Haitao Mi

R-Zero: Self-Evolving Reasoning LLM from Zero Data

The Pensieve Paradigm: Stateful Language Models Mastering Their Own Context

DeepMath-103K: A Large-Scale, Challenging, Decontaminated, and Verifiable Mathematical Dataset for Advancing Reasoning

THE END OF MANUAL DECODING: TOWARDS TRULY END-TO-END LANGUAGE MODELS

Vision-SR1: Self-Rewarding Vision-Language Model via Reasoning Decomposition and Multi-Reward Policy Optimization

DeepCompress: A Dual Reward Strategy for Dynamically Exploring and Compressing Reasoning Chains

CDE: Curiosity-Driven Exploration for Efficient Reinforcement Learning in Large Language Models

InComeS: Integrating Compression and Selection Mechanisms into LLMs for Efficient Model Editing

On the Evolution of Language Models without Labels: Majority Drives Selection, Novelty Promotes Variation

WebAggregator: Scaling Complex Logical Information Aggregation for Web Agents Foundation Models

DeepTheorem: Advancing LLM Reasoning for Theorem Proving Through Natural Language and Reinforcement Learning

One Token to Fool LLM-as-a-Judge

CLUE: Non-parametric Verification from Experience via Hidden-State Clustering

VOGUE: Guiding Exploration with Visual Uncertainty Improves Multimodal Reasoning

Thoughts Are All Over the Place: On the Underthinking of Long Reasoning Models

Improving LLM General Preference Alignment via Optimistic Online Mirror Descent

Trust, But Verify: A Self-Verification Approach to Reinforcement Learning with Verifiable Rewards

Two Experts Are All You Need for Steering Thinking: Reinforcing Cognitive Effort in MoE Reasoning Models Without Additional Training

MPS-Prover: Advancing Stepwise Theorem Proving by Multi-Perspective Search and Data Curation

UniGist: Towards General and Hardware-aligned Sequence-level Long Context Compression

DOTS: Learning to Reason Dynamically in LLMs via Optimal Reasoning Trajectories Search

Iterative Nash Policy Optimization: Aligning LLMs with General Preferences via No-Regret Learning

The First Few Tokens Are All You Need: An Efficient and Effective Unsupervised Prefix Fine-Tuning Method for Reasoning Models

Do NOT Think That Much for 2+3=? On the Overthinking of Long Reasoning Models

HDFlow: Enhancing LLM Complex Problem-Solving with Hybrid Thinking and Dynamic Workflows

Thoughts Are All Over the Place: On the Underthinking of o1-Like LLMs

SIaM: Self-Improving Code-Assisted Mathematical Reasoning of Large Language Models