AI Paper Atlas

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

Learning Decentralized LLM Collaboration with Multi-Agent Actor Critic

Shuo Liu, Tianle Chen, Ryan Amiri, Christopher Amato

ICML 2026 · Multi-agent systems and games › Cooperative multi-agent RL · approach: Offline and networked MARL · 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 a centralized critic instead of per-agent decentralized critics (ablation)CoLLM · CoopHEpass rate (%): 59.1 -> 75.2Improved
use a centralized critic instead of per-agent decentralized critics (ablation)CoLLM · StrBuild (n=2)IoU (%): 44.6 -> 68.5Improved
use a centralized critic instead of per-agent decentralized critics (ablation)CoLLM · HouseBuild (n=2)HP (%): 43.8 -> 86.4Improved
replace MA-REINFORCE (MAGRPO) with centralized-critic actor-criticCoLLM-CC · CoopHEpass rate (%): 74.3 -> 75.2No clear effect
replace MA-REINFORCE (MAGRPO) with centralized-critic actor-criticCoLLM-CC · StrBuild (n=2)IoU (%): 50.6 -> 68.5Improved
replace MA-REINFORCE (MAGRPO) with centralized-critic actor-criticCoLLM-CC · TLDRscore (%): 93.5 -> 95.2No clear effect
fine-tune agents with MARL instead of prompt-only multi-agentMAGRPO vs Parallel · TLDRscore (%): 22.9 -> 93.5Improved
fine-tune single model with GRPOQwen model · CoopHEpass rate (%): 56.3 -> 61.8Improved
add a critic to single-model GRPO (AC) (ablation)Single model · CoopHEpass rate (%): 61.8 -> 62.5No clear effect
increase agents from 2 to 3 (ablation)CoLLM-CC · StrBuildIoU (%): 68.5 -> 75.4Improved
increase agents from 2 to 3 (ablation)MAGRPO · StrBuildIoU (%): 50.6 -> 38.1Hurt
increase agents from 2 to 3 (ablation)CoLLM-DC · HouseBuildIoU (%): 46.8 -> 43.7Hurt

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