Welcome to Cooperative AI Lab website! Our research aim is to enable machines to exhibit cooperative and safe behavior in intelligent decision making tasks. To this end, we work on multi-agent cooperation, human-AI coordination, and scalable alignment, which includes human value learning, safety, and ethics.
Active openings
PostDoc Position: One PostDoc position available in Multi-Agent Reinforcement Learning and Foundation Models. Feel free to get in touch if you are interested.
PhD Positions:
We normally recruit 2~3 PhD students each year. If you are interested in joining us, please complete this form and email (yali.du AT kcl.ac.uk). See the form for requirements.
News
16 Oct 2025
Four papers are accepted to EMNLP 2025! We explored Chain-of-Thought, training LLM via verifier-free RL, embodied manipulation, and spiral of silence in LLM agents
16 Oct 2025Five papers are accepted to NeurIPS 2025! We explored CoT Reasoning, Counterfactuals for Language Model Agents, Design Policy Learning, Generalizable and Scalable Policy Learning, and Evaluating Generalization Capabilities of LLM-Based Agents.
18 Mar 2025🚀 We’ve released SocialJax — a suite of sequential social dilemma environments for multi-agent RL in JAX, developed together with DeepMind researcher Joel Leibo. Built for speed, SocialJax leverages JAX’s GPU/TPU acceleration to scale MARL research. Code is available here
5 Dec 2024Welcome Dr Hao Liang to join our group as a postdoc! He will work on causality-inspired multi-agent RL.
1 Oct 2024Welcome Zihao Guo to join our group as a PhD!
27 Sep 2024Our project on Discipline hop for high-fidelity, high-generalisation models of human behaviour is funded by EPSRC, paternering with Matteo Leonetti (KCL) and Gustav Markkula from University of Leeds.
26 Sep 2024Five papers are accepted to NeurIPS 2025! We explored RL for cooperative games and LLMs for werewolves, football, human instruction following!
6 Sep 2024Exciting news! Nature Machine Intelligence has featured our work on decentralized multi-agent reinforcement learning for large-scale networked control, on cracking the scaling challenge of MARL. Welcome to check it out here!
1 Sep 2024Exciting news! Our review on safe RL entitled A review of safe reinforcement learning: Methods, theory and applications has been accepted to IEEE TPAMI. Congrats to Shangding! Read the preprint here!
2 May 2024One paper is accepted to ICML 2024. We explore Zero-shot Cross-task Preference Alignment for Robotic Manipulation. Earlier version appears on NeurIPS 2023 workshop on Optimal Transport.
27 Mar 2024Yali Du has been invited to serve as an Area Chair for NeurIPS 2024.
01 Mar 2024Yali Du has joined the NeurIPS organizing committee as a Communications Chair for NeurIPS 2024.
01 Mar 2024Yali Du is appointed as a Turing Fellow at The Alan Turing Institute.
06 Feb 2024One PostDoc position available in Multi-Agent Reinforcement Learning and Foundation Models. Deadline: 12 Mar 2024. Please apply here.