inherently personal: the same image can appear too warm, too muted, or already satisfactory to different users. Most lookup table (LUT) and…