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.
V-C Reasoning System Evaluation Instructions
step1: Download our released model or other open source models
Supported models: Qwen3/DeepSeekV3.1/DeepSeek R1. You can modify the llm_client.py to use other models.
step2: Deploy service via VLLM
step3: Run the V-C reasoning system by the following commands:
1@article{Marsrl2025,
2 title = {MarsRL: Advancing Multi-Agent Reasoning System via Reinforcement Learning with Agentic Pipeline Parallelism},
3 author = {Shulin Liu, Dong Du, Tao Yang, Yang Li, Boyu Qiu}
4 year = {2025}
5}
