AI Paper Atlas

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

Sparse Topology-Aware Pairwise Scoring for Large-Scale Multi-Agent Reinforcement Learning

Zhibo Deng, Feng Liang, Yong Zhang, Xiaoxi Zhang, Xiping Hu

ICML 2026 · Multi-agent systems and games › Cooperative multi-agent RL · approach: Cooperative MARL algorithms

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
plug SOPS into QMIXQMIX · IMP Correlated, 50 agentsnormalized discounted reward: 18.04 -> 45.59Improved
plug SOPS into QPLEXQPLEX · IMP Correlated, 50 agentsnormalized discounted reward: 20.10 -> 46.26Improved
plug SOPS into SHAQSHAQ · IMP Correlated, 50 agentsnormalized discounted reward: -14.88 -> 32.41Improved
plug SOPS into SHAQSHAQ · IMP Uncorrelated, 50 agentsnormalized discounted reward: -7.82 -> 15.46Improved
plug SOPS into QPLEXQPLEX · IMP OWF, 50 agentsnormalized discounted reward: 60.74 -> 62.93Improved
plug SOPS into QMIXQMIX · IMP OWF, 50 agentsnormalized discounted reward: 61.38 -> 65.46Improved
replace static exponential backbone EC-S with learned sparse selection (SOPS)EC-S · IMP OWF, 100 agentsnormalized discounted reward: 63.76 -> 66.40Improved
replace one-peer exponential schedule EC-O with SOPSEC-O · IMP Correlated, 100 agentsnormalized discounted reward: 18.85 -> 23.55Improved
SOPS under test-time message dropout p=0.5 versus no dropoutSOPS · Battle-100 win ratewin rate: 0.98 -> 0.74Hurt
EC-O under test-time message dropout p=0.5 versus no dropoutEC-O · Battle-100 win ratewin rate: 0.93 -> 0.46Hurt
use linear temperature annealing instead of exponential annealing (ablation)SOPS · Battle 20 and 42 agentswin rateImproved
use linear annealing instead of fixed Gumbel temperature (0.1, 0.5, 1.0, 2.0) (ablation)SOPS · Battle 20 and 42 agentswin rateImproved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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