MarsRL: Advancing Multi-Agent Reasoning System via Reinforcement Learning with Agentic Pipeline Parallelism

TL;DR

MarsRL is a reinforcement learning framework that enhances multi-agent reasoning systems by jointly optimizing agents with pipeline parallelism, significantly improving accuracy on reasoning benchmarks.

Contribution

Introduces MarsRL, a novel RL framework with agent-specific rewards and pipeline training to improve multi-agent reasoning in open-source models.

Findings

Improves AIME2025 accuracy from 86.5% to 93.3%.

Enhances BeyondAIME accuracy from 64.9% to 73.8%.

Surpasses larger models in reasoning tasks.

Abstract

Recent progress in large language models (LLMs) has been propelled by reinforcement learning with verifiable rewards (RLVR) and test-time scaling. However, the limited output length of LLMs constrains the depth of reasoning attainable in a single inference process. Multi-agent reasoning systems offer a promising alternative by employing multiple agents including Solver, Verifier, and Corrector, to iteratively refine solutions. While effective in closed-source models like Gemini 2.5 Pro, they struggle to generalize to open-source models due to insufficient critic and correction capabilities. To address this, we propose MarsRL, a novel reinforcement learning framework with agentic pipeline parallelism, designed to jointly optimize all agents in the system. MarsRL introduces agent-specific reward mechanisms to mitigate reward noise and employs pipeline-inspired training to enhance efficiency in handling long trajectories. Applied to Qwen3-30B-A3B-Thinking-2507, MarsRL improves AIME2025 accuracy from 86.5% to 93.3% and BeyondAIME from 64.9% to 73.8%, even surpassing Qwen3-235B-A22B-Thinking-2507. These findings highlight the potential of MarsRL to advance multi-agent reasoning systems and broaden their applicability across diverse reasoning tasks.