Hugging Face Daily Papers · July 1, 2026 · 5 min read

Dockerless: Environment-Free Program Verifier for Coding Agents

Mirrored from Hugging Face Daily Papers for archival readability. Support the source by reading on the original site.

Like Read original ↗

Try environment-free RL!\n","updatedAt":"2026-07-01T02:01:08.333Z","author":{"_id":"645b0c3ec35da9c7afd95421","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/645b0c3ec35da9c7afd95421/vYBrCDagHsXAo6J2p-uG0.jpeg","fullname":"Yuling","name":"YerbaPage","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":109,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.7525107264518738},"editors":["YerbaPage"],"editorAvatarUrls":["https://cdn-avatars.huggingface.co/v1/production/uploads/645b0c3ec35da9c7afd95421/vYBrCDagHsXAo6J2p-uG0.jpeg"],"reactions":[{"reaction":"🚀","users":["azzzacs","shilhe","bbzs","SiyuYe1","sY713","ayanami-kitasan","AlexCuadron"],"count":7}],"isReport":false}},{"id":"6a44839e52f3fb621f790612","author":{"_id":"64b75ba000bac1088cea9231","avatarUrl":"/avatars/5a8bb7cccfe20dad3fda4405fb1eb3ff.svg","fullname":"Siyu Ye","name":"SiyuYe1","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":1,"isUserFollowing":false},"createdAt":"2026-07-01T03:03:58.000Z","type":"comment","data":{"edited":false,"hidden":false,"latest":{"raw":"Environment-free does not mean no repo access.\nGreat work scaling RL post-training without per-repo complex Docker setup.","html":"Environment-free does not mean no repo access. Great work scaling RL post-training without per-repo complex Docker setup.\n","updatedAt":"2026-07-01T03:03:58.739Z","author":{"_id":"64b75ba000bac1088cea9231","avatarUrl":"/avatars/5a8bb7cccfe20dad3fda4405fb1eb3ff.svg","fullname":"Siyu Ye","name":"SiyuYe1","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":1,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.8691291213035583},"editors":["SiyuYe1"],"editorAvatarUrls":["/avatars/5a8bb7cccfe20dad3fda4405fb1eb3ff.svg"],"reactions":[{"reaction":"👍","users":["YerbaPage","azzzacs"],"count":2}],"isReport":false}},{"id":"6a45c383e578d36124c8e9b3","author":{"_id":"63d3e0e8ff1384ce6c5dd17d","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/1674830754237-63d3e0e8ff1384ce6c5dd17d.jpeg","fullname":"Librarian Bot (Bot)","name":"librarian-bot","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":372,"isUserFollowing":false},"createdAt":"2026-07-02T01:48:51.000Z","type":"comment","data":{"edited":false,"hidden":false,"latest":{"raw":"This is an automated message from the [Librarian Bot](https://huggingface.co/librarian-bots). I found the following papers similar to this paper. \n\nThe following papers were recommended by the Semantic Scholar API \n\n* [Orchard: An Open-Source Agentic Modeling Framework](https://huggingface.co/papers/2605.15040) (2026)\n* [Automating Formal Verification with Reinforcement Learning and Recursive Inference](https://huggingface.co/papers/2605.30914) (2026)\n* [LiteCoder-Terminal: Scaling Long-Horizon Terminal Environments for Learning Language Agents](https://huggingface.co/papers/2605.29559) (2026)\n* [FastContext: Training Efficient Repository Explorer for Coding Agents](https://huggingface.co/papers/2606.14066) (2026)\n* [Socratic-SWE: Self-Evolving Coding Agents via Trace-Derived Agent Skills](https://huggingface.co/papers/2606.07412) (2026)\n* [Terminus-4B: Can a Smaller Model Replace Frontier LLMs at Agentic Execution Tasks?](https://huggingface.co/papers/2605.03195) (2026)\n* [CUA-Gym: Scaling Verifiable Training Environments and Tasks for Computer-Use Agents](https://huggingface.co/papers/2605.25624) (2026)\n\n\n Please give a thumbs up to this comment if you found it helpful!\n\n If you want recommendations for any Paper on Hugging Face checkout [this](https://huggingface.co/spaces/librarian-bots/recommend_similar_papers) Space\n\n You can directly ask Librarian Bot for paper recommendations by tagging it in a comment: `@librarian-bot recommend`","html":"This is an automated message from the <a href=\"https://huggingface.co/librarian-bots\">Librarian Bot</a>. I found the following papers similar to this paper. \nThe following papers were recommended by the Semantic Scholar API \n<ul>\n<li><a href=\"https://huggingface.co/papers/2605.15040\">Orchard: An Open-Source Agentic Modeling Framework</a> (2026)</li>\n<li><a href=\"https://huggingface.co/papers/2605.30914\">Automating Formal Verification with Reinforcement Learning and Recursive Inference</a> (2026)</li>\n<li><a href=\"https://huggingface.co/papers/2605.29559\">LiteCoder-Terminal: Scaling Long-Horizon Terminal Environments for Learning Language Agents</a> (2026)</li>\n<li><a href=\"https://huggingface.co/papers/2606.14066\">FastContext: Training Efficient Repository Explorer for Coding Agents</a> (2026)</li>\n<li><a href=\"https://huggingface.co/papers/2606.07412\">Socratic-SWE: Self-Evolving Coding Agents via Trace-Derived Agent Skills</a> (2026)</li>\n<li><a href=\"https://huggingface.co/papers/2605.03195\">Terminus-4B: Can a Smaller Model Replace Frontier LLMs at Agentic Execution Tasks?</a> (2026)</li>\n<li><a href=\"https://huggingface.co/papers/2605.25624\">CUA-Gym: Scaling Verifiable Training Environments and Tasks for Computer-Use Agents</a> (2026)</li>\n</ul>\n Please give a thumbs up to this comment if you found it helpful!\n If you want recommendations for any Paper on Hugging Face checkout <a href=\"https://huggingface.co/spaces/librarian-bots/recommend_similar_papers\">this</a> Space\n You can directly ask Librarian Bot for paper recommendations by tagging it in a comment: <code>@librarian-bot recommend</code>\n","updatedAt":"2026-07-02T01:48:51.720Z","author":{"_id":"63d3e0e8ff1384ce6c5dd17d","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/1674830754237-63d3e0e8ff1384ce6c5dd17d.jpeg","fullname":"Librarian Bot (Bot)","name":"librarian-bot","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":372,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.71711266040802},"editors":["librarian-bot"],"editorAvatarUrls":["https://cdn-avatars.huggingface.co/v1/production/uploads/1674830754237-63d3e0e8ff1384ce6c5dd17d.jpeg"],"reactions":[],"isReport":false}}],"primaryEmailConfirmed":false,"paper":{"id":"2606.28436","authors":[{"_id":"6a433a39763f63ca3757e9ab","name":"Wenhao Zeng","hidden":false},{"_id":"6a433a39763f63ca3757e9ac","user":{"_id":"645b0c3ec35da9c7afd95421","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/645b0c3ec35da9c7afd95421/vYBrCDagHsXAo6J2p-uG0.jpeg","isPro":false,"fullname":"Yuling","user":"YerbaPage","type":"user","name":"YerbaPage"},"name":"Yuling Shi","status":"claimed_verified","statusLastChangedAt":"2026-07-01T08:46:59.756Z","hidden":false},{"_id":"6a433a39763f63ca3757e9ad","name":"Xiaodong Gu","hidden":false},{"_id":"6a433a39763f63ca3757e9ae","name":"Chao Hu","hidden":false},{"_id":"6a433a39763f63ca3757e9af","name":"Chaofan Wang","hidden":false},{"_id":"6a433a39763f63ca3757e9b0","name":"Yuhao Cui","hidden":false},{"_id":"6a433a39763f63ca3757e9b1","name":"Hongting Zhou","hidden":false},{"_id":"6a433a39763f63ca3757e9b2","name":"Mengnan Qi","hidden":false},{"_id":"6a433a39763f63ca3757e9b3","name":"Jianqiao Wangni","hidden":false},{"_id":"6a433a39763f63ca3757e9b4","name":"Zhaojian Yu","hidden":false},{"_id":"6a433a39763f63ca3757e9b5","name":"Shuzheng Gao","hidden":false},{"_id":"6a433a39763f63ca3757e9b6","name":"Kai Cai","hidden":false},{"_id":"6a433a39763f63ca3757e9b7","name":"Shilin He","hidden":false}],"publishedAt":"2026-06-26T00:00:00.000Z","submittedOnDailyAt":"2026-07-01T00:00:00.000Z","title":"Dockerless: Environment-Free Program Verifier for Coding Agents","submittedOnDailyBy":{"_id":"645b0c3ec35da9c7afd95421","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/645b0c3ec35da9c7afd95421/vYBrCDagHsXAo6J2p-uG0.jpeg","isPro":false,"fullname":"Yuling","user":"YerbaPage","type":"user","name":"YerbaPage"},"summary":"Program verifiers play a central role in training coding agents, including selecting trajectories for supervised fine-tuning (SFT) and providing rewards for reinforcement learning (RL). Standard execution-based verification requires running unit tests inside per-repository environments such as Docker images, incurring substantial environment setup costs. We propose Dockerless, an environment-free agentic patch verifier that evaluates generated code patches without executing them. Rather than simply matching candidate patches to references, Dockerless judges patch correctness using evidence gathered through agentic repository exploration. On a verifier evaluation benchmark, Dockerless outperforms the strongest open-source verifier by 14.3 AUC points. Using Dockerless as both the SFT trajectory filter and the RL reward enables a fully environment-free post-training pipeline. The resulting model reaches 62.0%, 50.0%, and 35.2% resolve rate on SWE-bench Verified, Multilingual, and Pro, respectively. It surpasses the Qwen3.5-9B baseline by 2.4, 8.7, and 2.9 points, matching environment-based post-training.","upvotes":86,"discussionId":"6a433a39763f63ca3757e9b8","ai_summary":"A Dockerless environment-free agentic patch verifier improves code patch evaluation accuracy and enables effective post-training without execution-based verification costs.","ai_keywords":["program verifiers","supervised fine-tuning","reinforcement learning","Dockerless","agentic patch verifier","environment-free","repository exploration","SWE-bench Verified","Multilingual","Pro"],"ai_summary_model":"Qwen/Qwen2.5-Coder-32B-Instruct","organization":{"_id":"653b817d32c97d0655575872","name":"ByteDance","fullname":"ByteDance","avatar":"https://cdn-avatars.huggingface.co/v1/production/uploads/6535c9e88bde2fae19b6fb25/0clr54wj5Ly-RkYU9OXPp.png"}},"canReadDatabase":false,"canManagePapers":false,"canSubmit":false,"hasHfLevelAccess":false,"upvoted":false,"upvoters":[{"_id":"645b0c3ec35da9c7afd95421","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/645b0c3ec35da9c7afd95421/vYBrCDagHsXAo6J2p-uG0.jpeg","isPro":false,"fullname":"Yuling","user":"YerbaPage","type":"user"},{"_id":"68831681240aa3d8ce43e1bf","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/no-auth/iSjQ3OIkIvqm_0Wxv4Qib.png","isPro":false,"fullname":"Azzz","user":"azzzacs","type":"user"},{"_id":"664aef59691370727c27ad2a","avatarUrl":"/avatars/2f3a319723d3995b1f3696ea0acd5edd.svg","isPro":false,"fullname":"Daisy Chen","user":"HuckleberryPopo","type":"user"},{"_id":"695f6988a4b57c6042a16889","avatarUrl":"/avatars/bd5b040f58a650f24249dbb9b632d954.svg","isPro":false,"fullname":"Yaoning Wang","user":"Castria-fdu","type":"user"},{"_id":"6563f81b738013bf3084a172","avatarUrl":"/avatars/d3e4cca44deee36324d9c588c9392cc1.svg","isPro":false,"fullname":"wcf","user":"bbzs","type":"user"},{"_id":"65684c80a9a1a6a50d779f58","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/65684c80a9a1a6a50d779f58/it534ZdH5LxRub1M_o3uM.jpeg","isPro":false,"fullname":"Silin Chen","user":"Silin-Chen","type":"user"},{"_id":"66260fa6eb729227ec6c002a","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/66260fa6eb729227ec6c002a/bwhhv6xkP7M9YbCOXDpp1.jpeg","isPro":false,"fullname":"Heng Lian","user":"Pixelieee","type":"user"},{"_id":"6909aada9ceb1a82bd4f8bc3","avatarUrl":"/avatars/eba734ea9323a4068b72e9b15c754220.svg","isPro":false,"fullname":"Li Yanxiao","user":"yan-xiaoo","type":"user"},{"_id":"635b482862fb2bc1b52d6be3","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/1666926580685-noauth.png","isPro":false,"fullname":"FuCao","user":"ythere","type":"user"},{"_id":"6a30ba25b9ca290e25e9553c","avatarUrl":"/avatars/a5ffca330ec10dd641418a800f330d42.svg","isPro":false,"fullname":"Xingyuan Liu","user":"luneastral","type":"user"},{"_id":"62c6df026a092eda1f1ab6e5","avatarUrl":"/avatars/d58fff1a157b189ce2617889ef5f6e2f.svg","isPro":false,"fullname":"Shilin He","user":"shilhe","type":"user"},{"_id":"6a2da6c8ca070ee12c6e396c","avatarUrl":"/avatars/0355287dcabaa67dbc7f0b10b87451f9.svg","isPro":false,"fullname":"Joe Mama","user":"JoeMama123123123","type":"user"}],"acceptLanguages":["en"],"dailyPaperRank":2,"organization":{"_id":"653b817d32c97d0655575872","name":"ByteDance","fullname":"ByteDance","avatar":"https://cdn-avatars.huggingface.co/v1/production/uploads/6535c9e88bde2fae19b6fb25/0clr54wj5Ly-RkYU9OXPp.png"},"markdownContentUrl":"https://huggingface.co/buckets/huggingchat/papers-content/resolve/2606/2606.28436.md","query":{}}">

Papers

arxiv:2606.28436

Dockerless: Environment-Free Program Verifier for Coding Agents

Published on Jun 26

· Submitted by

Yuling on Jul 1

#2 Paper of the day

ByteDance

Upvote

Authors:

Yuling Shi ,

Abstract

A Dockerless environment-free agentic patch verifier improves code patch evaluation accuracy and enables effective post-training without execution-based verification costs.

Generated by Qwen/Qwen2.5-Coder-32B-Instruct

Program verifiers play a central role in training coding agents, including selecting trajectories for supervised fine-tuning (SFT) and providing rewards for reinforcement learning (RL). Standard execution-based verification requires running unit tests inside per-repository environments such as Docker images, incurring substantial environment setup costs. We propose Dockerless, an environment-free agentic patch verifier that evaluates generated code patches without executing them. Rather than simply matching candidate patches to references, Dockerless judges patch correctness using evidence gathered through agentic repository exploration. On a verifier evaluation benchmark, Dockerless outperforms the strongest open-source verifier by 14.3 AUC points. Using Dockerless as both the SFT trajectory filter and the RL reward enables a fully environment-free post-training pipeline. The resulting model reaches 62.0%, 50.0%, and 35.2% resolve rate on SWE-bench Verified, Multilingual, and Pro, respectively. It surpasses the Qwen3.5-9B baseline by 2.4, 8.7, and 2.9 points, matching environment-based post-training.