My name is Chao Yu（于超）. I received my Ph.D. from the Department of Electronic Engineering at Tsinghua University
in 2023. I am currently an Assistant Professor (Distinguished Research Fellow) at the Embodied Decision Intelligence Lab (EDI Lab) at Tsinghua Shenzhen International Graduate School (SIGS)
. I have been selected for the Youth Talent Support Program of the Chinese Institute of Electronics. My research has long focused on reinforcement learning–based decision intelligence. As first author or corresponding author, I have published more than 50 papers in top-tier international conferences and journals, including ICML, NeurIPS, ICLR, CVPR, ECCV, CoRL, IROS, ICRA, TMLR, and RAL, with over 6,000 citations on Google Scholar. My representative works include the multi-agent reinforcement learning algorithm MAPPO, which has received more than 3,000 Google Scholar citations, and RLinf, a large-scale reinforcement learning training framework for embodied intelligence, which has accumulated over 3,000 GitHub stars.
Feel free to reach out if you’d like to discuss research or explore potential collaboration!
📃 Research Interest
RL Infra
- My Technical Preference: Scalable reinforcement learning systems, training infrastructure, and system-algorithm co-design for large-scale policy optimization.
- Representative works on RL Infra include: RLinf and etc. covering efficient RL training, real-world online policy learning, and VLA+RL system design.
Strategic Agent
- My Research Focus: multi-agent RL, strategic reasoning, self-play, cooperation/competition, and language agents.
- Representative works include MAPPO, Fictitious Cross-Play, MARSHAL, WideSeek-R1, Werewolf game etc.
Embodied Agent
- My Application Interest: embodied intelligence with quadrupeds, drones, multi-robot systems, VLA models, and world-model-based robotic training.
- Representative works include WoVR, πRL, World4RL, RoboScape-R, VolleyBots, FlightBench, OmniDrones, and etc.
🔥 News
- 2026.01: 🎉 2 papers (1xfirst, 1xcontribute) are accepted by The Fourteenth International Conference on Learning Representations (ICLR 2026). See you in Rio de Janeiro🇧🇷!
🏫 Educations
-
2019 - 2023: Department of Electronic Engineering, Tsinghua University
.
Ph.D. in Electronic Science and Technology.
Outstanding Doctoral Graduate (Top 5%), Outstanding Doctoral Thesis (Top 10%).
Advisor: Prof. Yu Wang; Co-advisor: Assistant Prof. Yi Wu. -
2016 - 2019: Department of Mechanical Engineering, Tsinghua University
.
M.S. in Mechanical Engineering and Automation.
Outstanding Master’s Thesis (Top 10%).
Advisor: Prof. Xin-Jun Liu. -
2012 - 2016: School of Automation, Beijing Institute of Technology
.
B.S. in Automation.
Outstanding Graduate (Top 15%).
📃 Publications
ACM MM 2026
arXiv 2026
WoVR: World Models as Reliable Simulators for Post-Training VLA Policies with RL
arXiv preprint arXiv:2602.13977
arXiv 2026
Beyond Imitation: Reinforcement Learning-Based Sim-Real Co-Training for VLA Models
arXiv preprint arXiv:2602.12628
arXiv 2026
RLinf-USER: A Unified and Extensible System for Real-World Online Policy Learning in Embodied AI
arXiv preprint arXiv:2602.07837
Preprint 2026
WideSeek-R1: Exploring Width Scaling for Broad Information Seeking via Multi-Agent Reinforcement Learning
Preprint (2026)
Preprint 2025
RoboScape-R: Unified Reward-Observation World Models for Generalizable Robotics Training via RL
Preprint (2025)
ACL 2025
Preprint 2025
IROS 2025
Long-horizon Locomotion and Manipulation on a Quadrupedal Robot with Large Language Models
IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS 2025)
ICLR 2026
MARSHAL: Incentivizing Multi-Agent Reasoning via Self-Play with Strategic LLMs
The Fourteenth International Conference on Learning Representations (ICLR 2026)
arXiv 2025
RLinf-VLA: A Unified and Efficient Framework for VLA+RL Training
arXiv preprint arXiv:2510.06710
ICLR 2026
SAC Flow: Sample-Efficient Reinforcement Learning of Flow-Based Policies via Velocity-Reparameterized Sequential Modeling
ICLR 2026
ICLR 2026
RE-PO: Robust Enhanced Policy Optimization as a General Framework for LLM Alignment
The Fourteenth International Conference on Learning Representations (ICLR 2026)
arXiv 2025
JuggleRL: Mastering Ball Juggling with a Quadrotor via Deep Reinforcement Learning
arXiv preprint arXiv:2509.24892
arXiv 2025
World4RL: Diffusion World Models for Policy Refinement with Reinforcement Learning for Robotic Manipulation
arXiv preprint arXiv:2509.19080
arXiv 2025
arXiv 2025
D3P: Dynamic Denoising Diffusion Policy via Reinforcement Learning
arXiv preprint arXiv:2508.06804 (2025)
EMNLP 2025
Spec-VLA: Speculative Decoding for Vision-Language-Action Models with Relaxed Acceptance
EMNLP 2025 Main Conference
RA-L 2025
Online Planning for Multi-UAV Pursuit-Evasion in Unknown Environments Using Deep Reinforcement Learning
IEEE Robotics and Automation Letters (2025)
arXiv 2025
arXiv 2025
VS-Bench: Evaluating VLMs for Strategic Reasoning and Decision-Making in Multi-Agent Environments
ICLR 2026 Oral
NeurIPS 2025
NeurIPS 2025
CoRL 2025
Toward Real-World Cooperative and Competitive Soccer with Quadrupedal Robot Teams
Conference on Robot Learning (CoRL 2025)
Preprint 2025
Fine-tuning Diffusion Policies with Backpropagation Through Diffusion Timesteps
Preprint (2025)
CoRL 2025
Preprint 2025
Hysteresis-Aware Neural Network Modeling and Whole-Body Reinforcement Learning Control of Soft Robots
Preprint (2025)
AAAI 2026 Bridge
Survey 2025
Multi-Robot System for Cooperative Exploration in Unknown Environments: A Survey
Survey Paper (2025)
ICLR 2026
Translate Policy to Language: Flow Matching Generated Rewards for LLM Explanations
The Fourteenth International Conference on Learning Representations (ICLR 2026)
ICML 2025
Learning Strategic Language Agents in the Werewolf Game with Iterative Latent Space Policy Optimization
Proceedings of the 42nd International Conference on Machine Learning (ICML 2025)
NeurIPS 2025 D&B
ICML 2025
JMLR 2025
Learning Global Nash Equilibrium in Team Competitive Games with Generalized Fictitious Cross-Play
Journal of Machine Learning Research, 26 (2025), 1–30
PACM IMWUT 2024
SleepNetZero: Zero-Burden Zero-Shot Reliable Sleep Staging with Neural Networks Based on Ballistocardiograms
Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies (PACM IMWUT 2024)
RA-L 2025
IROS 2025
ICRA 2025
Preprint 2024
Survey 2024
RA-L 2025
FlightBench: A Comprehensive Benchmark of Spatial Planning Methods for Quadrotors
IEEE Robotics and Automation Letters (RA-L 2025)
Preprint 2024
CityLight: A Universal Model Towards Real-world City-scale Traffic Signal Control Coordination
Preprint (2024)
ICRA 2024
LAGOON: Language-Guided Motion Control
IEEE International Conference on Robotics and Automation (ICRA 2024)
ICML 2024 Oral
AAAI 2024
RA-L 2024
NeurIPS 2024
Sharing Minds during MARL Training for Enhanced Cooperative LLM Agents
The Thirty-eighth Conference on Neural Information Processing Systems (NeurIPS 2024)
ICLR 2025
Few-shot In-context Preference Learning using Large Language Models
The Thirteenth International Conference on Learning Representations (ICLR 2025)
AAMAS 2024
LLM-Powered Hierarchical Language Agent for Real-time Human-AI Coordination
Proc. of the 23rd International Conference on Autonomous Agents and Multiagent Systems (AAMAS 2024)
RA-L 2024
Preprint 2023
ICML 2024
Language Agents with Reinforcement Learning for Strategic Play in the Werewolf Game
Proceedings of the 41st International Conference on Machine Learning (ICML 2024)
AAMAS 2023
Fictitious Cross-Play: Learning Global Nash Equilibrium in Mixed Cooperative-Competitive Games
Proceedings of the 22nd International Conference on Autonomous Agents and Multiagent Systems (AAMAS 2023)
IJCAI 2023
ICLR 2023
Learning Zero-Shot Cooperation with Humans, Assuming Humans Are Biased
The Eleventh International Conference on Learning Representations (ICLR 2023)
AAMAS 2023
Asynchronous Multi-Agent Reinforcement Learning for Efficient Real-Time Multi-Robot Cooperative Exploration
Proceedings of the 22nd International Conference on Autonomous Agents and Multiagent Systems (AAMAS 2023)
AAMAS 2023
Learning Graph-Enhanced Commander-Executor for Multi-Agent Navigation
Proceedings of the 22nd International Conference on Autonomous Agents and Multiagent Systems (AAMAS 2023)
NeurIPS 2022 D&B
ROBIO 2022
A Benchmark of Planning-based Exploration Methods in Photo-Realistic 3D Simulator
IEEE International Conference on Robotics and Biomimetics (ROBIO 2022)
ECCV 2022
ICIP 2022
SAVE: Spatial-Attention Visual Exploration
IEEE International Conference on Image Processing (ICIP 2022)
CoG 2022
VMAPD: Generate Diverse Solutions for Multi-Agent Games with Recurrent Trajectory Discriminators
IEEE Conference on Games (CoG 2022)
ICML 2022
Preprint 2021
Multi-Agent Vulnerability Discovery for Autonomous Driving with Hazard Arbitration Reward
Preprint (2021)
CAAI 2021
Unlocking the Potential of MAPPO with Asynchronous Optimization
CAAI International Conference on Artificial Intelligence (2021)
ICLR 2021
DAC 2020
INCA: INterruptible CNN Accelerator for Multi-tasking in Embedded Robots
57th ACM/IEEE Design Automation Conference (DAC 2020)
IPDPSW 2020
CNN-based Monocular Decentralized SLAM on Embedded FPGA
IEEE International Parallel and Distributed Processing Symposium Workshops (IPDPSW 2020)
FCCM 2020
CNN-based Feature-point Extraction for Real-time Visual SLAM on Embedded FPGA
IEEE 28th Annual International Symposium on Field-Programmable Custom Computing Machines (FCCM 2020)
ICCRT 2019
Long-Sighted Imitation Learning for Partially Observable Control
Proceedings of the 2019 2nd International Conference on Control and Robot Technology (ICCRT 2019)
ROBIO 2019
A DenseNet Feature-based Loop Closure Method for Visual SLAM System
2019 IEEE International Conference on Robotics and Biomimetics (ROBIO 2019)
Workshop 2019
Learning Safety-Aware Policy with Imitation Learning for Context-Adaptive Navigation
Workshop Paper / Technical Report (2019)
IROS 2018
DS-SLAM: A Semantic Visual SLAM towards Dynamic Environments
IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS 2018)
🏆 Awards
- 2024: China Postdoctoral Excellent Special Foundation (Top 1,000 nationwide), Chinese Postdoctoral Science Foundation (CPSF).
- 2024: Postdoctoral Fellowship Program (Top 3,000 nationwide), Chinese Postdoctoral Science Foundation (CPSF).
- 2024: Runner-up for Outstanding Doctoral Thesis (Top 5), Chinese Intelligent Agent and Multi-Agent Systems.
- 2023: Shuimu Scholar Program, Tsinghua University.
- 2023: Chuanxin Future Scholar Program, Department of Electronic Engineering, Tsinghua University.
- 2023: Zhang Keqian Postdoctoral Fellowship, Department of Electronic Engineering, Tsinghua University.
- 2023: Outstanding Doctoral Thesis (Top 10%), Tsinghua University.
- 2023: Outstanding Doctoral Graduate (Top 5%), Tsinghua University.
- 2019: Outstanding Master’s Thesis (Top 10%), Tsinghua University.
- 2019 - 2023: First-Class Scholarship (3 times), Tsinghua University.
- 2015: National Scholarship, China Ministry of Education.
🎤 Talks
Nanjing, China🇨🇳
RLinf: A Highly Flexible Reinforcement Learning Post-Training Framework for Embodied Intelligence
智汇金陵 AI开源人才峰会暨魔搭开发者大会 2026.03.22 13:30 - 18:00
👓 Projects
- Projects: [1] 基于深度强化学习的多无人机追逃博弈决策和控制关键技术研究，国家自然科学基金委，青年科学基金项目（C类）, 2025-2027.
- Projects: [2] 多机协同高效机器学习系统研究，国家自然科学基金-中德合作交流基金, 2021-2025.
- Projects: [3] 具有强推理能力的大语言模型智能体关键技术研究，中国博士后基金特别资助, 2023-2025.
💼 Work Experience
- 2026.01 - Present: Assistant Professor, Tsinghua University.
- 2023.07 - 2025.12: Postdoctoral Researcher, Tsinghua University.
🙋 Recruitment
We are actively recruiting!
We are looking for Ph.D. and Master students, Postdocs at Tsinghua University, Ph.D. students of the Joint Program of Zhongguancun Academy and Tsinghua University and Undergraduate Interns with strong interests and motivation to work on frontier research topics including:
Candidates with hands-on systems building abilities and mathematical background are highly encouraged.
🔗 Bond
EE, THU
Energy Efficient Computing Group
Nanoscale Integrated Circuits and System Lab, Energy Efficient Computing Group (NICS-EFC) is leaded by Professor Yu Wang, in Electronic Engineering Department, Tsinghua University. The group is committed to the research of energy-efficient circuits and systems design methodology towards the Artificial Intelligence (AI) scenario: Multi-agent Reinforcement Learning Algorithm, Efficient and Robust DL system, Domain Specific Acceleration, and Multi-agent system. Each direction is headed by a research associate.