AI news archive.
10,659 articles filed under Research, newest first, from the outlets listed on the sources page.
- arXiv cs.AIresearchMOSAIC: Multi-Objective Search for Allocating Inference Cache Across LLM Decoder Layers
arXiv:2610.25431v1 Announce Type: cross Abstract: The Key-Value (KV) cache is the main memory bottleneck of long-context LLM inference. Token eviction methods such as SnapKV, H2O, AdaKV and L2Norm shrink it, but they giv…
- arXiv cs.AIresearchEnhanced Masked Diffusion Unlearning
arXiv:2610.25447v1 Announce Type: cross Abstract: Masked diffusion language models (MDLMs) generate text by iteratively unmasking tokens with bidirectional context. As these models reach real-world deployment, machine un…
- arXiv cs.AIresearchReliability-Preserving Compression for Deployable Medical Image Segmentation
arXiv:2610.25460v1 Announce Type: cross Abstract: Post-training compression of pretrained medical segmentation models commonly uses quantization, pruning, or distillation to reduce model size and computational cost. Exis…
- arXiv cs.AIresearchDiaLong: Benchmarking Reminiscence in Long-Context Dialogues
arXiv:2610.25463v2 Announce Type: replace-cross Abstract: How well can transformer-based Large Language Models (LLMs) retain and recall memories in long contexts of dialogue? LLMs are known to forget long-range factual i…
- arXiv cs.AIresearchMultimodal-Conditioned Latent Diffusion Models for Spatiotemporal Brain MRI Forecasting
arXiv:2610.25481v2 Announce Type: replace-cross Abstract: Spatiotemporal forecasting of 3D anatomical changes is a critical generative vision challenge for modeling Alzheimer's Disease (AD) progression. While brain MRI f…
- arXiv cs.AIresearchPRIMA: Pose Refined Inductive Motion Adapter for Zero Shot Social Bodily Behavior Recognition
arXiv:2610.25482v1 Announce Type: cross Abstract: Recognizing fine-grained bodily behaviors, such as self-grooming, postural shifts, and backchanneling, is a useful intermediate step for computational social signal analy…
- arXiv cs.AIresearchDynamics-Aligned Representation Learning under Induced Partial Observability
arXiv:2610.25511v1 Announce Type: cross Abstract: Model-based reinforcement learning (MBRL) has enabled sample-efficient learning of continuous control policies from high-dimensional observations through learned latent d…
- arXiv cs.AIresearchATEST: Adaptive Resource-Aware Test-Time Scaling for LLM Inference
arXiv:2610.25519v1 Announce Type: cross Abstract: Test-time scaling (TTS) improves large language model (LLM) reasoning by sampling multiple answers, but fixed sampling budgets ignore both query difficulty and the changi…
- arXiv cs.AIresearchModaliff: A Unified Token Diffusion Model for Any-to-Any Generation
arXiv:2610.25533v1 Announce Type: cross Abstract: Any-to-any generation seeks a single model that translates flexibly among text, image, and audio, yet existing approaches either require fully aligned multimodal tuples o…
- arXiv cs.AIresearchResidential Construction Image Dataset Using CSI Divisions for Deep Learning-based Object Detection
arXiv:2610.25535v1 Announce Type: cross Abstract: Accurate recognition of construction site elements is essential for the application of deep learning (DL) in the construction industry. Despite rapid advances in computer…
- arXiv cs.AIresearchSafety Offset Learning: When Benign Fine-Tuning Misaligns LLMs
arXiv:2610.25611v1 Announce Type: cross Abstract: Safety risks from fine-tuning are commonly characterized by the harmfulness of the training data:harmful fine-tuning is considered dangerous, while benign fine-tuning is…
- arXiv cs.AIresearchWeighted State Souping for Efficient Long-Context Retrieval
arXiv:2610.25621v1 Announce Type: cross Abstract: State space models (SSMs) combine fast inference and fixed state size with competitive language-modeling performance. Recently, state souping has been proposed as a proce…
- arXiv cs.AIresearchACorD: Adaptive Correspondence Discovery for Medical Image Registration
arXiv:2610.25638v1 Announce Type: cross Abstract: Medical image registration is a fundamental task for aligning anatomical structures across images. Exhaustively searching for point-wise correspondences is prohibitively…
- arXiv cs.AIresearchConsistency Training Suppresses Cue Following Without Teaching Cue Selectivity
arXiv:2610.25650v1 Announce Type: cross Abstract: Language models often change answers because of a cue, a hint such as "I believe the answer is B" that gives no valid evidence. Consistency training (CT) trains a model t…
- arXiv cs.AIresearchA "feature ODE" describing the learning behavior of shallow MLPs on simple functions
arXiv:2610.25652v1 Announce Type: cross Abstract: We study the gradient flow training dynamics of two-layer multi-layer perceptrons (MLPs) on anisotropic Gaussian data. Even in this simple setting, solving the dynamics i…
- arXiv cs.AIresearchBeyond Leaderboards: A Method for Application-Specific Capability Profiling of Large Language Models
arXiv:2610.25659v1 Announce Type: cross Abstract: Public leaderboards rank large language models (LLMs) in general, but an engineer has to choose a model for a specific application, architecture and cost. This paper pres…
- arXiv cs.AIresearchCross-Pollination Retrieval: Finding Transferable Ideas Across Fields in Scientific Literature
arXiv:2610.25678v1 Announce Type: cross Abstract: Many scientific advances come from cross-pollination, carrying an idea from one field over to a problem in another. Yet retrieving papers across fields remains underexplo…
- arXiv cs.AIresearchMT-CC: Multi-Group Temperature Scaling for Asymmetric Calibration Behavior in Class-Incremental Learning
arXiv:2610.25689v1 Announce Type: cross Abstract: As deep neural networks are required to continuously adapt to evolving data, class-incremental learning (CIL) has become an active research area. However, most existing w…
- arXiv cs.AIresearchProgress-Guided Preference Generation for Aligning Generative Control Policies
arXiv:2610.25692v1 Announce Type: cross Abstract: Generative policies have achieved strong performance in imitation learning tasks, but their success is often constrained by the limited availability of expert demonstrati…
- arXiv cs.AIresearchCausal Humility: Governing the Authority of Fairness Claims Under Causal Uncertainty
arXiv:2610.25709v1 Announce Type: cross Abstract: AI systems increasingly shape consequential decisions in employment, lending, healthcare, and public services, making fairness an important problem for auditing and gover…
- arXiv cs.AIresearchCorrected Block Sparse Spectral Optimization: Exact Models, Finite Identification, and Local Dynamics
arXiv:2610.25721v1 Announce Type: cross Abstract: The block generalized-power method pairs threshold-based support selection with inexpensive Stiefel updates, but the standard block formulation conflates two nonequivalen…
- arXiv cs.AIresearchAnatomy of Off-Policy Policy Gradient: Importance Sampling, KL Regularization, and Baselines
arXiv:2610.25725v1 Announce Type: cross Abstract: Reinforcement learning has become a central tool for large language model (LLM) post-training, where policy gradient methods are routinely deployed off-policy, even thoug…
- arXiv cs.AIresearchNo Matter Who You Are: Addressivity and Sycophancy in Large Language Models
arXiv:2610.25728v1 Announce Type: cross Abstract: Sycophancy, where an LLM becomes overly agreeable, deferential, or flattering toward the user, has been identified as a major failure mode. While post-training has been i…
- arXiv cs.AIresearchA Calibrated Instrument for Detecting Disposition Drift in Language Models
arXiv:2610.25735v1 Announce Type: cross Abstract: Before language models are released, they are tested for harmful traits such as willingness to carry out dangerous requests, deception, power seeking, and sycophancy. The…
- arXiv cs.AIresearchCognitively Decoupled Meta-Induction for Out-of-Distribution Generalization
arXiv:2610.25749v1 Announce Type: cross Abstract: In-context learning (ICL) tackles downstream tasks conditioned on demonstrations. This paradigm relies heavily on neighbor-based matching between queries and demonstratio…
- arXiv cs.AIresearchAirMPA: A Meteorology-to-Pollution Adapter for Global Air Quality Forecasting
arXiv:2610.25797v1 Announce Type: cross Abstract: Accurate real-time forecasting of atmospheric pollutants is essential for reducing exposure risks and improving public health. Traditional numerical forecasting systems s…
- arXiv cs.AIresearchOptimizing Marketing Budgets for Social Influence: A Generalized Fractional Incentives Framework
arXiv:2610.25823v1 Announce Type: cross Abstract: Marketing managers routinely allocate limited budgets across individuals in social networks whose responsiveness and influence vary substantially. We study the problem of…
- arXiv cs.AIresearchDiffusion Language Models Change Speed Across Tasks and Languages
arXiv:2610.25835v1 Announce Type: cross Abstract: Diffusion language models can write several tokens in one forward pass, and they decide during generation how many passes to spend. Two outputs of the same token length c…
- arXiv cs.AIresearchWhere RL Breaks: Localizing and Repairing Failures via Targeted Continued Pretraining
arXiv:2610.25836v1 Announce Type: cross Abstract: Reinforcement learning with verifiable rewards (RLVR) is widely used to improve reasoning in language models, yet growing evidence suggests that it mostly sharpens what t…
- arXiv cs.AIresearchOV-PEAR: Observation-Verified Pose-Scale Estimation and Reconstruction for Unknown Industrial Manipulation Objects
arXiv:2610.25840v1 Announce Type: cross Abstract: Fine industrial robotic manipulation requires 6D pose, scale, and an operation-ready 3D model. Existing industrial pose estimation relies on CAD models, object-specific t…
- arXiv cs.AIresearchSWIFT: Self-guided Weak-to-strong dIFfusion models via Target extrapolation
arXiv:2610.25844v2 Announce Type: replace-cross Abstract: Training high-quality diffusion models is computationally expensive, and inference-time guidance often adds further sampling cost. In this work, we introduce Swif…
- arXiv cs.AIresearchMeasurement Validity and Representation Dynamics of Confidence Signals in Large Language Models
arXiv:2610.25846v1 Announce Type: cross Abstract: Whether a language model "knows what it knows" matters for trustworthy deployment and for deciding when it should abstain. Prior work usually measures this through one ch…
- arXiv cs.AIresearchRubrics Beyond Evaluation: Exploiting Rubric Structure in Reinforcement Learning
arXiv:2610.25850v1 Announce Type: cross Abstract: Rubrics are widely used to evaluate open-ended question answering by breaking overall quality into explicit criteria. These explicit criteria make rubrics a rich source o…
- arXiv cs.AIresearchAdaptive Decoding Methods for Token-Efficient Reinforcement Learning
arXiv:2610.25860v1 Announce Type: cross Abstract: Rollout decoding affects both which trajectories are used for reinforcement learning with verifiable rewards (RLVR) and how many completion tokens are spent collecting th…
- arXiv cs.AIresearchManifold-Constrained Residual Adaptation with Frame-Adaptive Layer Aggregation for Speaker Verification
arXiv:2610.25892v1 Announce Type: cross Abstract: Large self-supervised speech models provide powerful representations for speaker verification, but full fine-tuning is expensive. Most conventional adapters insert small…
- arXiv cs.AIresearchUtilityBench: Measuring Language Model Utility When Errors Have Different Costs
arXiv:2610.25924v1 Announce Type: cross Abstract: As language models are increasingly deployed in real-world scenarios with important outcomes, the impact of their mistakes often extends far beyond what correctness can c…
- arXiv cs.AIresearchKeep the Lineages Alive: Diversity-Preserving SMC for Diffusion LLM Decoding
arXiv:2610.25932v1 Announce Type: cross Abstract: Inference-time sequential Monte Carlo (SMC) decoding can improve the reasoning accuracy of masked diffusion language models (diffusion LLMs) without post-training by resa…
- arXiv cs.AIresearchThe KV-Surgeon: Compensation-Aware Offline KV-Cache Compaction
arXiv:2610.25933v1 Announce Type: cross Abstract: Offline compaction of the key--value cache (KV-cache) reduces the cost of repeatedly querying a fixed context by constructing a smaller cache once and reusing it subseque…
- arXiv cs.AIresearchPaceVLA: Adaptive Temporal Action Quantization in Vision-Language-Action Models
arXiv:2610.25961v2 Announce Type: replace-cross Abstract: Fast and reliable manipulation is crucial for practical robot deployment. Although Vision-Language-Action (VLA) models have recently shown remarkable progress as…
- arXiv cs.AIresearchBehavioral Probes for Information Flow in LLM Swarms
arXiv:2610.25966v1 Announce Type: cross Abstract: Most existing approaches to LLM swarm optimization are utility-driven, focusing on final task utility while leaving the processing, integration, and reliance of informati…
- arXiv cs.AIresearchReconstruction and Discrete-Event Simulation Assessment of an Analytical
arXiv:2610.25967v1 Announce Type: cross Abstract: This article presents an analytical-DES benchmark for intra-aisle robotic picking in automated warehousing. The investigated cycle combines horizontal shuttle travel, tot…
- arXiv cs.AIresearchCorrectness-Dependent Asymmetric Target Distributions for Self-Distillation Policy Optimization
arXiv:2610.25973v1 Announce Type: cross Abstract: Self-distillation policy optimization provides dense token-level supervision for LLM post-training through feedback-conditioned self-teachers. However, directly matching…
- arXiv cs.AIresearchAuditor-Assisted Summary-Channel Verification for Hosted LLM Identity Substitution
arXiv:2610.25982v1 Announce Type: cross Abstract: Hosted LLM services increasingly expose model identity as a product claim, but external users and auditors often cannot inspect which model configuration served a request…
- arXiv cs.AIresearchAccelerated Drift of Scientific Prose in the LLM Era: Evidence from an Interrupted Time-Series Analysis
arXiv:2610.26008v1 Announce Type: cross Abstract: Every scientific text is embedded in the writing norms of its time: it reflects earlier publications while also shaping the texts that follow. The growing use of LLM-base…
- arXiv cs.AIresearchFrameGRPO: Incentivizing Video Reasoning via Multi-Frame Rollouts and Temporal Credit
arXiv:2610.26022v1 Announce Type: cross Abstract: Reinforcement learning with verifiable rewards, particularly Group Relative Policy Optimization (GRPO), has improved video reasoning in multimodal large language models.…
- arXiv cs.AIresearchMultimodal Decoupling for Diversified Recommendation through Semantic and Structural Shaping
arXiv:2610.26024v1 Announce Type: cross Abstract: Recommender systems often suffer from over-concentrated results when optimized for accuracy, limiting recommendation diversity. While diversity has become an important ob…
- arXiv cs.AIresearchTamedFlow: Trajectory-Aware Distillation for Teacher-Assisted Diffusion Sampling
arXiv:2610.26025v2 Announce Type: replace-cross Abstract: Diffusion Transformers (DiTs) deliver high-fidelity image and video generation, but inference requires many sequential model evaluations. Two acceleration strateg…
- arXiv cs.AIresearchHoneyed Words, Hidden Daggers: Benchmarking and Training LLMs to See Through Mock Politeness
arXiv:2610.26059v1 Announce Type: cross Abstract: The Chinese idiom "kou mi fu jian" ("honeyed words, hidden daggers") names speech that is polite in form but hostile in function. We present ZhenhuanBench, a 1,592-item b…
- arXiv cs.AIresearchNot All Similarity Is Leakage: Distribution-Aware Identity Leakage Assessment in Synthetic Face Datasets
arXiv:2610.26077v1 Announce Type: cross Abstract: Synthetic data provides the means for more ethical and less biased facial recognition. Nonetheless, the limitations of generative models and the conditioning process they…
- arXiv cs.AIresearchHow Robust Are Prognostic Radiomic Features? Comparing Selection Techniques on Pooled Glioblastoma Cohorts
arXiv:2610.26095v1 Announce Type: cross Abstract: Radiomic survival studies in glioblastoma report heterogeneous prognostic features, but how much of this stems from the methodology itself is rarely examined. We analysed…