AI news archive.
10,661 articles filed under Research, newest first, from the outlets listed on the sources page.
- arXiv cs.AIresearchSEmoEdit: Probing and Harnessing the Editability of Pre-trained Speech Flows
arXiv:2609.34648v1 Announce Type: cross Abstract: Existing training-based speech emotion editing methods often require substantial task-specific training and can be unstable. This motivates us to investigate whether the…
- arXiv cs.AIresearchEvidence-Aligned Multimodal On-Policy Self-Distillation for Fine-Grained Visual Understanding
arXiv:2609.34672v1 Announce Type: cross Abstract: Fine-grained visual understanding requires models to recognize small details within complex images. Multimodal on-policy self-distillation (OPSD) addresses this challenge…
- arXiv cs.AIresearchLearning What to Recall: Adaptive Multi-Cue Episodic Memory for World Models
arXiv:2609.34677v1 Announce Type: cross Abstract: World models predict future observations from current experience and actions, yet prediction can depend on observations seen far in the past. Episodic memory preserves pa…
- arXiv cs.AIresearchFair Fact-Checking: Closing the Cross-Lingual Gap in LLM Factual Judgement with RoSh
arXiv:2609.34678v1 Announce Type: cross Abstract: Misinformation on social media remains a critical problem, and more and more people settle it by asking a language model instead of a fact checker. Whether models judge s…
- arXiv cs.AIresearchUsing LLMs to Detect LLM-Generated Texts: A Cross-Generation Analysis
arXiv:2609.34691v1 Announce Type: cross Abstract: Automated detection of LLM-generated texts (LGTs) is critical, yet dedicated detectors often struggle to generalize across domains and models. While general-purpose LLMs…
- arXiv cs.AIresearchSufficiency of Zeroth-Order Reward Shaping for Policy Gradient in Stabilization Control
arXiv:2609.34695v1 Announce Type: cross Abstract: Reward shaping is fundamental to modern robotic control with deep reinforcement learning (RL), yet practitioners still rely heavily on heuristic principles borrowed from…
- arXiv cs.AIresearchTriangular Resampling for Long-Horizon Motion Generation
arXiv:2609.34697v2 Announce Type: replace-cross Abstract: We introduce Triangular Resampling (TR), a post-training method for mitigating long-horizon error accumulation in motion diffusion models. Built on FloodDiffusion…
- arXiv cs.AIresearchDynamic Flow, Static Graph: KV Cache Reuse for Efficient LLM Serving on Mobile NPUs
arXiv:2609.34727v1 Announce Type: cross Abstract: On-device large language model (LLM) serving is a cornerstone of local-first personal intelligence, offering users data sovereignty, strong privacy guarantees, and freedo…
- arXiv cs.AIresearchBeyond Token Alignment: Event Completion for Cross-Tokenizer On-Policy Distillation
arXiv:2609.34738v1 Announce Type: cross Abstract: On-policy distillation (OPD) transfers knowledge between language models through teacher supervision on student-generated trajectories. With different tokenizers, a singl…
- arXiv cs.AIresearchPredictive Dual Smoothing for Column Generation
arXiv:2609.34740v1 Announce Type: cross Abstract: Solving large-scale linear programs efficiently is an important challenge in many optimization settings. A key technique is column generation, which alternates between so…
- arXiv cs.AIresearchNo Pain, More Gain: Iterative Merging for Effective Multi-Teacher On-Policy Distillation
arXiv:2609.34745v1 Announce Type: cross Abstract: Multi-teacher on-policy distillation (MOPD) combines independently developed domain teachers into a single student by distilling their predictions on student-generated sa…
- arXiv cs.AIresearchA Unifying Framework of Concept-based Explainable AI with Completeness Guarantees
arXiv:2609.34750v1 Announce Type: cross Abstract: Concept-based explanations describe neural network predictions through human-understandable properties of inputs called concepts. The field encompasses approaches that di…
- arXiv cs.AIresearchWeaveData: A Multimodal Data Analysis System with Self-Critiquing and Self-Evolving LLM Plans
arXiv:2609.34764v1 Announce Type: cross Abstract: Multimodal data analysis, which answers questions over relational tables, text, and images, has attracted growing attention in the data management community. Large langua…
- arXiv cs.AIresearchWhen VLMs Trust Context: Evaluating Scene Text Recognition under Misleading Context
arXiv:2609.34781v1 Announce Type: cross Abstract: Vision-language models (VLMs) can read text in natural scenes, but their predictions may be influenced by the surrounding context. When the printed text conflicts with wh…
- arXiv cs.AIresearchCoHuB: A Simulation Benchmark for Multi-Humanoid Collaboration
arXiv:2609.34782v1 Announce Type: cross Abstract: Many physical tasks in human environments require collaboration, from assisting a partner to jointly manipulating an object. Yet, existing humanoid benchmarks largely foc…
- arXiv cs.AIresearchESTHER: Egocentric Stereo Hand Estimation and Reconstruction in the Wild
arXiv:2609.34817v2 Announce Type: replace-cross Abstract: Human dexterity is guided by two eyes watching two hands: binocular vision supplies the metric 3D structure that fine-grained manipulation consumes. Egocentric st…
- arXiv cs.AIresearchGaussian Neural Networks
arXiv:2609.34825v1 Announce Type: cross Abstract: Gaussian neural networks (GaNNs) are proposed as a novel regularization mechanism for neural networks. From a Bayesian perspective standard regularization techniques can…
- arXiv cs.AIresearchEviSplat: Preserving Multi-View Evidence in 3D Gaussian Splatting for Open-Vocabulary Segmentation
arXiv:2609.34853v1 Announce Type: cross Abstract: Open-vocabulary 3D scene understanding enables object localization and segmentation from free-form text queries without a fixed category vocabulary. Many recent methods b…
- arXiv cs.AIresearchBeyond Verbalized Confidence: Calibrating Reasoners with Differentiable Readouts
arXiv:2609.34857v1 Announce Type: cross Abstract: Reinforcement learning with verifiable rewards (RLVR) trains reasoning models to produce correct answers, but does not ensure that their stated confidence is calibrated.…
- arXiv cs.AIresearchFrom Attention Sensitivity to Layer Role: Revisiting Mixed-Precision Quantization of Transformers
arXiv:2609.34866v1 Announce Type: cross Abstract: Most post-training quantization pipelines fit each weight matrix to its pretrained counterpart, one matrix at a time. Whether that proxy tracks what an attention block ac…
- arXiv cs.AIresearchReference-Tail Trust:Certified Probability Floors for Learned Updates Inside a Deployed Network
arXiv:2609.34904v1 Announce Type: cross Abstract: Graph neural networks (GNNs) need to exploit improved message passing without surrendering control over predictions already trusted in deployment. We introduce Reference-…
- arXiv cs.AIresearchSincDPNet: Interpretable Raw-Waveform Bathroom Activity Recognition for Assistive Living
arXiv:2609.34907v1 Announce Type: cross Abstract: Bathroom acoustic-event recognition can support ambient assisted living in settings where continuous video monitoring is undesirable. However, practical deployment requir…
- arXiv cs.AIresearchDon't Throw Away the Tail: Action Upcycling for Policy Acceleration
arXiv:2609.34911v2 Announce Type: replace-cross Abstract: Modern robot policies predict a chunk of future actions from a single observation, execute only a prefix, and discard the rest before replanning. Choosing the len…
- arXiv cs.AIresearchDrug-Target Interaction Prediction via Hierarchical Sequential Cross-Attention over Chemical and Protein Language Models
arXiv:2609.34921v1 Announce Type: cross Abstract: Predicting Drug-Target Interactions~(DTIs) is a central task in computational drug discovery, with direct applications in virtual screening, drug repurposing, and therape…
- arXiv cs.AIresearchJazzSAMBA: A Synchronous and Asynchronous Multi-take Band Audio Dataset of Jazz Standards for Live Music Models
arXiv:2609.34931v1 Announce Type: cross Abstract: Machine learning has made strong progress on music tasks, both as assistive tools and as creative partners. However, most systems train on multitrack corpora that emphasi…
- arXiv cs.AIresearchJevVibe: Efficient Classification-Guided Secure Code Generation
arXiv:2609.34963v1 Announce Type: cross Abstract: Large language models can generate functionally correct code that still contains security weaknesses, motivating repair pipelines that first diagnose a weakness type befo…
- arXiv cs.AIresearchCyclostationary Phase Conditioning for Medical Time Series Diffusion
arXiv:2609.34965v2 Announce Type: replace-cross Abstract: Many physiological time series, such as cardiac and brain recordings, exhibit cyclostationarity: their statistics vary periodically with an underlying cycle phase…
- arXiv cs.AIresearchSemantic Uncertainty Quantification Needs Factual Equivalence
arXiv:2609.34967v1 Announce Type: cross Abstract: Semantic uncertainty quantification for large language models rests on a common template: sample several answers, measure how much they agree, and treat disagreement as u…
- arXiv cs.AIresearchRoboFL: Federated Expert Assembly for World Action Models
arXiv:2609.34968v1 Announce Type: cross Abstract: Vision-language-action and world-action models are increasingly popular, yet remain bottlenecked by physical interaction data that is scarce, institutionally siloed, and…
- arXiv cs.AIresearchJust MLPs: Efficient Visual State Reconstruction for Multimodal Language Models
arXiv:2609.34972v1 Announce Type: cross Abstract: Long visual token sequences often account for a substantial fraction of the computational overhead in multimodal large language models~(MLLMs). Existing approaches reduce…
- arXiv cs.AIresearchSPIDER: Multi-Layer Semantic Token Pruning and Adaptive Sub-Layer Skipping in Multimodal Large Language Models
arXiv:2609.34977v1 Announce Type: cross Abstract: Multimodal Large Language Models face significant efficiency challenges that stem from two distinct yet coupled sources: data redundancy and computational redundancy. Whi…
- arXiv cs.AIresearchComposable Decoding on the Probability Simplex: Theory and Implementation
arXiv:2609.34992v1 Announce Type: cross Abstract: Decoding for large language models is typically treated as a collection of isolated sampling strategies, with limited theoretical understanding of the behaviours they ind…
- arXiv cs.AIresearchStill There, No Longer Seen: Exposing Compression-Induced Risk in Large Vision-Language Models
arXiv:2609.35002v1 Announce Type: cross Abstract: Visual token compression reduces the inference cost of Large Vision-Language Models (LVLMs). However, aggregate robustness measures do not reveal whether a particular adv…
- arXiv cs.AIresearchLearning to Act under Visual Interruptions with Vision-Language-Action Models
arXiv:2609.35003v1 Announce Type: cross Abstract: Vision-language-action (VLA) models have demonstrated strong capabilities in robotic manipulation, but they are typically developed and evaluated with all camera streams…
- arXiv cs.AIresearchAddressing Spatial Indistinguishability in Spatiotemporal Prediction via Optimal Transport-Guided Masking
arXiv:2609.35021v1 Announce Type: cross Abstract: Spatiotemporal prediction aims to learn discriminative representations from correlated temporal signals over spatial structures for accurate future inference. A central c…
- arXiv cs.AIresearchVEX-Bench: Benchmarking Verification Complexity of LLM-Generated Misinformation
arXiv:2609.35028v1 Announce Type: cross Abstract: Large language models (LLMs) have made misinformation inexpensive to produce but not to verify, creating a growing asymmetry in the information ecosystem. Under tight tim…
- arXiv cs.AIresearchBA-DPO: Bias-Adjusted Direct Preference Optimization for Language Model Alignment
arXiv:2609.35044v1 Announce Type: cross Abstract: Preference-based alignment methods such as Direct Preference Optimization (DPO) use pairwise preferences labeled by human annotators to fine-tune language models. However…
- arXiv cs.AIresearchEMPIRIC: Experiment-Driven Learning of Residual World Models for Robot Planning
arXiv:2609.35047v1 Announce Type: cross Abstract: A robot should be able to learn through experiments how unfamiliar objects behave and interact, then plan with that knowledge. It need not start from scratch: physics eng…
- arXiv cs.AIresearchOPIS: An Input-Grounded Benchmark for Multi-Object Memory in Video World Models
arXiv:2609.35052v2 Announce Type: replace-cross Abstract: Video world models must preserve the visual state of the world over time, but existing evaluation protocols often rely on generated histories, video reference, or…
- arXiv cs.AIresearchTempoKV: Timely Staging of LLM KV Caches for Memory-Semantic Flash
arXiv:2609.35065v1 Announce Type: cross Abstract: Reusable prefix key-value (KV) caches can outgrow GPU memory in large language model (LLM) serving. A memory-semantic flash hierarchy offers SSD-backed capacity with a li…
- arXiv cs.AIresearchEchoes of Deeds: Moral History Can Shape and Steer LLM Behavioral Choices
arXiv:2609.35070v1 Announce Type: cross Abstract: Evaluations of Large Language Models (LLMs) morality typically consider decisions in isolation, thus overlooking whether an individual's unrelated prior conduct influence…
- arXiv cs.AIresearchSpikeLite: Lightweight Spiking Neural Networks for Time-Series Forecasting
arXiv:2609.35097v1 Announce Type: cross Abstract: Spiking neural networks (SNNs) offer an energy-efficient paradigm for time-series forecasting through spike-driven computation. However, recent SNN forecasters often purs…
- arXiv cs.AIresearchA mechanistic study of language model introspection
arXiv:2609.35108v1 Announce Type: cross Abstract: Large language models (LLMs) can sometimes report perturbations to their internal activations---even when the input provides no evidence that an intervention occurred. Ho…
- arXiv cs.AIresearchCTP-FL: Common-Trajectory Gradient Prediction for Federated Learning
arXiv:2609.35130v1 Announce Type: cross Abstract: Communication-efficient federated optimization commonly spends several gradient evaluations between server updates. Existing local-update methods use this computation to…
- arXiv cs.AIresearchLearning to Re-Draft: A Variational Stackelberg Game for Discrete Diffusion
arXiv:2609.35166v1 Announce Type: cross Abstract: Discrete diffusion models offer the ability to re-draft, revisiting and correcting earlier tokens throughout generation. This capability depends on the forward corruption…
- arXiv cs.AIresearchQAM: Quadratic-Accurate Checkpoint Merging via Sequential Consistency
arXiv:2609.35168v1 Announce Type: cross Abstract: Saved checkpoints record states along a training trajectory, but generally do not determine the updates at states that would be visited under a different schedule. We stu…
- arXiv cs.AIresearchCarveMix-RC: Addressing Rare-Class Imbalance Through Lesion-Aware Synthetic Augmentation for Brain Metastasis Segmentation
arXiv:2609.35195v1 Announce Type: cross Abstract: Accurate segmentation of post-treatment brain metastases is essential for treatment planning, longitudinal disease monitoring, and quantitative assessment of therapeutic…
- arXiv cs.AIresearchGAC-PINN: Geometry-Adaptive and Constraint-Enhanced Physics-Informed Neural Networks
arXiv:2609.35196v1 Announce Type: cross Abstract: For systems with steep gradients, sharp interfaces, or severe spatio-temporal coupling, Physics-informed neural networks (PINNs) suffer from spectral bias, geometric infl…
- arXiv cs.AIresearchAlignment Games: A Framework for Conceptual Repair in Human-AI Collaboration
arXiv:2609.35197v1 Announce Type: cross Abstract: The meaning of a concept in use is shaped by the situation, task, goals, and prior knowledge. For example, a request to make a poster "visually appealing for a five-year-…
- arXiv cs.AIresearchReCAT: Remember, Count, and Time: Structured Recurrent Memory for Robot Manipulation
arXiv:2609.35200v1 Announce Type: cross Abstract: Memory-dependent manipulation requires robots to make decisions using information that is no longer available to their current sensors, such as recalling an earlier visua…