self-verification in training loop.</p>\n","updatedAt":"2026-07-20T20:54:52.645Z","author":{"_id":"65d3b7ec8f6b98b34ee6bbe3","avatarUrl":"/avatars/53c2d4e4746147fc2559435d252e8951.svg","fullname":"Mingyuan Wu","name":"Mingyuan1997","type":"user","isPro":true,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":2,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.9433063864707947},"editors":["Mingyuan1997"],"editorAvatarUrls":["/avatars/53c2d4e4746147fc2559435d252e8951.svg"],"reactions":[],"isReport":false}}],"primaryEmailConfirmed":false,"paper":{"id":"2607.10966","authors":[{"_id":"6a5e8ae04fe5d1d13e84aa3b","name":"Mingyuan Wu","hidden":false},{"_id":"6a5e8ae04fe5d1d13e84aa3c","name":"Jingcheng Yang","hidden":false},{"_id":"6a5e8ae04fe5d1d13e84aa3d","name":"Shengyi Qian","hidden":false},{"_id":"6a5e8ae04fe5d1d13e84aa3e","name":"Xudong Wang","hidden":false},{"_id":"6a5e8ae04fe5d1d13e84aa3f","name":"Jize Jiang","hidden":false},{"_id":"6a5e8ae04fe5d1d13e84aa40","name":"Qifan Wang","hidden":false},{"_id":"6a5e8ae04fe5d1d13e84aa41","name":"Aashu Singh","hidden":false},{"_id":"6a5e8ae04fe5d1d13e84aa42","name":"Khoi Pham","hidden":false},{"_id":"6a5e8ae04fe5d1d13e84aa43","name":"Fei Liu","hidden":false},{"_id":"6a5e8ae04fe5d1d13e84aa44","name":"Zhaolun Su","hidden":false},{"_id":"6a5e8ae04fe5d1d13e84aa45","name":"Zhuokai Zhao","hidden":false},{"_id":"6a5e8ae04fe5d1d13e84aa46","name":"Klara Nahrstedt","hidden":false},{"_id":"6a5e8ae04fe5d1d13e84aa47","name":"Jianyu Wang","hidden":false},{"_id":"6a5e8ae04fe5d1d13e84aa48","name":"Hanchao Yu","hidden":false}],"publishedAt":"2026-07-13T00:00:00.000Z","submittedOnDailyAt":"2026-07-20T00:00:00.000Z","title":"SVR-R1: Bootstrapping Multi-modal Reasoning with Self-verification in Reinforcement Learning","submittedOnDailyBy":{"_id":"65d3b7ec8f6b98b34ee6bbe3","avatarUrl":"/avatars/53c2d4e4746147fc2559435d252e8951.svg","isPro":true,"fullname":"Mingyuan Wu","user":"Mingyuan1997","type":"user","name":"Mingyuan1997"},"summary":"We introduce Self-Verified Reasoner (SVR-R1), a multi-turn RL framework that turns a model's own verification into a learning signal for multimodal reasoning. For each query, the model proposes an answer using the same weights, and issues a binary self-verdict (Yes/No). A 'No' triggers a second-chance rethink; a 'Yes,' or a turn cap, finalizes the output for computing the outcome-based reward. SVR-R1 is implemented with GRPO and an asynchronous multi-turn rollout framework and needs no external supervision or auxiliary critics. We evaluate SVR-R1 on vision-language reasoning benchmarks and show that it improves accuracy by a large margin over strong standard GRPO baselines. Training dynamics show decreasing reliance on verification-fewer verification turns, yet higher test accuracy-indicating that the gap between verification and generation narrows as the policy internalizes self-correction and chooses the most confident answer via our framework. SVR-R1 bridges the less explored intersection of inference-time self-refinement and RL training for VLMs, offering a simple yet effective recipe for bootstrapping multimodal reasoning. We will open-source SVR-R1 to facilitate future research in VLMs.","upvotes":2,"discussionId":"6a5e8ae04fe5d1d13e84aa49","organization":{"_id":"65448bef5b5d9185ba3202b9","name":"UIUC-CS","fullname":"University of Illinois at Urbana-Champaign","avatar":"https://cdn-avatars.huggingface.co/v1/production/uploads/65448b21fcb96b8b48733729/ycqcXFayMTTD_KpE37067.jpeg"}},"canReadDatabase":false,"canManagePapers":false,"canSubmit":false,"hasHfLevelAccess":false,"upvoted":false,"upvoters":[{"_id":"6a2da6c8ca070ee12c6e396c","avatarUrl":"/avatars/0355287dcabaa67dbc7f0b10b87451f9.svg","isPro":false,"fullname":"Joe Mama","user":"JoeMama123123123","type":"user"},{"_id":"631e14ac473a6825f285e89d","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/631e14ac473a6825f285e89d/K-6QnoeGLg8XFvbTMMdqA.jpeg","isPro":false,"fullname":"Yury Panikov","user":"panikov","type":"user"}],"acceptLanguages":["en"],"dailyPaperRank":0,"organization":{"_id":"65448bef5b5d9185ba3202b9","name":"UIUC-CS","fullname":"University of Illinois at Urbana-Champaign","avatar":"https://cdn-avatars.huggingface.co/v1/production/uploads/65448b21fcb96b8b48733729/ycqcXFayMTTD_KpE37067.jpeg"},"markdownContentUrl":"https://huggingface.co/buckets/huggingchat/papers-content/resolve/2607/2607.10966.md","query":{}}">
SVR-R1: Bootstrapping Multi-modal Reasoning with Self-verification in Reinforcement Learning
Abstract
We introduce Self-Verified Reasoner (SVR-R1), a multi-turn RL framework that turns a model's own verification into a learning signal for multimodal reasoning. For each query, the model proposes an answer using the same weights, and issues a binary self-verdict (Yes/No). A 'No' triggers a second-chance rethink; a 'Yes,' or a turn cap, finalizes the output for computing the outcome-based reward. SVR-R1 is implemented with GRPO and an asynchronous multi-turn rollout framework and needs no external supervision or auxiliary critics. We evaluate SVR-R1 on vision-language reasoning benchmarks and show that it improves accuracy by a large margin over strong standard GRPO baselines. Training dynamics show decreasing reliance on verification-fewer verification turns, yet higher test accuracy-indicating that the gap between verification and generation narrows as the policy internalizes self-correction and chooses the most confident answer via our framework. SVR-R1 bridges the less explored intersection of inference-time self-refinement and RL training for VLMs, offering a simple yet effective recipe for bootstrapping multimodal reasoning. We will open-source SVR-R1 to facilitate future research in VLMs.
Community
self-verification in training loop.
Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.
Tap or paste here to upload images
Cite arxiv.org/abs/2607.10966 in a model README.md to link it from this page.
Cite arxiv.org/abs/2607.10966 in a dataset README.md to link it from this page.
Cite arxiv.org/abs/2607.10966 in a Space README.md to link it from this page.
Discussion (0)
Sign in to join the discussion. Free account, 30 seconds — email code or GitHub.
Sign in →No comments yet. Sign in and be the first to say something.