TL;DR
GrandCode is a multi-agent reinforcement learning system that achieves superhuman performance in competitive programming contests, surpassing top human players in recent live competitions.
Contribution
It introduces a novel multi-agent RL framework with agentic modules and a new training algorithm, Agentic GRPO, to excel in complex, multi-stage tasks like competitive programming.
Findings
GrandCode outperformed all human participants in three recent Codeforces contests.
It is the first AI system to consistently beat top human programmers in live competitive programming.
The system demonstrates AI's potential to surpass human expertise in complex coding tasks.
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…
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Code & Models
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
