Shuqing Shi 史述青

Email: shuqing.shi[at]kcl[dot]ac[dot]uk

I am a Ph.D. student in Computer Science at King's College London, supervised by Dr. Yali Du. Previously, I received my M.Sc. and B.Eng. from the University of Electronic Science and Technology of China (UESTC).

My research focuses on building AI agents that can effectively coordinate and cooperate in complex multi-agent environments. I'm particularly interested in the fundamental question: How can heterogeneous agents learn to cooperate in a complex environment in an efficient and scalable way?

Research Interests

Multi-Agent Reinforcement Learning: Developing algorithms for coordination and cooperation in complex environments.
Cooperative Game Theory: Understanding coalition formation and value distribution among agents.
Causal Inference: Applying counterfactual reasoning for better credit assignment and generalization.

📚 Publications

* Equal contribution

Preprints

Under Review

Resolving Complex Social Dilemmas by Aligning Preferences with Counterfactual Regret

Shuqing Shi, Yudi Zhang, Joel Z. Leibo, Yali Du

arXiv

We propose a novel approach to resolve complex social dilemmas by aligning agent preferences using counterfactual regret minimization. Our method enables agents to learn cooperative strategies in mixed-motive scenarios where traditional approaches fail.

2026

ICLR 2026

BRIDGE: Bi-level Reinforcement Learning for Dynamic Group Structure in Coalition Formation Games

Shuqing Shi, Nam Phuong Tran, Hao Liang, Debmalya Mandal, Long Tran-Thanh, Yali Du

In The Thirteenth International Conference on Learning Representations (ICLR), 2026

Paper

Coalition formation is fundamental to multi-agent cooperation, yet existing approaches typically treat it as a static problem. We propose BRIDGE, a bi-level reinforcement learning framework that jointly learns when to form coalitions (high-level) and how to coordinate within them (low-level).

@inproceedings{shi2026bridge, title={BRIDGE: Bi-level Reinforcement Learning for Dynamic Group Structure in Coalition Formation Games}, author={Shi, Shuqing and Tran, Nam Phuong and Liang, Hao and Mandal, Debmalya and Tran-Thanh, Long and Du, Yali}, booktitle={The Thirteenth International Conference on Learning Representations}, year={2026} }

ICLR 2026

SocialJax: An Evaluation Suite for Multi-agent Reinforcement Learning in Sequential Social Dilemmas

Zihao Guo*, Shuqing Shi*, Richard Willis, Tristan Tomilin, Joel Z. Leibo, Yali Du

In The Thirteenth International Conference on Learning Representations (ICLR), 2026

Paper

We present SocialJax, a JAX-based evaluation suite for multi-agent reinforcement learning in sequential social dilemmas. Built for speed, SocialJax runs thousands of episodes in seconds on a single GPU while providing standard implementations of classic social dilemmas.

@inproceedings{guo2026socialjax, title={SocialJax: An Evaluation Suite for Multi-agent Reinforcement Learning in Sequential Social Dilemmas}, author={Guo, Zihao and Shi, Shuqing and Willis, Richard and Tomilin, Tristan and Leibo, Joel Z. and Du, Yali}, booktitle={The Thirteenth International Conference on Learning Representations}, year={2026} }

2025

NeurIPS 2025 🌟 Spotlight (Top 3%)

Causality Meets Locality: Provably Generalizable and Scalable Policy Learning for Networked Systems

Hao Liang*, Shuqing Shi*, Yudi Zhang, Biwei Huang, Yali Du

In The Thirty-Ninth Annual Conference on Neural Information Processing Systems (NeurIPS), 2025

Paper

We propose a principled approach that combines causal inference with locality principles for provably generalizable and scalable policy learning in networked systems. Our method achieves state-of-the-art performance while providing theoretical guarantees.

@inproceedings{liang2025causality, title={Causality Meets Locality: Provably Generalizable and Scalable Policy Learning for Networked Systems}, author={Liang, Hao and Shi, Shuqing and Zhang, Yudi and Huang, Biwei and Du, Yali}, booktitle={The Thirty-Ninth Annual Conference on Neural Information Processing Systems}, year={2025} }

NeurIPS D&B 2025

Evaluating Generalization Capabilities of LLM-Based Agents in Mixed-Motive Scenarios Using Concordia

Chandler Smith et al., Shuqing Shi et al.

In NeurIPS Datasets and Benchmarks Track, 2025

Paper

We evaluate the generalization capabilities of LLM-based agents in mixed-motive scenarios using the Concordia framework, providing insights into how large language models perform in multi-agent social interactions.

@inproceedings{smith2025evaluating, title={Evaluating Generalization Capabilities of LLM-Based Agents in Mixed-Motive Scenarios Using Concordia}, author={Smith, Chandler and others}, booktitle={NeurIPS Datasets and Benchmarks Track}, year={2025} }

2024

NeurIPS 2024

Learning the Expected Core of Strictly Convex Stochastic Cooperative Games

Nam P. Tran, Shuqing Shi, Debmalya Mandal, Yali Du, Long Tran-Thanh

In The Thirty-Eighth Annual Conference on Neural Information Processing Systems (NeurIPS), 2024

Paper

We study the problem of learning the expected core in strictly convex stochastic cooperative games, providing theoretical analysis and practical algorithms for stable coalition value distribution under uncertainty.

@inproceedings{tran2024learning, title={Learning the Expected Core of Strictly Convex Stochastic Cooperative Games}, author={Tran, Nam P. and Shi, Shuqing and Mandal, Debmalya and Du, Yali and Tran-Thanh, Long}, booktitle={The Thirty-Eighth Annual Conference on Neural Information Processing Systems}, year={2024} }

2022

IJCNN 2022

Solving Poker Games Efficiently: Adaptive Memory Based Deep Counterfactual Regret Minimization

Shuqing Shi, Xiaobin Wang, Dong Hao, Zhiyou Yang, Hong Qu

In International Joint Conference on Neural Networks (IJCNN), 2022

Paper

We propose an adaptive memory-based approach to deep counterfactual regret minimization for solving large-scale poker games more efficiently, reducing memory requirements while maintaining solution quality.

@inproceedings{shi2022solving, title={Solving Poker Games Efficiently: Adaptive Memory Based Deep Counterfactual Regret Minimization}, author={Shi, Shuqing and Wang, Xiaobin and Hao, Dong and Yang, Zhiyou and Qu, Hong}, booktitle={International Joint Conference on Neural Networks}, year={2022} }

Jan 2026	Two papers accepted to ICLR 2026: BRIDGE and SocialJax!
Oct 2025	Recognized as NeurIPS 2025 Top Reviewer (Top 8–10%)! 🎉
Sep 2025	Our paper on causal policy learning received NeurIPS 2025 Spotlight (Top 3%)! 🌟
Sep 2025	Paper on LLM-based agents accepted to NeurIPS 2025 Datasets & Benchmarks Track.
Sep 2024	Paper on stochastic cooperative games accepted to NeurIPS 2024.

Shuqing Shi 史述青

Research Interests

📰 News

📚 Publications

Resolving Complex Social Dilemmas by Aligning Preferences with Counterfactual Regret

BRIDGE: Bi-level Reinforcement Learning for Dynamic Group Structure in Coalition Formation Games

SocialJax: An Evaluation Suite for Multi-agent Reinforcement Learning in Sequential Social Dilemmas

Causality Meets Locality: Provably Generalizable and Scalable Policy Learning for Networked Systems

Evaluating Generalization Capabilities of LLM-Based Agents in Mixed-Motive Scenarios Using Concordia

Learning the Expected Core of Strictly Convex Stochastic Cooperative Games

Solving Poker Games Efficiently: Adaptive Memory Based Deep Counterfactual Regret Minimization

💼 Experience

The Chinese University of Hong Kong, Shenzhen

Mohamed bin Zayed University of Artificial Intelligence (MBZUAI)

🏆 Awards

🎓 Academic Services