All entries

The Art of Building Verifiers for Computer Use Agents

arXiv:2604.06240v1 Announce Type: cross Abstract: Verifying the success of computer use agent (CUA) trajectories is a critical challenge: without reliable verification, neither evaluation nor training

agentsarxiv-cs-ai

The Art of (Mis)alignment: How Fine-Tuning Methods Effectively Misalign and Realign LLMs in Post-Training

All Blog X Post Paper YouTube Reddit GitHub

Clear filters

DGX agent

arXiv:2604.07754v1 Announce Type: cross Abstract: The deployment of large language models (LLMs) raises significant ethical and safety concerns. While LLM alignment techniques are adopted to improve m

safetyarxiv-cs-cl

10 Apr 2026

The ATOM Report: Measuring the Open Language Model Ecosystem

arXiv:2604.07190v1 Announce Type: cross Abstract: We present a comprehensive adoption snapshot of the leading open language models and who is building them, focusing on the ~1.5K mainline open models

The Defense Trilemma: Why Prompt Injection Defense Wrappers Fail?

arXiv:2604.06436v2 Announce Type: cross Abstract: We prove that no continuous, utility-preserving wrapper defense-a function D: Xo X that preprocesses inputs before the model sees them-can make al

The Depth Ceiling: On the Limits of Large Language Models in Discovering Latent Planning

arXiv:2604.06427v1 Announce Type: cross Abstract: The viability of chain-of-thought (CoT) monitoring hinges on models being unable to reason effectively in their latent representations. Yet little is

The Detection-Extraction Gap: Models Know the Answer Before They Can Say It

arXiv:2604.06613v2 Announce Type: cross Abstract: Modern reasoning models continue generating long after the answer is already determined. Across five model configurations, two families, and three ben

The End of the Foundation Model Era: Open-Weight Models, Sovereign AI, and Inference as Infrastructure

arXiv:2604.06217v1 Announce Type: cross Abstract: The foundation model era -- roughly 2020 to 2025 -- is over. The forces that defined it have inverted. Open source models have reached frontier perfor

The Geometry of Forgetting

arXiv:2604.06222v1 Announce Type: cross Abstract: Why do we forget? Why do we remember things that never happened? The conventional answer points to biological hardware. We propose a different one: ge

The Human Condition as Reflected in Contemporary Large Language Models

arXiv:2604.06206v1 Announce Type: cross Abstract: This study seeks to uncover evidence of a latent structure in evolved human culture as it is refracted through contemporary large language models (LLM

The Illusion of Stochasticity in LLMs

arXiv:2604.06543v1 Announce Type: cross Abstract: In this work, we demonstrate that reliable stochastic sampling is a fundamental yet unfulfilled requirement for Large Language Models (LLMs) operating

agentsarxiv-cs-lg

The Illusion of Superposition? A Principled Analysis of Latent Thinking in Language Models

arXiv:2604.06374v1 Announce Type: cross Abstract: Latent reasoning via continuous chain-of-thoughts (Latent CoT) has emerged as a promising alternative to discrete CoT reasoning. Operating in continuo

The Impact of Steering Large Language Models with Persona Vectors in Educational Applications

arXiv:2604.07102v1 Announce Type: cross Abstract: Activation-based steering can personalize large language models at inference time, but its effects in educational settings remain unclear. We study pe

The Master Key Hypothesis: Unlocking Cross-Model Capability Transfer via Linear Subspace Alignment

arXiv:2604.06377v1 Announce Type: cross Abstract: We investigate whether post-trained capabilities can be transferred across models without retraining, with a focus on transfer across different model

The Persistence of Cultural Memory: Investigating Multimodal Iconicity in Diffusion Models

arXiv:2511.11435v3 Announce Type: replace Abstract: The ambiguity between generalization and memorization in TTI diffusion models becomes pronounced when prompts invoke culturally shared visual refere

researcharxiv-cs-cv

The Planetary Cost of AI Acceleration, Part II: The 10th Planetary Boundary and the 6.5-Year Countdown

arXiv:2604.04956v2 Announce Type: replace-cross Abstract: The recent, super-exponential scaling of autonomous Large Language Model (LLM) agents signals a broader, fundamental paradigm shift from machi

agentsarxiv-cs-ai

The Rhetoric of Machine Learning

arXiv:2604.06754v1 Announce Type: new Abstract: I examine the technology of machine learning from the perspective of rhetoric, which is simply the art of persuasion. Rather than being a neutral and 'o

The Stepwise Informativeness Assumption: Why are Entropy Dynamics and Reasoning Correlated in LLMs?

arXiv:2604.06192v1 Announce Type: cross Abstract: Recent work uses entropy-based signals at multiple representation levels to study reasoning in large language models, but the field remains largely em

The Sustainability Gap in Robotics: A Large-Scale Survey of Sustainability Awareness in 50,000 Research Articles

arXiv:2604.07921v1 Announce Type: new Abstract: We present a large-scale survey of sustainability communication and motivation in robotics research. Our analysis covers nearly 50,000 open-access paper

safetyarxiv-cs-ro

The Theorems of Dr. David Blackwell and Their Contributions to Artificial Intelligence

arXiv:2604.06621v1 Announce Type: cross Abstract: Dr. David Blackwell was a mathematician and statistician of the first rank, whose contributions to statistical theory, game theory, and decision theor

safetyarxiv-cs-lg

The Theory and Practice of Highly Scalable Gaussian Process Regression with Nearest Neighbours

arXiv:2604.07267v1 Announce Type: cross Abstract: Gaussian process (GP) regression is a widely used non-parametric modeling tool, but its cubic complexity in the training size limits its use on mass

The Traveling Thief Problem with Time Windows: Benchmarks and Heuristics

arXiv:2604.06724v1 Announce Type: cross Abstract: While traditional optimization problems were often studied in isolation, many real-world problems today require interdependence among multiple optimiz

The Unreasonable Effectiveness of Data for Recommender Systems

arXiv:2604.06420v2 Announce Type: cross Abstract: In recommender systems, collecting, storing, and processing large-scale interaction data is increasingly costly in terms of time, energy, and computat

The Weaponization of Computer Vision: Tracing Military-Surveillance Ties through Conference Sponsorship

arXiv:2604.07803v1 Announce Type: cross Abstract: Computer vision, a core domain of artificial intelligence (AI), is the field that enables the computational analysis, understanding, and generation of

researcharxiv-cs-cv

Theory and interpretability of Quantum Extreme Learning Machines: a Pauli-transfer matrix approach

arXiv:2602.18377v3 Announce Type: replace-cross Abstract: Quantum reservoir computers (QRCs) have emerged as a promising approach to quantum machine learning, since they utilize the natural dynamics o

Thinking in Graphs with CoMAP: A Shared Visual Workspace for Designing Project-Based Learning

arXiv:2604.06200v1 Announce Type: cross Abstract: Designing project-based learning (PBL) demands managing highly interdependent components, a task that both traditional linear tools and purely convers

Thompson Sampling for Infinite-Horizon Discounted Decision Processes

arXiv:2405.08253v3 Announce Type: replace-cross Abstract: This paper develops a viable notion of learning for sampling-based algorithms that applies in broader settings than previously considered. Mor

Through the Magnifying Glass: Adaptive Perception Magnification for Hallucination-Free VLM Decoding

arXiv:2503.10183v4 Announce Type: replace Abstract: Existing vision-language models (VLMs) often suffer from visual hallucination, where the generated responses contain inaccuracies that are not groun

researcharxiv-cs-cv

Tight Convergence Rates for Online Distributed Linear Estimation with Adversarial Measurements

arXiv:2604.06282v1 Announce Type: cross Abstract: We study mean estimation of a random vector X in a distributed parameter-server-worker setup. Worker i observes samples of a_i^op X, where $a_

Time-Series Classification with Multivariate Statistical Dependence Features

arXiv:2604.06537v1 Announce Type: new Abstract: In this paper, we propose a novel framework for non-stationary time-series analysis that replaces conventional correlation-based statistics with direct

Tool-MCoT: Tool Augmented Multimodal Chain-of-Thought for Content Safety Moderation

arXiv:2604.06205v1 Announce Type: cross Abstract: The growth of online platforms and user content requires strong content moderation systems that can handle complex inputs from various media types. Wh

Tool Retrieval Bridge: Aligning Vague Instructions with Retriever Preferences via Bridge Model

model-releasesarxiv-cs-cl

arXiv:2604.07816v1 Announce Type: new Abstract: Tool learning has emerged as a promising paradigm for large language models (LLMs) to address real-world challenges. Due to the extensive and irregularl

TOOLCAD: Exploring Tool-Using Large Language Models in Text-to-CAD Generation with Reinforcement Learning

arXiv:2604.07960v1 Announce Type: cross Abstract: Computer-Aided Design (CAD) is an expert-level task that relies on long-horizon reasoning and coherent modeling actions. Large Language Models (LLMs)

agentsarxiv-cs-cl

Toward a Tractability Frontier for Exact Relevance Certification

arXiv:2604.07349v1 Announce Type: cross Abstract: Exact relevance certification asks which coordinates are necessary to determine the optimal action in a coordinate-structured decision problem. The tr

Toward a universal foundation model for graph-structured data

arXiv:2604.06391v1 Announce Type: cross Abstract: Graphs are a central representation in biomedical research, capturing molecular interaction networks, gene regulatory circuits, cell--cell communicati

Toward Memory-Aided World Models: Benchmarking via Spatial Consistency

arXiv:2505.22976v2 Announce Type: replace-cross Abstract: The ability to simulate the world in a spatially consistent manner is a crucial requirements for effective world models. Such a model enables

Local Ai

Toward Personalized Darts Training: A Data-Driven Framework Based on Skeleton-Based Biomechanical Analysis and Motion Modeling

arXiv:2604.01130v3 Announce Type: replace Abstract: As sports training becomes more data-driven, traditional dart coaching based mainly on experience and visual observation is increasingly inadequate

local-aiarxiv-cs-lg

Toward Reducing Unproductive Container Moves: Predicting Service Requirements and Dwell Times

arXiv:2604.06251v1 Announce Type: new Abstract: This article presents the results of a data science study conducted at a container terminal, aimed at reducing unproductive container moves through the

Towards Accurate and Calibrated Classification: Regularizing Cross-Entropy From A Generative Perspective

arXiv:2604.06689v1 Announce Type: new Abstract: Accurate classification requires not only high predictive accuracy but also well-calibrated confidence estimates. Yet, modern deep neural networks (DNNs

Towards Effective Long Video Understanding of Multimodal Large Language Models via One-shot Clip Retrieval

model-releasesarxiv-cs-cv

arXiv:2512.08410v2 Announce Type: replace Abstract: Due to excessive memory overhead, most Multimodal Large Language Models (MLLMs) can only process videos of limited frames. In this paper, we propose

Towards Hierarchical Multi-Step Reward Models for Enhanced Reasoning in Large Language Models

arXiv:2503.13551v5 Announce Type: replace Abstract: Recent studies show that Large Language Models (LLMs) achieve strong reasoning capabilities through supervised fine-tuning or reinforcement learning

researcharxiv-cs-cl

Towards Privacy-Preserving Large Language Model: Text-free Inference Through Alignment and Adaptation

arXiv:2604.06831v1 Announce Type: cross Abstract: Current LLM-based services typically require users to submit raw text regardless of its sensitivity. While intuitive, such practice introduces substan

Towards provable probabilistic safety for scalable embodied AI systems

arXiv:2506.05171v3 Announce Type: replace-cross Abstract: Embodied AI systems, comprising AI models and physical plants, are increasingly prevalent across various applications. Due to the rarity of sy

Towards Real-world Human Behavior Simulation: Benchmarking Large Language Models on Long-horizon, Cross-scenario, Heterogeneous Behavior Traces

model-releasesarxiv-cs-cl

arXiv:2604.08362v1 Announce Type: new Abstract: The emergence of Large Language Models (LLMs) has illuminated the potential for a general-purpose user simulator. However, existing benchmarks remain co

Towards Resilient Intrusion Detection in CubeSats: Challenges, TinyML Solutions, and Future Directions

arXiv:2604.06411v1 Announce Type: cross Abstract: CubeSats have revolutionized access to space by providing affordable and accessible platforms for research and education. However, their reliance on C

agentsarxiv-cs-ai

Towards Robust Content Watermarking Against Removal and Forgery Attacks

arXiv:2604.06662v1 Announce Type: cross Abstract: Generated contents have raised serious concerns about copyright protection, image provenance, and credit attribution. A potential solution for these p

Towards the Development of an LLM-Based Methodology for Automated Security Profiling in Compliance with Ukrainian Cybersecurity Regulations

arXiv:2604.06274v1 Announce Type: cross Abstract: In recent years, the pace of development of information technology in various areas has increased drastically, forcing cybersecurity specialists to co

ToxReason: A Benchmark for Mechanistic Chemical Toxicity Reasoning via Adverse Outcome Pathway

arXiv:2604.06264v1 Announce Type: cross Abstract: Recent advances in large language models (LLMs) have enabled molecular reasoning for property prediction. However, toxicity arises from complex biolog

TR-EduVSum: A Turkish-Focused Dataset and Consensus Framework for Educational Video Summarization

model-releasesarxiv-cs-cl

arXiv:2604.07553v1 Announce Type: new Abstract: This study presents a framework for generating the gold-standard summary fully automatically and reproducibly based on multiple human summaries of Turki