Browse publications on Google Scholar (top-right) ↗
Resume (English only)
Academic Achievements
[NeurIPS 2025 D&B (Spotlight🌟)] TIME: A Multi-level Benchmark for Temporal Reasoning of LLMs in Real-World Scenario, Super simple reproduction of Deepseek-R1-Zero and Deepseek-R1, using the '24-point game' as an example
Education
Peking University
Background
Interested in reasoning in LLMs, post-training, and interpretability of LLMs. Strive to seek ways to make LLMs generalize in reasoning ability.