Fengshuo Bai

Portrait of Fengshuo Bai

Ph.D. Student

Department of Computer Science and Engineering
Shanghai Jiao Tong University

My research spans dexterous manipulation, preference-based RL, and AI alignment, aiming to build reliable systems that learn from complex human objectives.

Dexterous Manipulation Preference-based RL AI Alignment

Email: fengshuobai [@] sjtu [DOT] edu [DOT] cn

Biography

I am a Ph.D. student in the Department of Computer Science and Engineering at Shanghai Jiao Tong University and a member of PAIR Lab. I am co-advised by Prof. Yaodong Yang and Prof. Ying Wen, and I am also part of the Wen-Tsun Wu AI Honorary Doctoral Class advised by Prof. Cewu Lu.

My research focuses on dexterous manipulation, preference-based reinforcement learning, and AI alignment, with the goal of building effective learning systems that align with complex human preferences. I am always open to collaboration and discussion on shared interests.

Education

Ph.D. in Computer Science and Engineering
Shanghai Jiao Tong University, 2023 – Present
Advised by Prof. Yaodong Yang & Prof. Ying Wen
M.S. in Control Theory and Control Engineering
University of Chinese Academy of Sciences, 2020 – 2023
B.Eng. in Computer Science and Engineering
Southeast University, 2016 – 2020

Experience

Institute for AI, Peking University
Research Intern, Jan 2022 – Oct 2022
Worked on RLHF with Dr. Yali Du and Dr. Yaodong Yang.
PKU-PsiBot Joint Lab
Research Intern, Aug 2024 – Feb 2026
Focused on Sim2Real transfer and Vision-Language-Action models.
Zhongguancun Academy
Student Project Leader, Sep 2024 – Jun 2027
Leading the AI Chemist project.

Honors & Awards

Innovation & Excellence Scholarship
Zhongguancun Academy, 2025
China National Scholarship
Ministry of Education, 2017
President Scholarship
Southeast University, 2018
Outstanding Student Leader
Jiangsu Province, 2019
Pacemaker to Merit Student
Southeast University, 2018

Competitions

Silver Prize (25 / 1178)
Kaggle Lux AI Challenge, 2021
3rd (Intro) & 7th (Research) Track
NeurIPS MineRL Diamond Competition, 2021
14th Place
IJCAI Mahjong AI Competition, 2021

News

2026.01 PoliCon was accepted to ICLR 2026.
2025.09 ValueDCG was accepted to the NeurIPS 2025 Workshop on Regulatable ML.
2025.09 Two papers were accepted to NeurIPS 2025, with one selected as Spotlight.
2025.08 AdaptFlow was accepted to EMNLP 2025.
2025.07 We released our VLA survey, A Survey on Vision-Language-Action Models.
2025.06 One paper was accepted for oral presentation at AGI 2025.

Selected Publications

* denotes equal contribution. Showing first-, co-first-, second-, and third-author papers by default.

PoliCon: Evaluating LLMs on Achieving Diverse Political Consensus Objectives

Zhaowei Zhang, Xiaobo Wang, Minghua Yi, Mengmeng Wang, Fengshuo Bai, Zilong Zheng, Yipeng Kang, Yaodong Yang

International Conference on Learning Representations (ICLR), 2026

Paper ArXiv Website
AdaptFlow: Adaptive Workflow Optimization via Meta-Learning

Runchuan Zhu, Bowen Jiang, Lingrui Mei, Fangkai Yang, Lu Wang, Haoxiang Gao, Fengshuo Bai, Pu Zhao, Qingwei Lin, Saravan Rajmohan, Dongmei Zhang

Conference on Empirical Methods in Natural Language Processing (EMNLP), 2025

Paper
GRAIT: Gradient-Driven Refusal-Aware Instruction Tuning for Effective Hallucination Mitigation

Runchuan Zhu, Xinke Jiang, Jiang Wu, Zhipeng Ma, Jiahe Song, Fengshuo Bai, Dahua Lin, Lijun Wu, Conghui He

NAACL Findings, 2025

Paper ArXiv

Academic Service

Conference Reviewer

NeurIPS2024–2025 ICLR2025–2026 ICML2025–2026 AAAI2025–2026 IJCAI2025 AAMAS2024–2026 ECCV2026 IROS2025 DAI2023–2024

Visitors