Search: “arxiv-cs-ai”

What Makes a Strong Model? A Unified Spectral Analysis of Knowledge Transfer over High-dimensional Linear Regression

arXiv:2606.01292v1 Announce Type: cross Abstract: Teacher-Student Knowledge Transfer (KT) is ubiquitous in modern machine learning, ranging from classical model compression via Knowledge Distillation

Model Releases

When AI Benchmarks Plateau: A Systematic Study of Benchmark Saturation

DGX agent

arXiv:2602.16763v2 Announce Type: replace Abstract: Artificial intelligence benchmarks are an important mechanism for measuring model progress and guiding deployment decisions. However, benchmarks qui

model-releasesarxiv-cs-ai

2 Jun 2026

When and How Much to Imagine: Adaptive Test-Time Scaling with World Models for Visual Spatial Reasoning

arXiv:2602.08236v2 Announce Type: replace-cross Abstract: Despite rapid progress in MLLMs, visual spatial reasoning remains unreliable when correct answers depend on how a scene would appear under uns

When Data Is Scarce: Scaling Sparse Language Models with Repeated Training

arXiv:2606.01155v1 Announce Type: cross Abstract: Scaling laws for dense LLMs under infinite data are well explored, but how sparsity interacts with limited data is not. In this work, we study sparse

When Do Attention Circuits Form? Developmental Trajectories of Capability and Attention-Sink Emergence Across Three 1B-ClassArchitectures

arXiv:2606.02378v1 Announce Type: cross Abstract: We track the developmental trajectory of attention-head circuit formation across three 1B-class language models spanning two architecture families (de

When Does Predictive Inverse Dynamics Outperform Behavior Cloning?

arXiv:2601.21718v2 Announce Type: replace-cross Abstract: Behavior cloning (BC) is a practical offline imitation learning method, but it often fails when expert demonstrations are limited. Recent work

When Jokes Cross the Line: Analyzing Regular Humor and Dark Humor in YouTube Shorts

arXiv:2606.00046v1 Announce Type: cross Abstract: Video platforms such as YouTube have reshaped how users engage with entertainment and information, emphasizing brief, highly engaging content such as

When Safe Skills Collide: Measuring Compositional Risk in Agent Skill Ecosystems

arXiv:2606.00448v1 Announce Type: cross Abstract: LLM agents increasingly rely on community-contributed skills that expand an agent's operational capability set. We study a core safety problem in agen

When Single Answer Is Not Enough: Rethinking Single-Step Retrosynthesis Benchmarks for LLMs

arXiv:2602.03554v2 Announce Type: replace-cross Abstract: Recent progress has expanded the use of large language models (LLMs) in drug discovery, including synthesis planning. However, objective evalu

When Softmax Fails at the Top: Extreme Value Corrections for InfoNCE

arXiv:2606.00262v1 Announce Type: cross Abstract: InfoNCE is the standard contrastive learning objective, but its softmax form is not only a computational convenience: it also encodes a statistical as

Where Do Deep-Research Agents Go Wrong? Span-Level Error Localization in Agent Trajectories

arXiv:2606.02060v1 Announce Type: new Abstract: Deep-research agents solve tasks through long trajectories of search, tool use, evidence inspection, and answer synthesis. Evaluation based on final ans

Who Annotates in NLP? A Large-scale Assessment of Human Annotation Reporting between 2018 and 2025

arXiv:2606.02255v1 Announce Type: cross Abstract: Human annotation is the empirical foundation of much NLP research, from dataset construction to model evaluation, but papers often leave unclear who p

Who Evaluates AI's Social Impacts? Mapping Coverage and Gaps in First and Third Party Evaluations

arXiv:2511.05613v2 Announce Type: replace-cross Abstract: Foundation models are increasingly central to high-stakes AI systems, and governance frameworks now depend on evaluations to assess their risk

Applications

Why Do Time Series Models Need Long Context Windows?

arXiv:2606.01999v1 Announce Type: cross Abstract: Modern deep learning models for forecasting groups of time series rely on increasingly longer observation windows. However, the benefit of increasing

applicationsarxiv-cs-ai

Why Not Hyperparameter-Friendly Optimisation? A Monotonic Adaptive Norm Rescaling Approach For Long-Tailed Recognition

arXiv:2606.02526v1 Announce Type: cross Abstract: Long-tailed recognition poses a significant challenge for deep learning. The two-stage decoupling paradigm, which separates representation learning fr

WorldCoder-Bench: Benchmarking Physically Grounded 3D World Synthesis

arXiv:2606.01869v1 Announce Type: new Abstract: Large language models (LLMs) are increasingly asked not only to write static interfaces, but to construct executable interactive worlds from natural lan

XAI-SOH-FL: Enhancing SOH-FL with Adaptive Aggregation and Explainable AI for Intrusion Detection in Heterogeneous IoT

arXiv:2606.00134v1 Announce Type: cross Abstract: Intrusion Detection Systems (IDS) in Internet of Things (IoT) environments face significant challenges due to data heterogeneity, lack of labeled data

You Can Learn Tokenization End-to-End with Reinforcement Learning

arXiv:2602.13940v2 Announce Type: replace-cross Abstract: Tokenization is a hardcoded compression step which remains in the training pipeline of Large Language Models (LLMs), despite a general trend t

Tutorials

You Don't Need All That Attention: Surgical Memorization Mitigation in Text-to-Image Diffusion Models

arXiv:2603.00133v2 Announce Type: replace-cross Abstract: Generative models have been shown to 'memorize' certain training data, leading to verbatim or near-verbatim generating images, which may cause

tutorialsarxiv-cs-ai

Zamba2-VL Technical Report

arXiv:2606.00390v1 Announce Type: cross Abstract: We present Zamba2-VL, a suite of vision-language models built on Zamba2, a hybrid language-model architecture combining Mamba2 state-space layers with

Zero-Shot Off-Policy Learning

arXiv:2602.01962v2 Announce Type: replace-cross Abstract: Off-policy learning methods seek to derive an optimal policy directly from a fixed dataset of prior interactions. This objective presents sign

A Behavioural and Representational Evaluation of Goal-Directedness in Language Model Agents

arXiv:2602.08964v2 Announce Type: replace-cross Abstract: Understanding an agent's goals helps explain and predict its behaviour, yet there is no established methodology for reliably attributing goals

A Kinetic Energy Perspective of Flow Matching

arXiv:2602.07928v2 Announce Type: replace-cross Abstract: Flow-based generative models can be viewed through a physics lens: sampling transports a particle from noise to data by integrating a learned

A Novel Global Context-aware Deep Neural Network for Enhanced Brain Tumor Segmentation using Magnetic Resonance Images

arXiv:2605.30510v1 Announce Type: cross Abstract: Brain cancer's severity necessitates precise brain tumor segmentation, which is crucial for effective brain tumor diagnosis. Manual identification, bu

A Persona-Based Evaluation Framework for Pluralistic Alignment in Generative AI

arXiv:2605.31021v1 Announce Type: new Abstract: Current alignment paradigms for generative artificial intelligence rely predominantly on monolithic benchmarking frameworks that reduce the plurality of

A Pilot Study on Curator-Guided Multilingual Art Description for Blind and Low-Vision Audiences with Small Vision-Language Models

arXiv:2605.31080v1 Announce Type: cross Abstract: Blind and low-vision (BLV) audiences remain underserved by visual art descriptions, particularly across languages and in museum settings where privacy

A Unified and Reproducible Experimentation Framework for Speech Understanding

arXiv:2605.30899v1 Announce Type: cross Abstract: Speech foundation models and Speech LLMs have advanced speech understanding, yet deployment-oriented model selection is hindered by non-comparable eva

A Unified Framework for Gradient Aggregation in Multi-Objective Optimization

arXiv:2605.30452v1 Announce Type: cross Abstract: Many machine learning problems involve multiple inherent trade-offs that are best addressed by gradient-based multi-objective optimization (MOO) algor

Active Timepoint Selection for Learning Measure-Valued Trajectories

arXiv:2605.30625v1 Announce Type: cross Abstract: Inferring continuous probability paths from sparse snapshots is a fundamental challenge in domains like single-cell biology, where high-fidelity data

AI Loss of Control Incident Management: Response & Resilience

arXiv:2605.30406v1 Announce Type: cross Abstract: Recent research demonstrating AI systems exhibiting deception and shutdown resistance suggests that AI loss of control (LOC) is an urgent policy conce

AMix-2: Establishing Protein as a Native Modality in Large Language Models

arXiv:2605.30963v1 Announce Type: cross Abstract: We present AMix-2, a protein-text foundation model that establishes protein as a native modality in large language models (LLMs), unifying protein und

An Odd Estimator for Shapley Values

arXiv:2602.01399v2 Announce Type: replace-cross Abstract: The Shapley value is a ubiquitous framework for attribution in machine learning, encompassing feature importance, data valuation, and causal i

Local Ai

An Organization-Scoped LLM Agent Runtime Architecture for Regulated Cybersecurity Operations

arXiv:2605.30604v1 Announce Type: cross Abstract: Regulated cybersecurity workflows lack a runtime substrate that enforces organization-level scope across retrieval, tool calls, memory, findings, repo

local-aiarxiv-cs-ai

AnchorSteer: Self-Discovered Concept Injection for Structure-Preserving Music Editing

arXiv:2605.31053v1 Announce Type: cross Abstract: Controllable music editing is to modify high-level attributes while strictly preserving rhythmic and melodic structures. However, this task is challen

Annealed Softmax Greedy in Many-Armed Bayesian Bandits

arXiv:2605.31034v1 Announce Type: cross Abstract: Reinforcement learning with verifiable rewards (RLVR) and group-based policy optimization methods such as GRPO update a stochastic policy by sampling

Answer-Set-Programming-based Abstractions for Reinforcement Learning

arXiv:2605.31444v1 Announce Type: new Abstract: Reinforcement Learning (RL) enables autonomous agents to learn policies from experience, but realistic problems often involve enormous state spaces, mak

Tutorials

Appropriateness of Empathy in AI: A Signal-Cost Perspective

arXiv:2605.31340v1 Announce Type: cross Abstract: The appropriateness of empathy in AI has emerged as a critical concern, as excessive empathy risks seeming manipulative while insufficient empathy app

tutorialsarxiv-cs-ai

Auto-Discovery-Bench: Diagnosing Structured State Tracking in Oracle-Guided Discovery

arXiv:2502.15224v2 Announce Type: replace-cross Abstract: Interactive discovery requires agents to maintain and update structured beliefs over many rounds of feedback. Before evaluating agents in nois

Automatically Attacking Software Reverse Engineering AI Agents

arXiv:2605.30667v1 Announce Type: cross Abstract: Software tools for reverse engineering executable binary files, such as Ghidra, enable malware analysts to safely conduct robust static analysis witho

Autoregressive Visual Generation Needs a Prologue

arXiv:2605.06137v2 Announce Type: replace-cross Abstract: In this work, we propose Prologue, an approach to bridging the reconstruction-generation gap in autoregressive (AR) image generation. Instead

AutoSci: A Memory-Centric Agentic System for the Full Scientific Research Lifecycle

arXiv:2605.31468v1 Announce Type: new Abstract: Scientific research has traditionally been human-intensive, requiring researchers to coordinate literature, ideas, experiments, manuscripts, and review

Benchmarking and Enhancing Text-to-Image Models for Generating Visual Representations in Early Arithmetic Education

arXiv:2605.31212v1 Announce Type: cross Abstract: AI systems are increasingly used to support educational content creation, yet it remains unclear whether they can generate outputs that faithfully rep

Benchmarking Machine Learning Uncertainty Quantification Methodologies for Predicting Turbine Gas Temperature Degradation

arXiv:2605.30585v1 Announce Type: cross Abstract: Effective prognostics and health management of modern engines relies on accurate turbine gas temperature predictions and robust uncertainty quantifica

Beyond Agreement: Scoring Panel-Surfaced Biomedical Entity Candidates for Curator Triage

arXiv:2605.30826v1 Announce Type: cross Abstract: Biomedical NER is deceptively simple for modern LLMs: plausible biomedical mentions are easy to surface, but corpus-convention correctness depends on

Beyond Classification: Dynamic Adapter Routing for Continual Multimodal Retrieval

arXiv:2605.31229v1 Announce Type: cross Abstract: While retrieval is a core function of vision-language models, continually updating these models for retrieval tasks remains critically underexplored.

Applications

Beyond Memorization: Assessing Semantic Generalization in Large Language Models Using Phrasal Constructions

arXiv:2501.04661v3 Announce Type: replace-cross Abstract: The web-scale of pretraining data has created an important evaluation challenge: to disentangle linguistic competence on cases well-represente

applicationsarxiv-cs-ai

Biases in the Blind Spot: Detecting What LLMs Fail to Mention

arXiv:2602.10117v5 Announce Type: replace-cross Abstract: Large Language Models (LLMs) often provide chain-of-thought (CoT) reasoning traces that appear plausible, but may hide internal biases. We cal

BilliardPhys-Bench: Benchmarking Physical Reasoning and Visual Dynamics of Multimodal LLMs

arXiv:2605.30900v1 Announce Type: new Abstract: Current multimodal models handle static image recognition well, but intuitive physical reasoning remains a weakness. Predicting how objects will move an