Chao Qu
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Chao Qu has a total of 19 co-authors. Here are the top co-authors by number of shared publications:
- Xiaoyu Tan
National University of Singapore
6 shared publications - Xihe Qiu
Shanghai University of Engineering Science
6 shared publications - Bin Li
Shanghai University of Engineering Science, Shanghai, China
2 shared publications - James Zhang
Ant Group Hangzhou China
2 shared publications - Yinghui Xu
Fudan University, Shanghai, China
2 shared publications - Dakuan Lu
Fudan University, Shanghai, China
1 shared publication
Latest Publications
- Open Access
- 20 April 2026
- Published by Association for Computing Machinery (ACM)
- p. 1378-1389
AURORA: Automated Training Framework of Universal Process Reward Models via Ensemble Prompting and Reverse Verification
- 20 April 2026
- Published by Association for Computing Machinery (ACM)
- p. 1378-1389
Open AccessAURORA: Automated Training Framework of Universal Process Reward Models via Ensemble Prompting and Reverse Verification
- Open Access
- 8 May 2025
- Published by Association for Computing Machinery (ACM)
- p. 458-467
Meta-Agent-Workflow: Streamlining Tool Usage in LLMs through Workflow Construction, Retrieval, and Refinement
- 8 May 2025
- Published by Association for Computing Machinery (ACM)
- p. 458-467
Open AccessMeta-Agent-Workflow: Streamlining Tool Usage in LLMs through Workflow Construction, Retrieval, and Refinement
- Open Access
- 21 October 2024
- Published by Association for Computing Machinery (ACM)
- p. 3989-3993
ILTS: Inducing Intention Propagation in Decentralized Multi-Agent Tasks with Large Language Models
Achieving coordination while avoiding suboptimal equilibria poses a significant challenge in decentralized multi-agent reinforcement learning (MARL) systems operating under limited global information. Conventional decentralized approaches have struggled to effectively induce cooperative behaviors between agents. We propose a novel hierarchical framework that synergistically combines large language models (LLMs) and deep reinforcement learning to address this challenge. Our proposed learning-to-share (ILTS) method decomposes the global objective into a two-level hierarchy: high-level LLM policy determines how to share rewards between neighboring agents to shape emergent collaboration, while low-level policies optimize the induced localized objectives using Q value networks. A meta-learning sophisticated dynamic reward-sharing scheme via LLMs is developed to facilitate decentralized cooperation without explicit communication. Experimental results demonstrate that ILTS outperforms prior MARL algorithms across cooperative multi-agent tasks by inducing collaborative strategies and performing intention propagation from lightweight learned signals. This hierarchical framework avoids the need for hand-engineered rewards or explicit communication while promoting scalable learning of intricate symbiotic behaviors between agents perceiving only local observations.
- 21 October 2024
- Published by Association for Computing Machinery (ACM)
- p. 3989-3993
Open AccessILTS: Inducing Intention Propagation in Decentralized Multi-Agent Tasks with Large Language Models
Achieving coordination while avoiding suboptimal equilibria poses a significant challenge in decentralized multi-agent reinforcement learning (MARL) systems operating under limited global information. Conventional decentralized approaches have struggled to effectively induce cooperative behaviors between agents. We propose a novel hierarchical framework that synergistically combines large language models (LLMs) and deep reinforcement learning to address this challenge. Our proposed learning-to-share (ILTS) method decomposes the global objective into a two-level hierarchy: high-level LLM policy determines how to share rewards between neighboring agents to shape emergent collaboration, while low-level policies optimize the induced localized objectives using Q value networks. A meta-learning sophisticated dynamic reward-sharing scheme via LLMs is developed to facilitate decentralized cooperation without explicit communication. Experimental results demonstrate that ILTS outperforms prior MARL algorithms across cooperative multi-agent tasks by inducing collaborative strategies and performing intention propagation from lightweight learned signals. This hierarchical framework avoids the need for hand-engineered rewards or explicit communication while promoting scalable learning of intricate symbiotic behaviors between agents perceiving only local observations.
- 12 March 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Emerging Topics in Computational Intelligence
- Vol. 8 (4) , 2974-2986
Model-Based Off-Policy Deep Reinforcement Learning With Model-Embedding
Model-based reinforcement learning (MBRL) has shown its advantages in sample efficiency over model-free reinforcement learning (MFRL) by leveraging control-based domain knowledge. Despite the impressive results it achieves, MBRL is still outperformed by MFRL due to the lack of unlimited interactions with the environment. While imaginary data can be generated by imagining the trajectories of future states, a trade-off between the usage of data generation and the influence of model bias remains to be resolved. In this paper, we propose a simple and elegant off-policy model-based deep reinforcement learning algorithm with a model embedded in the framework of probabilistic reinforcement learning, called MEMB. To balance the sample-efficiency and model bias, we exploit both real and imaginary data in training. In particular, we embed the model in the policy update and learn value functions from the real data set. We also provide a theoretical analysis of MEMB with the Lipschitz continuity assumption on the model and policy, proving the reliability of the short-term imaginary rollout. Finally, we evaluate MEMB on several benchmarks and demonstrate that our algorithm can achieve state-of-the-art performance.
- 12 March 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Emerging Topics in Computational Intelligence
- Vol. 8 (4) , 2974-2986
Model-Based Off-Policy Deep Reinforcement Learning With Model-Embedding
Model-based reinforcement learning (MBRL) has shown its advantages in sample efficiency over model-free reinforcement learning (MFRL) by leveraging control-based domain knowledge. Despite the impressive results it achieves, MBRL is still outperformed by MFRL due to the lack of unlimited interactions with the environment. While imaginary data can be generated by imagining the trajectories of future states, a trade-off between the usage of data generation and the influence of model bias remains to be resolved. In this paper, we propose a simple and elegant off-policy model-based deep reinforcement learning algorithm with a model embedded in the framework of probabilistic reinforcement learning, called MEMB. To balance the sample-efficiency and model bias, we exploit both real and imaginary data in training. In particular, we embed the model in the policy update and learn value functions from the real data set. We also provide a theoretical analysis of MEMB with the Lipschitz continuity assumption on the model and policy, proving the reliability of the short-term imaginary rollout. Finally, we evaluate MEMB on several benchmarks and demonstrate that our algorithm can achieve state-of-the-art performance.
- Open Access
- 26 November 2023
- Published by Association for Computing Machinery (ACM)
- p. 205-211
Adaptive Learning on User Segmentation: Universal to Specific Representation via Bipartite Neural Interaction
- 26 November 2023
- Published by Association for Computing Machinery (ACM)
- p. 205-211
Open AccessAdaptive Learning on User Segmentation: Universal to Specific Representation via Bipartite Neural Interaction