arXiv cs.AI· Ornith Team, Xiaoya Li, Guoyin Wang, Songqiao Su, Jiwei Li, Chris Shum·· 3 小时前精选AI 评分70
GrandCode 通过智能体强化学习在竞技编程中达到大师级水平
GrandCode: Achieving Grandmaster Level in Competitive Programming via Agentic Reinforcement Learning
AI 导读
Ornith Team 等发布论文，介绍多智能体强化学习系统 GrandCode，称其为首个在实时竞技编程比赛中稳定击败所有人类参赛者的 AI 系统。系统编排假设提出、求解、测试生成、总结等智能体模块，通过后训练与在线测试时强化学习联合改进，并提出面向多阶段智能体 rollout、延迟奖励和严重 off-policy 漂移的 Agentic GRPO。
推荐理由
原文报告了多智能体强化学习系统在 Codeforces 三场实时比赛中均夺冠的细节，可了解其智能体编排与 Agentic GRPO 设计。
正文 · 原文
Abstract:Competitive programming remains one of the last few human strongholds in coding against AI. The best AI system to date still underperforms the best humans competitive programming: the most recent best result, Google's Gemini~3 Deep Think, attained 8th place even not being evaluated under live competition conditions. In this work, we introduce GrandCode, a multi-agent RL system designed for competitive programming. The capability of GrandCode is attributed to two key factors: (1) It orchestrates a variety of agentic modules (hypothesis proposal, solver, test generator, summarization, etc) and jointly improves them through post-training and online test-time RL; (2) We introduce Agentic GRPO specifically designed for multi-stage agent rollouts with delayed rewards and the severe off-policy drift that is prevalent in agentic RL. GrandCode is the first AI system that consistently beats all human participants in live contests of competitive programming: in the most recent three Codeforces live competitions, i.e., Round~1087 (Mar 21, 2026), Round~1088 (Mar 28, 2026), and Round~1089 (Mar 29, 2026), GrandCode placed first in all of them, beating all human participants, including legendary grandmasters. GrandCode shows that AI systems have reached a point where they surpass the strongest human programmers on the most competitive coding tasks.
| Comments: | Tech Report; Pre-print |
| Subjects: | Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2604.02721 [cs.AI] |
| (or arXiv:2604.02721v4 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2604.02721 arXiv-issued DOI via DataCite |
Submission history
From: Jiwei Li [view email]
[v1]
Fri, 3 Apr 2026 04:26:56 UTC (9,653 KB)
[v2]
Mon, 13 Jul 2026 00:56:35 UTC (9,653 KB)
[v3]
Wed, 5 Aug 2026 18:24:09 UTC (9,653 KB)
[v4]
Wed, 11 Nov 2026 15:59:14 UTC (9,601 KB)
来源：arXiv cs.AI · arxiv.org