Search: “reasoning”

AsyncOPD: How Stale Can On-Policy Distillation Be?

arXiv:2606.24143v1 Announce Type: new Abstract: On-policy distillation (OPD) trains a student on its own rollouts guided by teacher feedback and is becoming increasingly important for large language m

safetyarxiv-cs-lg

Autonomous Video Generation with Counterfactual Controllability for Self-Evolving World Models

Human

Content type

All Blog X Post Paper YouTube Reddit GitHub

Clear filters

DGX agent

arXiv:2606.24152v1 Announce Type: new Abstract: Existing literature claims that video generation essentially is world modelling. On the one hand, the claim is productive because it pushes generative A

agentsarxiv-cs-cv

24 Jun 2026

CAVEWOMAN: How Large Language Models Behave Under Linguistic Input and Output Compression

arXiv:2606.24083v1 Announce Type: cross Abstract: 'Talk short. Drop grammar. Save token.' This caveman style is widely promoted as a way to cut inference cost, but whether it actually saves anything d

Critique of Agent Model

arXiv:2606.23991v1 Announce Type: new Abstract: What is an agent? What constitutes agency? With the rise of Large Language Model (LLM) systems marketed as ``coding agents'', ``AI co-scientists'', and

safetyarxiv-cs-ai

Elon Musk built one of the largest AI compute clusters on earth. Yann LeCun just explained why xAI now rents it out to rivals instead of win…

Elon Musk built one of the largest AI compute clusters on earth. Yann LeCun just explained why xAI now rents it out to rivals instead of winning with it. Musk has antagonized so much AI talent he stru

researchyann-lecun--x

Emergent Relational Order in LLM Agent Societies: From Collective Affect to Authority Stratification

arXiv:2606.23764v1 Announce Type: cross Abstract: Fei Xiaotong's Differential Order Pattern characterizes rural society as egocentric and relationally graded, with cooperation attenuating over social

agentsarxiv-cs-ai

Gemini 3 Pro was the first model to achieve at least 23% on ARC-AGI-2, which it did in November, 2025 (it actually scored 31%). So the 8-12 …

model-releasesethan-mollick--x

Gemini 3 Pro was the first model to achieve at least 23% on ARC-AGI-2, which it did in November, 2025 (it actually scored 31%). So the 8-12 month gap between closed and open weights models still seems

GENA3D: Generative Amodal 3D Modeling by Bridging 2D Priors and 3D Coherence

arXiv:2511.21945v3 Announce Type: replace Abstract: Generating complete 3D objects under partial occlusions (i.e., amodal scenarios) is a practically important yet challenging problem, as large portio

safetyarxiv-cs-cv

HANCLIP: A Family of Hyperbolic Angular Negation Vision Language Models

arXiv:2606.23843v1 Announce Type: new Abstract: Vision-Language Models (VLMs) are typically pre-trained on large-scale image-text datasets to capture semantic correspondences between visual content an

MedBench v5: A Dynamic, Process-Oriented, and Hallucination-Aware Benchmark for Clinical Multimodal Models

model-releasesarxiv-cs-cl

arXiv:2606.24155v1 Announce Type: new Abstract: Existing medical AI benchmarks lack process visibility, atomic skill evaluation, and integrated hallucination detection. We introduce MedBench v5, a red

Multimedia and Visual Analytics in the Agentic Era

arXiv:2504.06138v3 Announce Type: replace-cross Abstract: Professional users need tools to help them gain actionable insights from large multimedia collections. Foundation models and AI agents have ra

NavWM: A Unified Navigation World Model for Foresight-Driven Planning

arXiv:2606.24101v1 Announce Type: cross Abstract: Conventional visual navigation policies often struggle with myopic decision-making and mode collapse in complex environments. While world models offer

agentsarxiv-cs-cv

🧠 Paradigm II — Agent Foundation Model: world modeling as agent capability. Single-turn, non-agentic environment prediction → tested direct…

🧠 Paradigm II — Agent Foundation Model: world modeling as agent capability. Single-turn, non-agentic environment prediction → tested directly on multi-turn, tool-calling agent tasks. No agentic RL, no

agentsqwen--x

PETRA: Transforming Web Text for Petroleum-Engineering Domain Adaptation

model-releasesarxiv-cs-cl

arXiv:2606.24346v1 Announce Type: cross Abstract: Petroleum-engineering search exposes a supervision gap for strong general retrievers: relevant evidence exists in public web text, but domain relevanc

Quantifying Prior Dominance in RAG Systems

arXiv:2606.23695v1 Announce Type: cross Abstract: Retrieval-Augmented Generation (RAG) grounds Large Language Models in external knowledge, yet current evaluations rely on discrete heuristics that suf

researcharxiv-cs-ai

Robust Dual-Signal Fusion: Hybrid Neuro-Symbolic Gating with Compressed Chain-of-Thought Refinement for Irony Detection in Social Media Texts

arXiv:2606.16845v2 Announce Type: replace-cross Abstract: Small-scale Large Language Models (LLMs) natively default to literal semantic interpretations, making few-shot irony detection a persistent ch

researcharxiv-cs-ai

RoPE-Aware Bit Allocation for KV-Cache Quantization

model-releasesarxiv-cs-cl

arXiv:2606.24033v1 Announce Type: cross Abstract: Existing low-bit KV-cache quantizers often treat each cached key as a flat vector. Under RoPE, however, a key's contribution to a future attention log

Rule2Text: A Framework for Generating and Evaluating Natural Language Explanations of Knowledge Graph Rules

arXiv:2508.10971v2 Announce Type: replace-cross Abstract: Knowledge graphs (KGs) can be enhanced through rule mining; however, the resulting logical rules are often difficult for humans to interpret d

SAFARI: Scaling Long Horizon Agentic Fault Attribution via Active Investigation

arXiv:2606.24626v1 Announce Type: new Abstract: As autonomous agents tackle increasingly complex multi-step, multi-agent tasks, their execution trajectories have scaled beyond the constraints of even

Applications

Scaling Laws for Task-Specific LLM Distillation

arXiv:2606.24747v1 Announce Type: new Abstract: Large Language Models (LLMs) achieve strong performance across a growing range of domains, yet their scale poses deployment challenges in applications w

applicationsarxiv-cs-ai

TACTFUL: Tactile-Driven Exploration For Object Localization and Identification in Confined Environments

arXiv:2606.24712v1 Announce Type: cross Abstract: Humans effortlessly locate and identify objects by touch alone, even without vision. In contrast, robotic systems rely heavily on vision and struggle

safetyarxiv-cs-ai

This is the strongest ARC-AGI-2 performance to date by an open-source model.

model-releasesfrancois-chollet--x

This is the strongest ARC-AGI-2 performance to date by an open-source model. GLM-5.2 from @Zai_org on ARC-AGI (Verified) - ARC-AGI-2: 22.8%, 0.25 - ARC-AGI-1: 77.0%, 0.19 Performance is comparable wit

Trimming the Long-Tail of Visual World Modeling Evaluation

arXiv:2606.24256v1 Announce Type: new Abstract: Physical interactions follow a long-tailed distribution: a set of common and regular interactions dominates human experience and visual data, while a br

Trustworthy Image Authentication using Forensic Knowledge Graphs

arXiv:2606.23917v1 Announce Type: new Abstract: Advances in generative AI have made image falsification highly realistic, demanding trustworthy authentication systems. Existing forensic detectors can

researcharxiv-cs-cv

Tuning without Peeking: Provable Generalization Bounds and Robust LLM Post-Training

arXiv:2507.01752v4 Announce Type: replace-cross Abstract: Gradient-based optimization is the workhorse of deep learning, offering efficient and scalable training via backpropagation. However, exposing

Universal Guideline-Driven Image Clustering via a Hybrid LLM Agent

arXiv:2606.24094v1 Announce Type: new Abstract: Unifying image clustering across different clustering scenarios remains challenging due to fundamental gaps among tasks. We introduce a Guideline-Driven

agentsarxiv-cs-cv

video-SALMONN-R^3: Learning to ReWatch, ReAsk, and ReAnswer for Efficient Video Understanding

arXiv:2606.24477v1 Announce Type: cross Abstract: Video large language models (LLMs) are often constrained by computation and memory budgets, leading them to use reduced frame rates and spatial resolu

VieSpeaker: A Large-Scale Vietnamese Speaker Recognition Dataset Beyond Visual Dependency

arXiv:2606.24066v1 Announce Type: cross Abstract: Speaker recognition has advanced rapidly with large-scale training datasets, yet Vietnamese remains under-resourced, with existing corpora limited in

researcharxiv-cs-cl

VisCritic: Visual State Comparison as Process Reward for GUI Agents

arXiv:2606.24525v1 Announce Type: new Abstract: GUI agents powered by vision-language models show strong potential for automating digital tasks, yet frequently fail in long-horizon scenarios due to th

When Helpfulness Overrides Causal Caution: Context-Dependent Suppression and Recovery in LLMs

arXiv:2606.24370v1 Announce Type: new Abstract: Large language models (LLMs) are increasingly integrated into decision-support roles in business and policy contexts. While prior benchmark studies have

Local Ai

A DVDrive Approach for doScenes Instructed Driving Challenge

arXiv:2606.21623v1 Announce Type: new Abstract: Instruction-conditioned trajectory prediction is an emerging problem in autonomous driving, where a model predicts the future ego trajectory not only fr

local-aiarxiv-cs-cv

A-Evolve-Training: Autonomous Post-Training of a 30B Model

model-releasesarxiv-cs-lg

arXiv:2606.20657v1 Announce Type: cross Abstract: Post-training a frontier model is normally weeks of human work: proposing data and recipe changes, launching runs, reading evals, deciding what to kee

A Verifiable Search Is Not a Learnable Chain-of-Thought

model-releasesarxiv-cs-lg

arXiv:2606.21884v1 Announce Type: new Abstract: It is tempting to assume any task solvable by a short program can be taught to a model as its chain-of-thought: write the steps out, fine-tune, and the

AI Capex pushback from GS: 'If frontier intelligence can increasingly be developed in the East at a fraction of the cost incurred in the Wes…

AI Capex pushback from GS: 'If frontier intelligence can increasingly be developed in the East at a fraction of the cost incurred in the West … then the largest capital allocators are also the ones mo

safetygary-marcus--x

An Empirical Study of OpenPangu Quantization on Ascend NPUs

arXiv:2606.21257v1 Announce Type: new Abstract: OpenPangu models are attractive targets for private and domestic large-language-model deployment, yet their robustness under aggressive post-training qu

Any-Body Guard: Universal Safeguarding for Manipulation Policies via Action Masking

arXiv:2606.22278v1 Announce Type: cross Abstract: Ensuring safety of learning-enabled robotic manipulation across diverse embodiments and tasks still requires significant manual engineering. Existing

safetyarxiv-cs-lg

Bayesian Adaptation Gym: A Benchmark for the Bayesian Low-Rank Adaptation of Multi-Modal Language Models

model-releasesarxiv-cs-lg

arXiv:2606.22188v1 Announce Type: new Abstract: Large multi-modal language models are increasingly deployed in high-stakes domains, making well-calibrated uncertainty essential. Traditional Bayesian m

Benchmarking Vision-Language Models for Microscopic Plant Image Understanding

arXiv:2606.22497v1 Announce Type: new Abstract: Microscopic imaging provides essential visual evidence for studying plant biology and pathology at the cellular and subcellular levels. However, existin

BIT-Nav: Brain-Inspired Trajectory Memory for Embodied Navigation

arXiv:2606.21398v1 Announce Type: new Abstract: Vision-Language Models (VLMs) for embodied navigation rely on selecting a fixed number of frames from a growing trajectory history. As episodes extend,

researcharxiv-cs-ro

Hardware

Build an AI Scientist for Life Science Discovery with NVIDIA BioNeMo Agent Toolkit

The NVIDIA BioNeMo Agent Toolkit enables AI scientists to use structure prediction, molecular generation, docking, sequence analysis, design, and genomics as callable tools , allowing AI agents to exe

hardwarenvidia-developer

Build real agentic apps using CUGA: two dozen working examples on a lightweight harness

CUGA is a lightweight framework from IBM Research for building agentic AI applications, featuring two dozen practical working examples. The framework enables developers to create autonomous AI agents

agentshugging-face

Calibrated Sampling-Free Uncertainty Estimation in Bayesian Deep Learning

arXiv:2606.16214v3 Announce Type: replace Abstract: Modern deep learning models remain notoriously prone to overconfidence, limiting their reliability in high-stakes applications. Bayesian methods aim

Causal Reward World Models: Zero-shot Reward Design for Automated Skill Generation

arXiv:2606.23280v1 Announce Type: new Abstract: Automated Reward Design (ARD) aims to replace manual reward engineering in reinforcement learning with language-driven reward function synthesis. Howeve

researcharxiv-cs-ro

Cloak: Zero-Shot Cross-Embodiment Manipulation by Masking the End-Effector from the VLA

arXiv:2606.22836v1 Announce Type: new Abstract: We present Cloak, a training recipe that endows a Vision-Language-Action (VLA) model with zero-shot cross-embodiment transfer by cloaking the end-effect

researcharxiv-cs-ro

Closed-Loop Verbal Reinforcement Learning for Task-Level Robotic Planning

arXiv:2603.22169v2 Announce Type: replace Abstract: We propose a new Verbal Reinforcement Learning (VRL) framework for interpretable task-level planning in mobile robotic systems operating under execu

safetyarxiv-cs-ro

Cohort-Anchored Foundation Models for Electronic Health Records: From Risk Scores to Auditable Peer Cohorts

arXiv:2606.21885v1 Announce Type: new Abstract: Foundation models have achieved remarkable performance across medical question answering, imaging, and electronic health record (EHR) tasks, yet reliabl

safetyarxiv-cs-lg

CQD-SHAP: Explainable Complex Query Answering via Shapley Values

arXiv:2510.15623v2 Announce Type: replace Abstract: Complex query answering (CQA) goes beyond the widely studied link prediction task by addressing more sophisticated queries that require multi-hop re

CuratorKIT : Data Curation and Synthetic Data Generation for LLM Post-Training

arXiv:2606.21631v1 Announce Type: cross Abstract: Data curation is a critical part of post-training pipelines for large language models, yet existing tools often treat ingestion, deduplication, synthe