Search: “reasoning”

How Much Online RL is Enough? Informative Rollouts for Offline Preference Optimization in RLVR

arXiv:2605.21266v1 Announce Type: new Abstract: Reinforcement Learning from Verifiable Rewards (RLVR) has emerged as a powerful paradigm for reasoning in language models, with GRPO as its primary exam

Human

Content type

All Blog X Post Paper YouTube Reddit GitHub

Clear filters

Industry

Marc Andreessen (@pmarca) on what it’s like to work at SpaceX: “It’s like being dropped into a zone of shocking competence.” “Everybody is u…

DGX agent

Marc Andreessen (@pmarca) on what it’s like to work at SpaceX: “It’s like being dropped into a zone of shocking competence.” “Everybody is ultra competent, and the reason everybody’s ultra competent i

industryelon-musk--x

21 May 2026

Hardware

PlexRL: Cluster-Level Orchestration of Serviceized LLM Execution for RLVR

arXiv:2605.20863v1 Announce Type: cross Abstract: Reinforcement learning with verifiable rewards (RLVR) has recently unlocked strong reasoning capabilities in large language models (LLMs), triggering

hardwarearxiv-cs-lg

Pseudo-Formalization for Automatic Proof Verification

arXiv:2605.20531v1 Announce Type: cross Abstract: Reliable verification of proofs remains a bottleneck for training and evaluating AI systems on hard mathematical reasoning. Fully formal proofs, in la

🚀Qwen3.7-Max just landed at 56.6 on the Artificial Analysis Intelligence Index — a solid 4.8pt jump over Qwen3.6-Max-Preview. @ArtificialAn…

🚀Qwen3.7-Max just landed at 56.6 on the Artificial Analysis Intelligence Index — a solid 4.8pt jump over Qwen3.6-Max-Preview. @ArtificialAnlys ⚡️Sharper sci reasoning, stronger agentic chops, better c

agentsqwen--x

Retrospective Sparse Attention for Efficient Long-Context Generation

arXiv:2508.09001v2 Announce Type: replace Abstract: Large Language Models (LLMs) are increasingly deployed in long-context tasks such as reasoning, code generation, and multi-turn dialogue. However, i

researcharxiv-cs-cl

Strategy-Induct: Task-Level Strategy Induction for Instruction Generation

arXiv:2605.20924v1 Announce Type: new Abstract: Designing effective task-level prompts is crucial for improving the performance of Large Language Models (LLMs). While prior work on instruction inducti

researcharxiv-cs-cl

VISTAQA: Benchmarking Joint Visual Question Answering and Pixel-Level Evidence

model-releasesarxiv-cs-cv

arXiv:2605.20676v1 Announce Type: new Abstract: Establishing a clear link between model predictions and the visual evidence that supports them is critical for transparency and reliability in multimoda

You Only Need Minimal RLVR Training: Extrapolating LLMs via Rank-1 Trajectories

model-releasesarxiv-cs-cl

arXiv:2605.21468v1 Announce Type: cross Abstract: Reinforcement learning with verifiable rewards (RLVR) has become a dominant paradigm for improving reasoning in large language models (LLMs), yet the

Artifact-Bench: Evaluating MLLMs on Detecting and Assessing the Artifacts of AI-Generated Videos

model-releasesarxiv-cs-cv

arXiv:2605.18984v1 Announce Type: new Abstract: Recent video generative models have greatly improved the realism of AI-generated videos, yet their outputs still exhibit artifacts such as temporal inco

Artificial Phantasia: Emergent Mental Imagery in Large Language Models

arXiv:2509.23108v2 Announce Type: replace Abstract: Can visual imagery be driven solely by language? This idea goes against cognitive science's traditional view that visual mental imagery is only poss

Beyond Majority Voting: LLM Aggregation by Leveraging Higher-Order Information

arXiv:2510.01499v2 Announce Type: replace-cross Abstract: With the rapid progress of multi-agent large language model (LLM) reasoning, how to effectively aggregate answers from multiple LLMs has emerg

agentsarxiv-cs-ai

CEPO: RLVR Self-Distillation using Contrastive Evidence Policy Optimization

arXiv:2605.19436v1 Announce Type: cross Abstract: When a model produces a correct solution under reinforcement learning with verifiable rewards (RLVR), every token receives the same reward signal rega

safetyarxiv-cs-cl

DarkLLM: Learning Language-Driven Adversarial Attacks with Large Language Models

arXiv:2605.18868v1 Announce Type: cross Abstract: While vision and multimodal foundation models underpin critical tasks from perception to complex reasoning, they remain highly vulnerable to adversari

Formal Skill: Programmable Runtime Skills for Efficient and Accurate LLM Agents

arXiv:2605.19604v1 Announce Type: new Abstract: Large Language Model (LLM) agents increasingly act inside real workspaces, where tools and skills determine whether model reasoning becomes reliable act

Hybrid-LoRA: Bridging Full Fine-Tuning and Low-Rank Adaptation for Post-Training

arXiv:2605.18822v1 Announce Type: cross Abstract: Post-training has become essential for adapting large language models (LLMs) to complex downstream behaviors, including instruction following, prefere

Local Ai

iDiff: Interpretable Difference-aware Framework for Pairwise Image Quality Assessment

arXiv:2605.19522v1 Announce Type: new Abstract: Pairwise image quality assessment (IQA) in professional photography requires a model not only to identify the preferred image between two candidates, bu

local-aiarxiv-cs-cv

Inferring Sensitive Attributes from Knowledge Graph Embeddings: Attack and Defense Strategies

arXiv:2605.19644v1 Announce Type: cross Abstract: Knowledge Graphs (KGs) are a powerful representation of linked data, offering flexibility, semantic richness, and support for knowledge enrichment and

applicationsarxiv-cs-lg

LP-Eval: Rubric and Dataset for Measuring the Quality of Legal Proposition Generation

arXiv:2605.19815v1 Announce Type: cross Abstract: Legal proposition generation is central to legal reasoning and doctrinal scholarship, yet remain under-examined in Legal NLP. This paper investigates

Lying Is Just a Phase: The Hidden Alignment Transition in Language Model Scaling

arXiv:2605.18838v1 Announce Type: cross Abstract: Scaling laws predict loss from compute but not how capabilities interact. We measure the coupling between reasoning and truthfulness across 63 base mo

MOCHA: Multi-Objective Chebyshev Annealing for Agent Skill Optimization

arXiv:2605.19330v1 Announce Type: new Abstract: LLM agents organize behavior through skills - structured natural-language specifications governing how an agent reasons, retrieves, and responds. Unlike

agentsarxiv-cs-ai

Hardware

OScaR: The Occam's Razor for Extreme KV Cache Quantization in LLMs and Beyond

arXiv:2605.19660v1 Announce Type: cross Abstract: The rapid advancement toward long-context reasoning and multi-modal intelligence has made the memory footprint of the Key-Value (KV) cache a dominant

hardwarearxiv-cs-cl

PAVE: A Cognitive Architecture for Legitimate Violation in Generative Agent Societies

arXiv:2605.19351v1 Announce Type: cross Abstract: Generative agents based on large language models reproduce believable human behavior in cooperative settings, but how they should reason in situations

agentsarxiv-cs-ai

Probabilistic Tiny Recursive Model

arXiv:2605.19943v1 Announce Type: new Abstract: Tiny Recursive Models (TRM) solve complex reasoning tasks with a fraction of the parameters of modern large language models (LLMs) by iteratively refini

Sequential Consensus for Multi-Agent LLM Debates: A Wald-SPRT compute governor with calibration-based failure detection

arXiv:2605.19193v1 Announce Type: new Abstract: Multi-agent LLM debate improves factuality and reasoning, but most recipes pick a fixed round count, over-spending on easy items and under-spending on h

Spatial-MLLM: Boosting MLLM Capabilities in Visual-based Spatial Intelligence

arXiv:2505.23747v2 Announce Type: replace-cross Abstract: Recent advancements in Multimodal Large Language Models (MLLMs) have significantly enhanced performance on 2D visual tasks. However, improving

The Growing Pains of Frontier Models: When Leaderboards Stop Separating and What to Measure Next

arXiv:2605.18840v1 Announce Type: cross Abstract: Leaderboards rank frontier models on independent axes but do not reveal whether capabilities reinforce or trade off across releases -- and at the fron

Trustworthy Agent Network: Trust in Agent Networks Must Be Baked In, Not Bolted On

arXiv:2605.19035v1 Announce Type: new Abstract: The rapid advancement of Large Language Models has given rise to autonomous LLM-based agents capable of complex reasoning and execution. As these agents

When the Majority Votes Wrong, the Intervention Timing for Test-Time Reinforcement Learning Hides in the Extinction Window

arXiv:2605.19444v1 Announce Type: cross Abstract: Test-time reinforcement learning (TTRL) reports substantial accuracy gains on mathematical reasoning benchmarks using majority vote as a pseudo-label

When to Stop Reusing: Dynamic Gradient Gating for Sample-Efficient RLVR

arXiv:2605.19425v1 Announce Type: cross Abstract: Reinforcement Learning with Verifiable Rewards (RLVR) has become the dominant paradigm for advanced reasoning in Large Language Models (LLMs), but rol

Whispers of Wealth: Red-Teaming Google's Agent Payments Protocol via Prompt Injection

arXiv:2601.22569v2 Announce Type: replace-cross Abstract: Large language model (LLM) based agents are increasingly used to automate financial transactions, yet their reliance on contextual reasoning e

AI of the People, by the People, for the People: A Social Choice Approach to Collective Control of Artificial Intelligence

arXiv:2605.16291v1 Announce Type: cross Abstract: With the growing adoption of AI systems, reasoning about how society can exert control over AI becomes an increasingly urgent problem. Existing work o

AMR-SD: Asymmetric Meta-Reflective Self-Distillation for Token-Level Credit Assignment

arXiv:2605.18529v1 Announce Type: new Abstract: The alignment of Large Language Models (LLMs) for complex reasoning heavily relies on Reinforcement Learning with Verifiable Rewards (RLVR). However, st

An Empirical Study of Privacy Leakage Chains via Prompt Injection in Black-Box Chatbot Environments

arXiv:2605.18133v1 Announce Type: cross Abstract: LLM-based chatbot agents increasingly process user requests by combining natural-language reasoning with external tools such as web browsing. These ca

AscendOptimizer: Episodic Agent for Ascend NPU Operator Optimization

arXiv:2603.23566v2 Announce Type: replace-cross Abstract: Optimizing AscendC (Ascend C) operators for Ascend NPUs is difficult for two reasons. First, unlike CUDA, the ecosystem offers few public kern

Beyond Imperfect Alternatives with Rulemapping: A Neuro-Symbolic Case Study on Online Hate Speech

arXiv:2605.16280v1 Announce Type: cross Abstract: Automating legal reasoning forces a choice between imperfect alternatives: symbolic systems offer transparency but struggle with ambiguity, whereas ne

BioProAgent: Neuro-Symbolic Grounding for Constrained Scientific Planning

arXiv:2603.00876v2 Announce Type: replace Abstract: Large language models (LLMs) have demonstrated significant reasoning capabilities in scientific discovery but struggle to bridge the gap to physical

Local Ai

BLAgent: Agentic RAG for File-Level Bug Localization

arXiv:2605.17965v1 Announce Type: cross Abstract: Bug localization remains a key bottleneck in downstream software maintenance tasks, including root cause analysis, triage, and automated program repai

local-aiarxiv-cs-ai

CAM-Bench: A Benchmark for Computational and Applied Mathematics in Lean

arXiv:2605.17255v1 Announce Type: new Abstract: Formal theorem-proving benchmarks enable mechanically verifiable evaluation of mathematical reasoning in large language models. However, existing benchm

ChemVA: Advancing Large Language Models on Chemical Reaction Diagrams Understanding

arXiv:2605.17214v1 Announce Type: new Abstract: While Large Language Models (LLMs) have revolutionized scientific text processing, they exhibit a significant capability gap when interpreting chemical

Concise and Logically Consistent Conformal Sets for Neuro-Symbolic Concept-Based Models

arXiv:2605.18202v1 Announce Type: cross Abstract: Neuro-Symbolic Concept-based Models (NeSy-CBMs) are a family of architectures that integrate neural networks with symbolic reasoning for enhanced reli

Local Ai

CounterCount: A Diagnostic Framework for Counting Bias in Vision Language Models

arXiv:2605.17826v1 Announce Type: cross Abstract: Vision-Language Models (VLMs) excel at multimodal reasoning, yet it remains unclear whether their answers are grounded in visual evidence or driven by

local-aiarxiv-cs-ai

DecoupleSearch: Decouple Planning and Search via Hierarchical Reward Modeling

arXiv:2510.21712v2 Announce Type: replace-cross Abstract: Retrieval-Augmented Generation (RAG) systems have emerged as a pivotal methodology for enhancing Large Language Models (LLMs) through the dyna

DISA: Offline Importance Sampling for Distribution-Matching LLM-RL

arXiv:2605.17295v1 Announce Type: cross Abstract: Modern reasoning agents are increasingly evaluated on their ability to generate multiple valid solution paths, plans, or tool-use traces for a given i

safetyarxiv-cs-cl

DisasterVQA: A Visual Question Answering Benchmark Dataset for Disaster Scenes

model-releasesarxiv-cs-cv

arXiv:2601.13839v2 Announce Type: replace Abstract: Social media imagery provides a low-latency source of situational information during natural and human-induced disasters, enabling rapid damage asse

Distribution Transformers: Fast Approximate Bayesian Inference With On-The-Fly Prior Adaptation

arXiv:2502.02463v3 Announce Type: replace-cross Abstract: While Bayesian inference provides a principled framework for reasoning under uncertainty, its widespread adoption is limited by the intractabi

Do Vision-Language-Models show human-like logical problem-solving capability in point and click puzzle games?

arXiv:2605.11223v2 Announce Type: replace Abstract: Vision-Language(-Action) Models (VLMs) are increasingly applied to interactive environments, yet existing benchmarks often overlook the complex phys

Embodied Task Planning via Graph-Informed Action Generation with Large Language Models

model-releasesarxiv-cs-cl

arXiv:2601.21841v3 Announce Type: replace Abstract: While Large Language Models (LLMs) have demonstrated strong zero-shot reasoning capabilities, their deployment as embodied agents still faces fundam