AI Paper Atlas

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

AgentPO: Enhancing Multi-Agent Collaboration via Reinforcement Learning

Lin Sun, Chuang Liu, Can Zhang, Yubin Wu, Weijia Lu, Ning Wu

ICLR 2026 · Multi-agent systems and games › Cooperative multi-agent RL · approach: MARL benchmarks and grouping

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
train a lightweight hint collaborator with GRPO in front of a frozen actorLlama-3.1-8B-Instruct actor · Average over 5 math benchmarksPass@1: 19.4 (CoT) -> 31.5Improved
train collaborator with GRPO (with AgentPO) vs untrained hint model (ablation)Qwen-Plus actor, Hint-Actor · Average over 5 math benchmarksPass@1: 55.1 -> 58.8Improved
train critic with AgentPO vs untrained critic (ablation)Qwen-Plus actor, Critic-Actor · Average over 5 math benchmarksPass@1: 57.9 -> 58.9Improved
fine-tune only the 3B hint collaborator instead of the 7B actorQwen2.5-Math-7B · Average over 5 math benchmarksPass@1: 48.0 (Prime-Zero-7B) -> 49.4Improved
add AgentPO collaborator to multi-agent debateLlama-3.2-3B-Instruct debate · OlympiadBenchPass@1: 15.3 -> 21.1Improved
add AgentPO collaborator to role assignmentLlama-3.2-3B-Instruct role assignment · MATH500Pass@1: 48.3 -> 52.2Improved
increase training samples from 500 to 700 and 1,000 (ablation)AgentPO hint model · Average over 5 math benchmarks (Qwen2.5-Math-7B actor)Pass@1: 49.4 at 500 samples, slight dip at 700 and 1000Hurt
use hint model Qwen2.5-3B-Instruct instead of larger Qwen2.5-7B-Instruct (ablation)AgentPO with Llama-3.1-8B-Instruct actor · Average Pass@1Pass@1: 29.1 -> 31.5Improved
add AgentPO hint model to API-based Qwen-Plus actorQwen-Plus · OlympiadBenchPass@1: 49.6 -> 49.6No 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