I am a final-year Ph.D. candidate in Computer Science at Fudan University (expected June 2026), working with
Prof. Xipeng Qiu and
Prof. Xuanjing Huang at the
FudanNLP Group.
I received my B.S. in Data Science from East China Normal University in 2021.
My research focuses on building more reliable and capable reasoning systems for large language models. Key directions include:
◈ LLM Reasoning — uncertainty-aware guidance, cross-model communication, and hierarchical answer aggregation for complex multi-step reasoning. ◈ Reward Design — dynamic, generalizable process reward models with fine-grained multi-dimensional criteria for robust supervision. ◈ Evaluation — adaptive metrics for test-time scaling, error classification frameworks, and self-knowledge assessment of LLMs.
News
2026.01Started internship at Tencent HY, working on reward design.
2025.09Interned at Optiver AI as an Alpha Researcher (Sep–Nov 2025).
20252 papers accepted to ACL 2025: DG-PRM and financial bias in LLMs.
2024.12Awarded the National Scholarship (国家奖学金).
2024Papers accepted to ACL, CVPR, EMNLP, LREC-COLING, NAACL, ICML, COLM, AAAI in 2024–2025.
Selected Publications
* denotes equal contribution. See full list on
Google Scholar.
First Author
Do Large Language Models Know What They Don't Know?
Introduced SelfAware, a dataset and automated methodology to evaluate LLMs' ability to recognize unanswerable questions, revealing intrinsic self-knowledge capabilities.
Exchange-of-Thought: Enhancing LLM Capabilities through Cross-Model Communication
Introduced UAG, a plug-and-play approach that dynamically identifies uncertainty signals and intervenes by retracting to reliable states with certified reasoning clues.
Explicit Memory Learning with Expectation Maximization
Proposed DG-PRM with reward trees for fine-grained multi-dimensional criteria and Pareto dominance estimation, achieving superior cross-domain generalization.
Aggregation of Reasoning: A Hierarchical Framework for Enhancing Answer Selection in LLMs
