Leitian Tao

I am a Ph.D. student in the Computer Sciences Department at the University of Wisconsin–Madison, advised by Prof. Sharon Li. My research focuses on post-training for agentic systems and reasoning tasks.

I am currently a Research Intern at Microsoft Research, working with Dr. Baolin Peng and Dr. Jianfeng Gao on agentic reinforcement learning and coding-agent self-reflection for scaling.

Email / Google Scholar / Github / LinkedIn / CV

Research

My research focuses on post-training for agentic systems and reasoning tasks, with an emphasis on agentic reinforcement learning, long-horizon credit assignment, self-improving coding agents, and learning beyond binary verifiable rewards. Representative papers are highlighted.

Experience

Microsoft Research — Research Intern
Feb 2026 – Present · Redmond, WA
Mentors: Dr. Baolin Peng and Dr. Jianfeng Gao. Research on agentic RL and coding-agent self-reflection for scaling.

Meta (Facebook AI Research) — Research Scientist Intern
May 2025 – Dec 2025 · Bellevue, WA
Mentors: Dr. Ping Yu and Dr. Jason Weston. Research on improving LLM reasoning beyond binary verifiable rewards.

Adobe Research — Research Scientist Intern
May 2024 – Dec 2024 · San Jose, CA
Mentors: Dr. Xiang Chen and Dr. Tong Yu. Developed CodeLutra, a preference-learning framework that improves code generation using success and failure signals.

Shanghai AI Lab — Research Intern
Oct 2022 – Mar 2023 · Shanghai, China
Mentor: Dr. Peng Gao. Studied fine-tuning methods for multimodal model alignment.

Preprints
	Executable Verification Is All You Need for the Self-Improvement Agentic Loop Leitian Tao, et al. Ongoing
	TRACE: Turn-Level Reward Assignment via Credit Estimation for Long-Horizon Agents Leitian Tao, Baolin Peng, Wenlin Yao, Tao Ge, Hao Cheng, Mike Hang Wang, Jianfeng Gao, Sharon Li Preprint
	The Era of Real-World Human Interaction: RL from User Conversations Chuanyang Jin, Jing Xu, Bo Liu, Leitian Tao, Olga Golovneva, Tianmin Shu, Wenting Zhao, Xian Li, Jason Weston Preprint, 2025 [arXiv]

Publications
	Hybrid Reinforcement: When Reward Is Sparse, It's Better to Be Dense Leitian Tao, Ilia Kulikov, Swarnadeep Saha, Tianlu Wang, Jing Xu, Yixuan Li, Jason Weston, Ping Yu ICLR 2026 [arXiv] [BibTeX] @inproceedings{tao2026hybrid, title = {Hybrid Reinforcement: When Reward Is Sparse, It's Better to Be Dense}, author = {Tao, Leitian and Kulikov, Ilia and Saha, Swarnadeep and Wang, Tianlu and Xu, Jing and Li, Yixuan and Weston, Jason E. and Yu, Ping}, booktitle = {International Conference on Learning Representations (ICLR)}, year = {2026} }
	RESTRAIN: From Spurious Votes to Signals -- Self-Driven RL with Self-Penalization Zhaoning Yu, Will Su, Leitian Tao, Haozhu Wang, Aashu Singh, Hanchao Yu, Jianyu Wang, Hongyang Gao, Weizhe Yuan, Jason Weston, Ping Yu, Jing Xu ICLR 2026 [arXiv] [BibTeX] @inproceedings{yu2026restrain, title = {RESTRAIN: From Spurious Votes to Signals -- Self-Driven RL with Self-Penalization}, author = {Yu, Zhaoning and Su, Will and Tao, Leitian and Wang, Haozhu and Singh, Aashu and Yu, Hanchao and Wang, Jianyu and Gao, Hongyang and Yuan, Weizhe and Weston, Jason and Yu, Ping and Xu, Jing}, booktitle = {International Conference on Learning Representations (ICLR)}, year = {2026} }
	Limited Preference Data? Learning Better Reward Model with Latent Space Synthesis Leitian Tao, Xuefeng Du, Yixuan Li NeurIPS 2025 [arXiv]
	Your Weak LLM is Secretly a Strong Teacher for Alignment Leitian Tao, Yixuan Li ICLR 2025
	CodeLutra: Boosting LLM Code Generation via Preference-Guided Refinement Leitian Tao, Xiang Chen, Tong Yu, Tung Mai, Ryan Rossi, Yixuan Li, Saayan Mitra TMLR 2025
	Non-Parametric Outlier Synthesis Leitian Tao, Xuefeng Du, Xiaojin Zhu, Yixuan Li ICLR 2023
	Predicate Correlation Learning for Scene Graph Generation Leitian Tao, Li Mi, Nannan Li, Xianhang Cheng, Yaosi Hu, Zhenzhong Chen IEEE TIP
	Activate and Reject: Safe Domain Generalization under Category Shift Chaoqi Chen, Luyao Tang, Leitian Tao, Hong-Yu Zhou, Yue Huang, Xiaoguang Han, Yizhou Yu ICCV 2023
	Position: Challenges and Future Directions of Data-Centric AI Alignment Min-Hsuan Yeh, Jeffrey Wang, Xuefeng Du, Seongheon Park, Leitian Tao, Shawn Im, Yixuan Li ICML 2025

Template inspired by Jon Barron.
Last updated: July 2026