Tianyi Wang
I work on post-training algorithms for large language models, with a focus on reinforcement learning and reliable optimization.
My current research explores stable and efficient learning for long-horizon reasoning. I am also interested in optimizer design, implementation, and reproducible machine learning systems.
Currently
I have joined ByteDance Seed for research, following my research internship at Tencent Youtu.
Updates
News
- DyJR was accepted to NeurIPS 2026. Congratulations to all the authors! 🎉
- I have joined ByteDance Seed for research after completing my research internship at Tencent Youtu.
- SPPO was selected for an oral presentation at ACL 2026.
- SPPO was accepted to ACL 2026.
- APO was accepted to ICML 2026.
Research
All publications →Selected Works
SPPO: Sequence-Level PPO for Long-Horizon Reasoning
Reformulates long-horizon reasoning as a sequence-level contextual bandit with a decoupled scalar value function to improve stability and reduce memory cost.
DyJR: Preserving Diversity in Reinforcement Learning with Verifiable Rewards via Dynamic Jensen-Shannon Replay
Introduces Dynamic Jensen-Shannon Replay to preserve diversity in reinforcement learning through verifiable rewards mechanism.
Anchored Policy Optimization: Mitigating Exploration Collapse via Support-Constrained Rectification
Introduces support-constrained rectification for RLVR to mitigate exploration collapse and improve both Pass@1 and response diversity.
Background
Education
Beijing University of Posts and Telecommunications
B.S. in Computer Science and Technology
Internships
Experience
Current
ByteDance Seed
Research Intern
Completed
Tencent Youtu
Research Intern