Nucleolus Credit Assignment for Effective Coalitions in Multi-agent Reinforcement Learning Conference: Proc. of the 24th International Conference on Autonomous Agents and Multiagent Systems (AAMAS 2025); May 19 – 23, 2025; Detroit, Michigan, USA <<fp0207>>

Yugu Li, Zehong Cao, Jianglin Qiao, Siyi Hu · 2025

Read Nucleolus Credit Assignment for Effective Coalitions in Multi-agent Reinforcement Learning Conference: Proc. of the 24th International Conference on Autonomous Agents and Multiagent Systems (AAMAS 2025); May 19 – 23, 2025; Detroit, Michigan, USA <<fp0207>> by Yugu Li, Zehong Cao, Jianglin Qiao, Siyi Hu (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2503-00372.txt.

Description

Yugu Li, Zehong Cao, Jianglin Qiao, and Siyi Hu assign credit in multi-agent learning via game theory in a 2025 paper. Cooperative agents usually form one grand coalition for a composite task. That single-team habit often yields suboptimal performance. Fair credit could let agents split into effective subteams on their own.

Their nucleolus-based method grounds credit in cooperative game theory. A nucleolus Q-learning rule assigns fair shares to each agent. The operator carries guarantees for convergence and coalition stability. Predator-prey and StarCraft scenarios tested the approach across difficulties. Multiple effective coalitions emerged during training. Learning ran faster with wins and rewards above four baselines. Hard and super-hard settings showed the largest advantages. Interpretability of the formed teams is stressed throughout. Composite tasks needing subteams are the target class. The paper bridges game theory and deep multi-agent practice.

Catalog record

CREATORYugu Li, Zehong Cao, Jianglin Qiao, Siyi Hu
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE41 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2503-00372

Browse the full library · About the Tower of Babel