About Me
I am currently a Ph.D. student in Data Science and Artificial Intelligence at The Hong Kong Polytechnic University, supervised by Prof.
KC Tan. Previously, I received my M.S. in Applied Statistics from
Fudan University and my B.S. in Statistics from
Sun Yat-sen University.
My research spans policy diversity in traditional reinforcement learning and reinforcement learning for LLM reasoning (RL for LLM). In policy diversity, I study how to learn and recover diverse behaviors while maintaining policy quality. I also investigate how diverse opponent populations can improve multi-agent training. (ICLR 2023, NeurIPS 2023, ICLR 2025) In RL for LLM, I study how diversity-aware optimization and self-play can strengthen reasoning, and how to reduce redundant reasoning. I also develop robust evaluation methods to assess whether performance gains reflect genuine reasoning ability. (NeurIPS 2025 Spotlight, VAR-MATH, ICML 2026, NeurIPS 2026)
I am always happy to discuss research ideas around RL, LLM Reasoning, Interpretability, and EC + RL. Please feel free to contact me at nigelyaoj@gmail.com.
我目前是香港理工大学数据科学与人工智能方向的博士生,导师为 KC Tan 教授。此前,我在复旦大学获得应用统计硕士学位,在中山大学获得统计学理学学士学位。
我的研究从传统强化学习中的策略多样性,延伸至面向大语言模型推理的强化学习(RL for LLM)。在策略多样性方面,我研究如何在保持策略质量的同时,学习与恢复多样化行为。我也探索如何利用对手种群的多样性,改善多智能体对抗训练。(ICLR 2023、NeurIPS 2023、ICLR 2025)在 RL for LLM 方面,我关注如何通过多样性感知优化与自博弈训练提升推理能力,并减少冗余推理。我也研究稳健的评估方法,以检验性能提升是否反映真实推理能力。(NeurIPS 2025 Spotlight、VAR-MATH、ICML 2026、NeurIPS 2026)
欢迎交流 RL、LLM Reasoning、可解释性相关研究问题,也可以通过 nigelyaoj@gmail.com 联系我。
Experience
ByteDance Seed
Aug 2025 - Jan 2026Research Intern · LLM Evaluation & Efficient Reasoning
Built LLM evaluation benchmarks and automated scoring pipelines, with statistical analysis to assess evaluation reliability. Developed SLAT to reduce chain-of-thought length by about 50% while maintaining comparable accuracy (ICML 2026).
Amazon AI Lab
Oct 2020 - Feb 2022Research Intern · Generative Modeling
Developed generative models for interpretable multi-agent trajectory prediction (NeurIPS 2021). Proposed a self-supervised framework for amodal video object segmentation (NeurIPS 2022 Spotlight).
Tencent AI Lab
Jul 2022 - Apr 2024Independent Researcher · Reinforcement Learning
Studied policy diversity and population-based reinforcement learning for game AI (ICLR 2023, NeurIPS 2023). Developed methods to recover diverse policies from human gameplay and applied them to game AI (ICLR 2025).
Publications
🤖 RL for LLM
- [ICML 2026] SLAT: Segment-Level Adaptive Trimming for Efficient CoT Reasoning
This work studies how to reduce structural redundancy in chain-of-thought reasoning. It proposes segment-level adaptive trimming to selectively suppress low-utility redundant reasoning segments, improving the accuracy-efficiency trade-off for large reasoning models.
- [NeurIPS 2026] CasePlay: Self-Play Reinforcement Learning from Case Reports for Medical Reasoning
This work develops a self-play reinforcement learning framework that alternates between question generation and reasoning on clinical case reports to improve medical reasoning.
- [NeurIPS 2025 Spotlight] Diversity-Aware Policy Optimization for Large Language Model Reasoning
This work studies the relationship between solution diversity and reasoning potential in LLM reasoning, and proposes a diversity-aware policy optimization method for reinforcement learning training.
- [arXiv 2025] VAR-MATH: Probing True Mathematical Reasoning in LLMs via Symbolic Multi-Instance Benchmarks
This work introduces VAR-MATH, a symbolic multi-instance benchmark that probes whether LLM mathematical gains reflect genuine reasoning rather than benchmark-specific overfitting or memorization.
🎮 Diversity in RL
- [ICLR 2025] Diverse Policies Recovering via Pointwise Mutual Information Weighted Imitation Learning
This work focuses on recovering stylistically diverse policies from expert trajectories by weighting state-action pairs with pointwise mutual information.
- [NeurIPS 2023] Policy Space Diversity for Non-Transitive Games
This work studies diversity in non-transitive games and proposes a policy-space diversity measure that is more aligned with Nash approximation quality.
- [ICLR 2023] Quality-Similar Diversity via Population Based Reinforcement Learning
This work studies how to learn user-controllable and task-relevant diverse policy sets while maintaining similar policy quality.
🧩 Others (CV & GNN)
- [NeurIPS 2022 Spotlight] Self-supervised Amodal Video Object Segmentation
This work proposes a self-supervised video segmentation framework for inferring complete object shapes in occluded scenes using temporal information.
- [NeurIPS 2021] GRIN: Generative Relation and Intention Network for Multi-agent Trajectory Prediction
This work combines conditional generative modeling with graph neural networks to model agent intentions and social relations for multi-agent trajectory prediction.
Academic Services
- Reviewer for NeurIPS, ICLR, and ICML, 2023-present
- Reviewer for AAAI, 2025-present
- Reviewer for IEEE Transactions on Evolutionary Computation (TEVC)
- Reviewer for IEEE Transactions on Cognitive and Developmental Systems (TCDS)
Teaching Assistant
- ENG 2003: Information Technology, The Hong Kong Polytechnic University, Fall 2024
- COMP 6707: Computational Intelligence, The Hong Kong Polytechnic University, Spring 2025
- DSAI 4205: Big Data Analysis, The Hong Kong Polytechnic University, Fall 2025 and Spring 2026
