CV

Education

• Ph.D. in Artificial Intelligence, Zhejiang University, 2022 – present
  • Member of REAL Lab, advised by Dr. Yongliang Shen and Dr. Jian Shao
  • Research directions: LLM reasoning, AI agents
  • Coursework: Natural Language Processing, Knowledge Graphs, Knowledge Reasoning & Representation
• B.S. in Information Management and Information Systems, Beijing Normal University, 2018 – 2022
  • Minor: Data Science and Big Data Technology, Beijing Normal University, 2020 – 2022

Honors & Awards

• National Scholarship (国家奖学金)
• Beijing Merit Student (北京市三好学生)
• Beijing Outstanding Graduate (北京市优秀毕业生)

Internships

• Tencent, Hunyuan VLM Post-Training Team (Qingyun Program), Apr 2026 to present
  • Vision post-training: SFT, OPD, and RL optimization
  • Subjective-task RL for Yuanbao user scenarios; SFT-RL post-training strategies
  • Multi-teacher distillation (OPD) for large-scale MoE models
• Ant Group, Ling Foundation Model Team, Aug 2025 to Apr 2026
  • Capability optimization of the Ring series of reasoning LLMs; adaptive reasoning-effort tuning
  • Co-authored the Ring-1T technical report (trillion-parameter reasoning model)
  • Agent-based SWE task capabilities; RL research for LLMs
• Meituan, LongCat Foundation Model Team, Jun 2023 to Jul 2025
  • Math-specialized data cleaning, preprocessing, and mixing across pre-training, annealing, and alignment stages
  • Training experience with Dense and MoE models (1B / 7B / >100B parameters)
  • Built data-synthesis and evaluation pipelines and code frameworks

Skills

• Programming: Python (advanced features), Git, Linux
• LLM training & inference: Megatron-Core (pre-training / fine-tuning), veRL (RL framework), vLLM, SGLang

Publications

Milestone-Guided Policy Learning for Long-Horizon Language Agents

InftyThink+: Effective and Efficient Infinite-Horizon Reasoning via Reinforcement Learning

InftyThink: Breaking the Length Limits of Long-Context Reasoning in Large Language Models

MathFimer: Enhancing Mathematical Reasoning by Expanding Reasoning Steps through Fill-in-the-Middle Task

VerifyBench: Benchmarking Reference-based Reward Systems for Large Language Models

SpatialLadder: Progressive Training for Spatial Reasoning in Vision-Language Models

OmniEAR: Benchmarking Agent Reasoning in Embodied Tasks

Teaching LLMs According to Their Aptitude: Adaptive Switching Between CoT and TIR for Mathematical Problem Solving

Test-Time Reinforcement Learning for GUI Grounding via Region Consistency

Let LRMs Break Free from Overthinking via Self-Braking Tuning

Mind the Gap: Bridging Thought Leap for Improved Chain-of-Thought Tuning

Do Large Language Models excel in Complex Logical Reasoning with Formal Language?

AskToAct: Enhancing LLMs Tool Use via Self-Correcting Clarification

Every Step Evolves: Scaling Reinforcement Learning for Trillion-Scale Thinking Model (Ring-1T)

SVGenius: Benchmarking LLMs in SVG Understanding, Editing and Generation

GSM8K-V: Can Vision Language Models Solve Grade School Math Word Problems in Visual Contexts

Cooper: Co-Optimizing Policy and Reward Models in Reinforcement Learning for Large Language Models

UGPhysics: A Comprehensive Benchmark for Undergraduate Physics Reasoning with Large Language Models

LogicPro: Improving Complex Logical Reasoning via Program-Guided Learning

S³cMath: Spontaneous Step-Level Self-Correction Makes Large Language Models Better Mathematical Reasoners

Triad: A Framework Leveraging a Multi-Role LLM-based Agent to Solve Knowledge Base Question Answering