[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-85557-en":3,"doc-seo-85557-105":29,"detail-sidebar-cat-0-en-105":91},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":20,"is_deleted":4,"is_public":20,"is_downloadable":20,"audit_status":20,"page_count":21,"language":22,"language_code":23,"site_id":24,"html_lang":23,"table_of_contents":25,"faqs":26,"seo_title":13,"seo_description":14,"update_tm":27,"read_time":28},85557,16904993612988,"Olivia Brown","https://ap-avatar.wpscdn.com/davatar_a8503ba1806abce46bf441b54a3ca4cd",8,"Research & Report","GrandCode: Achieving Grandmaster Level in Competitive Programming via Agentic Reinforcement Learning","Competitive programming is presented as a key human stronghold where current AI systems still trail top competitors under live conditions. GrandCode is introduced as a multi-agent reinforcement learning system for competitive programming, coordinating hypothesis proposal, solving, test generation, and summarization. Its performance is attributed to joint improvement via post-training and online test-time RL, and to Agentic GRPO for multi-stage rollouts with delayed rewards, mitigating severe off-policy drift. GrandCode placed first across three recent Codeforces live rounds.","arXiv :2604 .0272 1v2 [ cs .AI] 13 Jul 2026  \nGrandCode: Achieving Grandmaster Level in Competitive Programming via Agentic Reinforcement Learning  \nXiaoya Li, Guoyin Wang, Songqiao Su, Chris Shum and Jiwei Li  \nDeepReinforce Team  \nAbstract  \nCompetitive 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. B  \nFigure 1: Codeforces standings overview for the three live contests in which GrandCode participated. GrandCode ranked first place in all three contests and being the first to finish all tasks in each of them.  \nB Email: {xiaoya_li, songqiao_su, chris_shum, [jiwei_li}@deep-reinforce.com](jiwei_li}@deep-reinforce.com)  \n1 Introduction  \nDespite rapid progress in AI for coding, the strongest current AI systems still fall short of the best human competitors in competitive programming. At the same time, the rapid improvement of large language models [21; 22 ; 23 ; 18 ; 7 ; 20 ; 35 ; 4] has driven substantial gains and has also spurred a growing literature on competitive-programming benchmarks, evaluation, and datasets [26; 6 ; 36 ; 17 ; 33 ; 14 ; 32 ; 19 ; 5 ; 40] . AlphaCode achieved a Codeforces rating of approximately 1300, placing it in the top 54% of competitors [16]; AlphaCode2 improved this to the 85th percentile [1]; and OpenAI’s o3 ranks 175th globally [24] . Most recently, Gemini 3 Deep Think attained a ranking of 8th place, though this result was obtained on historical problems rather than under live contest conditions.  \nIn this work, we introduce GrandCode, a multi-agent reinforcement learning system designed for competitive programming. GrandCode orchestrates a variety of agents and modules, and is optimized through both posttraining and online adaptation with test-time RL in an explicitly agentic loop: the hypothesis model proposes structural conjectures, the main solver takes the main responsibility of reasoning and solution generation, the summarization model maintains a compact memory of long context, and the test-case generator produces edge test cases to challenge proposed solutions. The goal of this design is to enable an agentic loop of reasoning, verification, and feedback, and continually refining its solutions.  \nTo address the severe off-policy issue in multi-turn agentic RL, we introduce Agentic GRPO, a variant of Group Relative Policy Optimization [27] that combines immediate reward updates with delayed correction, enabling more effective credit assignment under long, multi-stage rollouts and asynchronous training.  \nGrandCode is the first AI system to consistently surpass the best human competitors in competitive programming under live contest conditions: in the three most recent Codeforces rounds under standard live contest condit","cbCainIam1Zq1O05","https://ap.wps.com/l/cbCainIam1Zq1O05","pdf",9860119,1,31,"English","en",105,"# Introduction\n## Codeforces Competition Results\n## Codeforces","[{\"question\":\"What is GrandCode, and what problem does it target?\",\"answer\":\"GrandCode is a multi-agent reinforcement learning system designed for competitive programming. It targets the gap between current AI performance and top human competitors, especially under live contest conditions.\"},{\"question\":\"How does GrandCode’s multi-agent architecture work?\",\"answer\":\"GrandCode orchestrates multiple agentic modules, including hypothesis proposal, a main solver for reasoning and solution generation, summarization for compact long-context memory, and a test-case generator for edge cases.\"},{\"question\":\"What is Agentic GRPO and why is it important?\",\"answer\":\"Agentic GRPO is a variant of Group Relative Policy Optimization that combines immediate reward updates with delayed correction. It is designed to address severe off-policy drift in multi-turn, multi-stage agentic reinforcement learning.\"}]",1784204534,78,{"code":4,"msg":30,"data":31},"ok",{"site_id":24,"language":23,"slug":32,"title":13,"keywords":33,"description":14,"schema_data":34,"social_meta":86,"head_meta":88,"extra_data":90,"updated_unix":27},"grandcode-achieving-grandmaster-level-in-competitive-programming-via-agentic-reinforcement-learning","",{"@graph":35,"@context":85},[36,53,68],{"@type":37,"itemListElement":38},"BreadcrumbList",[39,43,47,50],{"item":40,"name":41,"@type":42,"position":20},"https://docshare.wps.com","Home","ListItem",{"item":44,"name":45,"@type":42,"position":46},"https://docshare.wps.com/document/","Document",2,{"item":48,"name":12,"@type":42,"position":49},"https://docshare.wps.com/document/research-report/",3,{"item":51,"name":13,"@type":42,"position":52},"https://docshare.wps.com/document/grandcode-achieving-grandmaster-level-in-competitive-programming-via-agentic-reinforcement-learning/85557/",4,{"url":51,"name":13,"@type":54,"author":55,"headline":13,"publisher":57,"fileFormat":60,"inLanguage":23,"description":14,"dateModified":61,"datePublished":62,"encodingFormat":60,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":56},"Person",{"url":40,"name":58,"@type":59},"DocShare","Organization","application/pdf","2026-07-17","2026-07-16",true,{"@type":65,"interactionType":66,"userInteractionCount":20},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71,77,81],{"name":72,"@type":73,"acceptedAnswer":74},"What is GrandCode, and what problem does it target?","Question",{"text":75,"@type":76},"GrandCode is a multi-agent reinforcement learning system designed for competitive programming. It targets the gap between current AI performance and top human competitors, especially under live contest conditions.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"How does GrandCode’s multi-agent architecture work?",{"text":80,"@type":76},"GrandCode orchestrates multiple agentic modules, including hypothesis proposal, a main solver for reasoning and solution generation, summarization for compact long-context memory, and a test-case generator for edge cases.",{"name":82,"@type":73,"acceptedAnswer":83},"What is Agentic GRPO and why is it important?",{"text":84,"@type":76},"Agentic GRPO is a variant of Group Relative Policy Optimization that combines immediate reward updates with delayed correction. It is designed to address severe off-policy drift in multi-turn, multi-stage agentic reinforcement learning.","https://schema.org",{"og:url":51,"og:type":87,"og:title":13,"og:site_name":58,"og:description":14},"article",{"robots":89,"canonical":51},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":92},[93,97,101,105,110,115,120,123,128,131,135],{"id":20,"doc_module":4,"doc_module_name":45,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":46,"doc_module":4,"doc_module_name":45,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":52,"doc_module":4,"doc_module_name":45,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":106,"doc_module":4,"doc_module_name":45,"category_name":107,"show_sort_weight":108,"slug":109},5,"Comic",60,"comic",{"id":111,"doc_module":4,"doc_module_name":45,"category_name":112,"show_sort_weight":113,"slug":114},6,"Technology",50,"technology",{"id":116,"doc_module":4,"doc_module_name":45,"category_name":117,"show_sort_weight":118,"slug":119},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":45,"category_name":12,"show_sort_weight":121,"slug":122},30,"research-report",{"id":124,"doc_module":4,"doc_module_name":45,"category_name":125,"show_sort_weight":126,"slug":127},9,"Religion & Spirituality",20,"religion-spirituality",{"id":126,"doc_module":4,"doc_module_name":45,"category_name":129,"show_sort_weight":126,"slug":130},"World Cup","world-cup",{"id":132,"doc_module":4,"doc_module_name":45,"category_name":133,"show_sort_weight":132,"slug":134},10,"Lifestyle","lifestyle",{"id":136,"doc_module":4,"doc_module_name":45,"category_name":137,"show_sort_weight":106,"slug":138},19,"General","general"]