Muyao Li

Muyao Li

First-year Ph.D. student at the Institute for Artificial Intelligence, Peking University

Hi! My name is Muyao Li (李沐遥). I am a first-year Ph.D. student at the Institute for Artificial Intelligence, Peking University, advised by Prof. Yaodong Yang. Previously, I studied in the Tong Class at Yuanpei College, Peking University, and interned with the CraftJarvis team. You can reach me at li_muyao@stu.pku.edu.cn.

My research focuses on building generalized agents that unify perception, reasoning, and decision-making within an end-to-end foundation model. My current work centers on deploying such agents in games.

Research Interests: Machine Learning, Reinforcement Learning, Reasoning, Embodied Intelligence, LLM Agents

🔥 News

  • 2026:  🎉 AdaMAP has been accepted to NeurIPS 2026.
  • 2026:  🎉 DeepHA and OpenHA have been accepted to ICML 2026.
  • 2026:  🎉 CrossAgent has been accepted to CVPR 2026.
  • 2025.05:  🎉🎉 The Paper “JARVIS-VLA: Post-Training Large-Scale Vision Language Models to Play Visual Games with Keyboards and Mouse” are accepted by ACL 2025

📝 Publications

AdaMAP: Learning Adaptive Multi-Action Prediction with Grounded Dreaming Guidance

Muyao Li, Zihao Wang, Xujing Li, Xiangyang Li, Jiabo Ye, Zhiyong Wu, Yaodong Yang, Yujia Qin

NeurIPS 2026 · Paper · Conference

DeepHA: Scaling Action Chains Elicits Deep Hierarchical Agents

Zihao Wang, Muyao Li, Kaichen He, Haowei Lin, Xiaojian Ma, Anji Liu, Yitao Liang

ICML 2026 · Paper · Conference

OpenHA: A Series of Open-Source Hierarchical Agentic Models in Minecraft

Zihao Wang, Muyao Li, Kaichen He, Xiangyu Wang, Zhancun Mu, Minghao Liu, Anji Liu, Yitao Liang

ICML 2026 · Paper · arXiv · Code

Training One Model to Master Cross-Level Agentic Actions via Reinforcement Learning

Kaichen He, Zihao Wang, Muyao Li, Anji Liu, Yitao Liang

CrossAgent · CVPR 2026 · Paper · arXiv · Code

ACL 2025
sym

JARVIS-VLA: Post-Training Large-Scale Vision Language Models to Play Visual Games with Keyboards and Mouse

Muyao Li*, Zihao Wang, Kaichen He, Xiaojian Ma, Yitao Liang

Project

  • The first VLA models in Minecraft that can follow human instructions on over 1k different atomic tasks
  • Introduce a novel approach, Act from Visual Language Post-Training, which refines Visual Language Models (VLMs) through visual and linguistic guidance in a self-supervised manner.

📖 Education

  • Ph.D. student (first year), Institute for Artificial Intelligence, Peking University, Beijing. Advisor: Prof. Yaodong Yang.
  • Undergraduate studies, Yuanpei College (Tong Class), Peking University, Beijing. Enrolled in September 2022.

💼 Experience

  • 2025.11 – Present: Intern, Seed team.