3 Papers Accepted at AAMAS 2025

Congratulations to the following scientists from A*STAR Centre for Frontier AI Research (A*STAR CFAR) on having their papers accepted at the 24^th International Conference on Autonomous Agents and Multiagent Systems (AAMAS 2025):

Prof Ong Yew Soon, Chief Artificial Intelligence (AI) Scientist and Advisor
Prof Ivor Tsang, Director and Distinguished Principal Scientist
Dr Cheston Tan, Senior Principal Scientist
Dr David Bossens, Senior Scientist
Dr Guo Qing, Senior Scientist
Dr Fan Xiaofeng, Scientist
Dr Lyu Yueming, Scientist
Dr Yu Xingrui, Scientist
Mr Neoh Tzeh Yuan, Engineer

List of accepted papers:

1.	Imitation from Diverse Behaviours: Wasserstein Quality Diversity Imitation Learning with Single-Step Archive Exploration *Xingrui Yu, Zhenglin Wan, David Mark Bossens, Yueming Lyu, Qing Guo, Ivor Tsang* Learning diverse and high-performance behaviours from a limited set of demonstrations is a grand challenge. This work introduces Wasserstein Quality Diversity Imitation Learning (WQDIL), which 1) improves the stability of imitation learning in the quality diversity setting with latent adversarial training based on a Wasserstein Auto-Encoder (WAE), and 2) mitigates a behaviour-overfitting issue using a measure-conditioned reward function with a single-step archive exploration bonus.
2.	Temporal Fair Division of Indisible Items Edith Elkind, Alexander Lam, Mohamad Latifian, Tzeh Yuan Neoh* and Nicholas Teh* This paper studies TEF1 allocations for fair division of indivisible items over time, showing existence through polynomial-time algorithms for several special cases and proving that determining TEF1 existence is computationally intractable in general.
3.	FedRLHF: A Convergence-Guaranteed Federated Framework for Privacy-Preserving and Personalised RLH Flint Xiaofeng Fan, Cheston Tan, *Yew-Soon Ong*, Roger Wattenhofer, Wei-Tsang Ooi FedRLHF introduces a federated approach to reinforcement learning with human feedback (RLHF), aiming to protect user privacy through distributed training while tailoring decision policies to individual participants’ preferences. By leveraging theoretical insights on convergence and personalisation, the framework promises both privacy and efficiency, addressing the heterogeneous nature of user-generated feedback in real-world RLHF scenarios, such as refining LLMs.

Held from 19 – 23 May in Detroit, Michigan, USA, AAMAS is a prestigious conference in the realm of agents and multiagent systems. The internationally renowned high-profile forum provides a platform for researchers and practitioners in all areas of agent technology to discuss the latest developments in the field.

Find out more about AAMAS 2025.

News

3 Papers Accepted at AAMAS 2025

A*STAR celebrates International Women's Day