AI Paper Atlas

Research map / Multi-agent systems and games / Cooperative multi-agent RL

Vulnerable Agent Identification in Large-Scale Multi-Agent Reinforcement Learning

Simin Li, Zihao Mao, Zheng Yuwei, Linhao Wang, Ruixiao Xu, Chengdong Ma, Zhiqian Liu, Xin Yu, Yuqing Ma, Xin Wang, Jie Luo, Bo An and 3 more

ICML 2026 · Multi-agent systems and games › Cooperative multi-agent RL · approach: Cooperative MARL algorithms · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
use VAI (Bellman-operator value reward) instead of baseline selectorsvulnerable agent selection · Battle/Taxi/Vicsek, 18 taskssystem reward under attack: outperforms all baselines in 17 out of 18 tasksImproved
use degree centrality or PIANO instead of random selectionvulnerable agent selection · Battle/Taxi/Vicseksystem reward under attackNo clear effect
use dense reward from the regularized Bellman value instead of sparse bi-level RL reward (ablation)upper-level selection · Battle/Taxi/Vicseksystem reward under attack: Battle 64/8: 295.09 (Bi-level) -> 281.50 (VAI-RL)Improved
use RL instead of greedy for upper-level selection (ablation)VAI · Battle/Taxi/Vicsek, 18 taskssystem reward under attack: VAI-RL better in 10 of 18 tasksMixed
use the Bellman-operator value as predictor of RL-trained attack resultregularized mean-field Bellman operator · Battle 64 agentsPearson correlation: r = 0.97Improved
add more rollouts for training the value function (ablation)VAI step 1 · pilot studyperformanceNo clear effect

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy