Yao Tang

Yao Tang 唐尧

Hi there! I am a 2nd-year CIS PhD at University of Pennsylvania, working on agentic reasoning and reinforcement learning.

I'm fortunate to be advised by Prof. Jiatao Gu. I received my bachelor's degree from Shanghai Jiao Tong University. I also had the privilege of working closely with Dr. Li Dong and Prof. Kai-Wei Chang.

News

[2026.10] ProgramLab is released! It offers 1,261 RL environments where agents rebuild real-world programs, for both single-agent and multi-agent RL. Code and dataset are available now!

[2026.01] Multiplex Thinking is released! Paper, code, and checkpoints are available now!

[2025.08] Start my Ph.D. journey at Penn! 🎉

[2025.06] Reinforcement Pre-Training is released!

[2025.05] 🎉 QLASS is accepted by ICML 2025! 🎉

[2024.09] 🎉 Our work CurrMask is accepted by NeurIPS 2024! See you in Vancouver! 🎉

Selected Publications

	ProgramLab: Train Single and Multi Agents to Rebuild Programs with RL Yao Tang+, Daixuan Cheng+, Li Dong, Zhihao Zhan, Shaohan Huang, Jiatao Gu, Furu Wei (+core contributor) Blog Post, 2026. Blog \| Code \| Dataset
	Multiplex Thinking: Reasoning via Token-wise Branch-and-Merge Yao Tang, Li Dong, Yaru Hao, Qingxiu Dong, Furu Wei, Jiatao Gu Preprint, 2026. Project \| Paper \| Code \| Checkpoints \| Post
	Reinforcement Pre-Training Qingxiu Dong, Li Dong, Yao Tang, Tianzhu Ye, Yutao Sun, Zhifang Sui, Furu Wei (equal contribution) Preprint, 2025.* Paper \| Post
	QLASS: Boosting Language Agent Inference via Q-Guided Stepwise Search Zongyu Lin, Yao Tang, Xingcheng Yao, Da Yin, Ziniu Hu, Yizhou Sun, Kai-Wei Chang (equal contribution) ICML 2025.* Paper \| Code \| Checkpoints \| Post
	Learning Versatile Skills with Curriculum Masking Yao Tang, Zhihui Xie, Zichuan Lin, Deheng Ye, Shuai Li (equal contribution) NeurIPS 2024.* Paper \| Code

Research Interests 🧋

I am interested in building generalized AI agents under limited supervision. Take developing a 🧋Boba-Agent that can make all kinds of boba as an example.

🤔To train a Boba-Agent, the most straightforward way is to collect a bunch of boba expert demos and train the agent via imitation learning. But data collection is too costly and the model can poorly generalize to making 🥭mango pomelo sago if the data is on 🍋lemon tea.

🤔Another choice is to allow the Boba-Agent learning from a large amount of trial-and-errors. The agent can keep exploring the kitchen, making different boba, receiving my reward, and improving its boba-making policy via reinforcement learning. Unfortunately, this process is friendly for neither me nor the kitchen.

😀The ideal case is that the agent can firstly search online and learn from offline video demos and textual instructions, extract reusable knowledge, and learn to safely navigate the kitchen, and quickly adapt to personal preferences. (For instance, learning that my highest praise is reserved for boba that's "not too sweet"!)

To summarize, my interests lie in how to enable effective internal world modeling to help the agent learn from large-scale boba-making offline data and extract reusable knowledge, as well as external world interaction, e.g., searching for new recipes online, exploring the kitchen, and getting feedback from me, so that it can quickly make delicious newly invented Boba under limited supervision.

Experiences

	Microsoft Research Asia 2025.03 ~ 2025.08 Research Internship, General Artificial Intelligence Group, Beijing, China Advisor: Dr. Li Dong
	University of California, Los Angeles 2024.07 ~ 2024.10 Los Angeles, CA, USA Visiting Student, working with Johnson Lin Advisor: Prof. Kai-Wei Chang
	Shanghai Jiao Tong University 2020.09 ~ 2025.03 Shanghai, China B.Eng. in Computer Science and Technology Student Researcher (2023.01 ~ 2024.05), working with Zhihui Xie Advisor: Prof. Shuai Li

Misc.

I enjoy jazz dance and am a beginner at guitar.

I am a fan of David Tao.

My family dog Happy loves to play with humans through RL. Here is a video when she was 3 months old.