🤖 How can we teach dexterous robots to perform precise, contact-rich assembly?</p>\n<p>Introducing Play2Perfect: first learn to play with objects, then perfect the policy for tight insertion, multi-part assembly, and screwing.</p>\n<p>Sound on! 🔊</p>\n","updatedAt":"2026-07-01T17:35:26.813Z","author":{"_id":"669093ca3a86663c1e4ae97c","avatarUrl":"/avatars/e3c514c6dbeae3df367c239b80616d0b.svg","fullname":"Tyler Lum","name":"tylerlum","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.9017700552940369},"editors":["tylerlum"],"editorAvatarUrls":["/avatars/e3c514c6dbeae3df367c239b80616d0b.svg"],"reactions":[],"isReport":false}},{"id":"6a45c3899ccf9a0e8c325567","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":"2026-07-02T01:48:57.000Z","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* [From Grasps to Dexterity: Large-Scale Grasp Pretraining for Dexterous Manipulation](https://huggingface.co/papers/2606.30749) (2026)\n* [Pose-Agnostic Robotic Functional Grasping via Observation-Action Canonicalization](https://huggingface.co/papers/2606.21148) (2026)\n* [TopoRetarget: Interaction-Preserving Retargeting for Dexterous Manipulation](https://huggingface.co/papers/2606.16272) (2026)\n* [Blind Dexterous Grasping via Real2Sim2Real Tactile Policy Learning](https://huggingface.co/papers/2606.11767) (2026)\n* [CoorDex: Coordinating Body and Hand Priors for Continuous Dexterous Humanoid Loco-Manipulation](https://huggingface.co/papers/2606.23680) (2026)\n* [Support-Constrained RL Enables Real-World Policy Improvement without Real-World Experience](https://huggingface.co/papers/2606.27475) (2026)\n* [TacCoRL: Integrating Tactile Feedback into VLA via Simulation](https://huggingface.co/papers/2606.11743) (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":"<p>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. </p>\n<p>The following papers were recommended by the Semantic Scholar API </p>\n<ul>\n<li><a href=\"https://huggingface.co/papers/2606.30749\">From Grasps to Dexterity: Large-Scale Grasp Pretraining for Dexterous Manipulation</a> (2026)</li>\n<li><a href=\"https://huggingface.co/papers/2606.21148\">Pose-Agnostic Robotic Functional Grasping via Observation-Action Canonicalization</a> (2026)</li>\n<li><a href=\"https://huggingface.co/papers/2606.16272\">TopoRetarget: Interaction-Preserving Retargeting for Dexterous Manipulation</a> (2026)</li>\n<li><a href=\"https://huggingface.co/papers/2606.11767\">Blind Dexterous Grasping via Real2Sim2Real Tactile Policy Learning</a> (2026)</li>\n<li><a href=\"https://huggingface.co/papers/2606.23680\">CoorDex: Coordinating Body and Hand Priors for Continuous Dexterous Humanoid Loco-Manipulation</a> (2026)</li>\n<li><a href=\"https://huggingface.co/papers/2606.27475\">Support-Constrained RL Enables Real-World Policy Improvement without Real-World Experience</a> (2026)</li>\n<li><a href=\"https://huggingface.co/papers/2606.11743\">TacCoRL: Integrating Tactile Feedback into VLA via Simulation</a> (2026)</li>\n</ul>\n<p> Please give a thumbs up to this comment if you found it helpful!</p>\n<p> 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</p>\n<p> You can directly ask Librarian Bot for paper recommendations by tagging it in a comment: <code>@librarian-bot recommend</code></p>\n","updatedAt":"2026-07-02T01:48:57.033Z","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":0.7231401205062866},"editors":["librarian-bot"],"editorAvatarUrls":["https://cdn-avatars.huggingface.co/v1/production/uploads/1674830754237-63d3e0e8ff1384ce6c5dd17d.jpeg"],"reactions":[],"isReport":false}}],"primaryEmailConfirmed":false,"paper":{"id":"2606.26428","authors":[{"_id":"6a43fb5d41f04ae4d7ad959b","name":"Tyler Ga Wei Lum","hidden":false},{"_id":"6a43fb5d41f04ae4d7ad959c","name":"Kushal Kedia","hidden":false},{"_id":"6a43fb5d41f04ae4d7ad959d","name":"C. Karen Liu","hidden":false},{"_id":"6a43fb5d41f04ae4d7ad959e","name":"Jeannette Bohg","hidden":false}],"mediaUrls":["https://cdn-uploads.huggingface.co/production/uploads/669093ca3a86663c1e4ae97c/Z4yiyZK-dbONqkaE9LT9P.mp4"],"publishedAt":"2026-06-24T00:00:00.000Z","submittedOnDailyAt":"2026-07-01T00:00:00.000Z","title":"Play2Perfect: What Matters in Dexterous Play Pretraining for Precise Assembly?","submittedOnDailyBy":{"_id":"669093ca3a86663c1e4ae97c","avatarUrl":"/avatars/e3c514c6dbeae3df367c239b80616d0b.svg","isPro":false,"fullname":"Tyler Lum","user":"tylerlum","type":"user","name":"tylerlum"},"summary":"Multi-fingered robots promise the speed and dexterity of human hands, yet challenging problems such as precise assembly have remained out of reach. These tasks are contact-rich, making data collection for imitation learning difficult, and sparse-reward, making direct exploration with reinforcement learning (RL) intractable. Consequently, prior work has made progress by structuring the problem with specialized grippers, tool attachments, and environment fixtures. In this work, we argue that before a robot can perfect precise assembly, it must first learn to play. We further ask the question: what factors in the process of learning to play matter for precise assembly? We propose Play2Perfect, an RL framework for task-agnostic pretraining through play on diverse objects and goals, which is then perfected on precise assembly. The goal of play is to acquire reusable manipulation priors, such as grasping, in-hand reorientation and pose reaching. Finetuning then adapts this general prior to assembly, focusing exploration on the final contact-rich, high-precision interactions needed for success. We systematically study key design choices in play pretraining, including object diversity, training objective, trajectory diversity, and goal precision. We show that our prior is 33x more sample-efficient than RL training from scratch, even when provided with dense, multi-stage rewards. We demonstrate zero-shot sim-to-real transfer, achieving 60% success on tight insertions with only 0.5 mm contact clearance, and over 50% success on long-horizon multi-part assembly and screwing.","upvotes":10,"discussionId":"6a43fb5e41f04ae4d7ad959f","projectPage":"https://play2perfect.github.io/","githubRepo":"https://github.com/kushal2000/play2perfect","githubRepoAddedBy":"user","ai_summary":"A reinforcement learning framework called Play2Perfect enables sample-efficient robotic assembly tasks by first learning general manipulation skills through playful interaction with diverse objects, then adapting these skills for precise assembly through fine-tuning.","ai_keywords":["reinforcement learning","imitation learning","contact-rich tasks","sparse-reward","task-agnostic pretraining","manipulation priors","grasping","in-hand reorientation","pose reaching","fine-tuning","sample efficiency","sim-to-real transfer","assembly"],"ai_summary_model":"Qwen/Qwen2.5-Coder-32B-Instruct","githubStars":13,"organization":{"_id":"672c672dcf09d152f4da04c4","name":"StanfordUniversity","fullname":"Stanford University","avatar":"https://cdn-avatars.huggingface.co/v1/production/uploads/68e396f2b5bb631e9b2fac9a/vJI0POlzGMXL2878t1vz2.jpeg"}},"canReadDatabase":false,"canManagePapers":false,"canSubmit":false,"hasHfLevelAccess":false,"upvoted":false,"upvoters":[{"_id":"669093ca3a86663c1e4ae97c","avatarUrl":"/avatars/e3c514c6dbeae3df367c239b80616d0b.svg","isPro":false,"fullname":"Tyler Lum","user":"tylerlum","type":"user"},{"_id":"65e1ff68ec41b21de6201cbd","avatarUrl":"/avatars/498f8d0b5d83c18e9aefdf730f5be170.svg","isPro":false,"fullname":"Kushal Kedia","user":"Kushal20","type":"user"},{"_id":"6344fbbc87964b331810d35d","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/1665466347462-6344fbbc87964b331810d35d.jpeg","isPro":false,"fullname":"Guy Tevet","user":"guytevet","type":"user"},{"_id":"65e20780ec41b21de623e65c","avatarUrl":"/avatars/5aaa4540eca3cbddc0a956d8dea2b27c.svg","isPro":false,"fullname":"Max Pace","user":"map438","type":"user"},{"_id":"65a6140947b88de066dd581f","avatarUrl":"/avatars/d01d8502bdc903ff998747bbcd3d262a.svg","isPro":false,"fullname":"Jisang Park","user":"alsichan","type":"user"},{"_id":"6480902ecacb1c4a069587ed","avatarUrl":"/avatars/90b8b9bf29a286eb3bb055121c989c7c.svg","isPro":false,"fullname":"ria doshi","user":"rdoshi21","type":"user"},{"_id":"64be8a7a5b8d826146f8a308","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/noauth/xxAFeuhAOxtvlmRC9EglO.jpeg","isPro":false,"fullname":"Dylan Zhou","user":"dylanzhou2","type":"user"},{"_id":"64f8fbd95515d7dcceb906b1","avatarUrl":"/avatars/1c7d034de408930b166592465e65fc31.svg","isPro":false,"fullname":"Yunhai Feng","user":"yunhaif","type":"user"},{"_id":"685106a0997075850b709eff","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/no-auth/2WLaSoecb190--Gr0_WHr.png","isPro":false,"fullname":"liu-ht23","user":"liu-ht23","type":"user"},{"_id":"636a8b90c95145940bfdec8c","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/1667926906903-noauth.jpeg","isPro":false,"fullname":"Huy Ha","user":"huy-ha","type":"user"}],"acceptLanguages":["en"],"dailyPaperRank":0,"organization":{"_id":"672c672dcf09d152f4da04c4","name":"StanfordUniversity","fullname":"Stanford University","avatar":"https://cdn-avatars.huggingface.co/v1/production/uploads/68e396f2b5bb631e9b2fac9a/vJI0POlzGMXL2878t1vz2.jpeg"},"markdownContentUrl":"https://huggingface.co/buckets/huggingchat/papers-content/resolve/2606/2606.26428.md","query":{}}">

Get this paper in your agent:

Don't have the latest CLI?

Cite arxiv.org/abs/2606.26428 in a model README.md to link it from this page.

Cite arxiv.org/abs/2606.26428 in a dataset README.md to link it from this page.

Cite arxiv.org/abs/2606.26428 in a Space README.md to link it from this page.

Add this paper to a collection to link it from this page.