Librarian Bot. I found the following papers similar to this paper.

The following papers were recommended by the Semantic Scholar API

Please give a thumbs up to this comment if you found it helpful!

If you want recommendations for any Paper on Hugging Face checkout this Space

You can directly ask Librarian Bot for paper recommendations by tagging it in a comment: \n\n@librarian-bot\n\t recommend

\n","updatedAt":"2025-06-06T01:38:24.003Z","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":318,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.7137890458106995},"editors":["librarian-bot"],"editorAvatarUrls":["https://cdn-avatars.huggingface.co/v1/production/uploads/1674830754237-63d3e0e8ff1384ce6c5dd17d.jpeg"],"reactions":[],"isReport":false}}],"primaryEmailConfirmed":false,"paper":{"id":"2506.04178","authors":[{"_id":"68417762d777f13c594dd03c","name":"Etash Guha","hidden":false},{"_id":"68417762d777f13c594dd03d","user":{"_id":"626c182a030a6e7363b6fe0a","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/626c182a030a6e7363b6fe0a/NaMcIOPVgJ6JFt8fojZWU.jpeg","isPro":false,"fullname":"Ryan Marten","user":"ryanmarten","type":"user"},"name":"Ryan Marten","status":"claimed_verified","statusLastChangedAt":"2025-06-07T05:49:20.703Z","hidden":false},{"_id":"68417762d777f13c594dd03e","name":"Sedrick Keh","hidden":false},{"_id":"68417762d777f13c594dd03f","name":"Negin Raoof","hidden":false},{"_id":"68417762d777f13c594dd040","name":"Georgios Smyrnis","hidden":false},{"_id":"68417762d777f13c594dd041","name":"Hritik Bansal","hidden":false},{"_id":"68417762d777f13c594dd042","user":{"_id":"62500f361684e0335e527bc6","avatarUrl":"/avatars/8d2bc4c9cfa8ea82049196431cc3ebea.svg","isPro":false,"fullname":"Marianna Nezhurina","user":"marianna13","type":"user"},"name":"Marianna Nezhurina","status":"claimed_verified","statusLastChangedAt":"2025-06-07T05:49:22.779Z","hidden":false},{"_id":"68417762d777f13c594dd043","name":"Jean Mercat","hidden":false},{"_id":"68417762d777f13c594dd044","user":{"_id":"6338cb22768611eccab73983","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/6338cb22768611eccab73983/AgdbEwS8TSWZS2vOLkR2u.jpeg","isPro":false,"fullname":"Trung Vu","user":"trungtvu","type":"user"},"name":"Trung Vu","status":"claimed_verified","statusLastChangedAt":"2025-06-11T08:36:55.579Z","hidden":false},{"_id":"68417762d777f13c594dd045","name":"Zayne Sprague","hidden":false},{"_id":"68417762d777f13c594dd046","user":{"_id":"628c29a54c5a62a1d216c560","avatarUrl":"/avatars/d21b4da766f87f47228112958666643b.svg","isPro":false,"fullname":"Ashima Suvarna","user":"Ashima","type":"user"},"name":"Ashima Suvarna","status":"claimed_verified","statusLastChangedAt":"2025-07-10T09:13:01.862Z","hidden":false},{"_id":"68417762d777f13c594dd047","user":{"_id":"62f7f4efe7c1c9bf10c81465","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/62f7f4efe7c1c9bf10c81465/AYlOg0fkP1o4GAP-8Y3xt.jpeg","isPro":true,"fullname":"Benjamin Feuer","user":"penfever","type":"user"},"name":"Benjamin Feuer","status":"claimed_verified","statusLastChangedAt":"2025-06-07T05:49:16.121Z","hidden":false},{"_id":"68417762d777f13c594dd048","name":"Liangyu Chen","hidden":false},{"_id":"68417762d777f13c594dd049","name":"Zaid Khan","hidden":false},{"_id":"68417762d777f13c594dd04a","name":"Eric Frankel","hidden":false},{"_id":"68417762d777f13c594dd04b","name":"Sachin Grover","hidden":false},{"_id":"68417762d777f13c594dd04c","name":"Caroline Choi","hidden":false},{"_id":"68417762d777f13c594dd04d","name":"Niklas Muennighoff","hidden":false},{"_id":"68417762d777f13c594dd04e","name":"Shiye Su","hidden":false},{"_id":"68417762d777f13c594dd04f","name":"Wanjia Zhao","hidden":false},{"_id":"68417762d777f13c594dd050","name":"John Yang","hidden":false},{"_id":"68417762d777f13c594dd051","user":{"_id":"6444e4417a7b94ddc2d14e1d","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/6444e4417a7b94ddc2d14e1d/kC9bUAHujBf2XQs1c2QfS.png","isPro":true,"fullname":"Shreyas Pimpalgaonkar","user":"pimpalgaonkar","type":"user"},"name":"Shreyas Pimpalgaonkar","status":"claimed_verified","statusLastChangedAt":"2025-06-06T07:41:50.414Z","hidden":false},{"_id":"68417762d777f13c594dd052","name":"Kartik Sharma","hidden":false},{"_id":"68417762d777f13c594dd053","name":"Charlie Cheng-Jie Ji","hidden":false},{"_id":"68417762d777f13c594dd054","user":{"_id":"63f39ca90be81bdc5d94acc5","avatarUrl":"/avatars/2eb069023671d8a9c118dd3c771d2d74.svg","isPro":false,"fullname":"Ethan Deng","user":"SGEthan","type":"user"},"name":"Yichuan Deng","status":"claimed_verified","statusLastChangedAt":"2025-06-06T07:41:52.871Z","hidden":false},{"_id":"68417762d777f13c594dd055","name":"Sarah Pratt","hidden":false},{"_id":"68417762d777f13c594dd056","name":"Vivek Ramanujan","hidden":false},{"_id":"68417762d777f13c594dd057","name":"Jon Saad-Falcon","hidden":false},{"_id":"68417762d777f13c594dd058","name":"Jeffrey Li","hidden":false},{"_id":"68417762d777f13c594dd059","name":"Achal Dave","hidden":false},{"_id":"68417762d777f13c594dd05a","user":{"_id":"611a7ec4289467cafea62d13","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/611a7ec4289467cafea62d13/pck-0fmPQkoU7yzh6-WoL.jpeg","isPro":false,"fullname":"Alon Albalak","user":"alon-albalak","type":"user"},"name":"Alon Albalak","status":"claimed_verified","statusLastChangedAt":"2025-06-07T05:49:18.246Z","hidden":false},{"_id":"68417762d777f13c594dd05b","name":"Kushal Arora","hidden":false},{"_id":"68417762d777f13c594dd05c","name":"Blake Wulfe","hidden":false},{"_id":"68417762d777f13c594dd05d","user":{"_id":"631620f6894404e25068856f","avatarUrl":"/avatars/52c30caa0ee11347f82420a14ec19996.svg","isPro":false,"fullname":"Chinmay Hegde","user":"chegde","type":"user"},"name":"Chinmay Hegde","status":"claimed_verified","statusLastChangedAt":"2025-06-07T05:49:13.873Z","hidden":false},{"_id":"68417762d777f13c594dd05e","name":"Greg Durrett","hidden":false},{"_id":"68417762d777f13c594dd05f","name":"Sewoong Oh","hidden":false},{"_id":"68417762d777f13c594dd060","name":"Mohit Bansal","hidden":false},{"_id":"68417762d777f13c594dd061","name":"Saadia Gabriel","hidden":false},{"_id":"68417762d777f13c594dd062","name":"Aditya Grover","hidden":false},{"_id":"68417762d777f13c594dd063","name":"Kai-Wei Chang","hidden":false},{"_id":"68417762d777f13c594dd064","name":"Vaishaal Shankar","hidden":false},{"_id":"68417762d777f13c594dd065","name":"Aaron Gokaslan","hidden":false},{"_id":"68417762d777f13c594dd066","name":"Mike A. Merrill","hidden":false},{"_id":"68417762d777f13c594dd067","name":"Tatsunori Hashimoto","hidden":false},{"_id":"68417762d777f13c594dd068","name":"Yejin Choi","hidden":false},{"_id":"68417762d777f13c594dd069","user":{"_id":"6355b485b8b79340d4630dd5","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/6355b485b8b79340d4630dd5/HIZO4ybweRy48VdCtk2MB.jpeg","isPro":false,"fullname":"Jenia Jitsev","user":"JJitsev","type":"user"},"name":"Jenia Jitsev","status":"claimed_verified","statusLastChangedAt":"2025-06-16T07:17:50.768Z","hidden":false},{"_id":"68417762d777f13c594dd06a","name":"Reinhard Heckel","hidden":false},{"_id":"68417762d777f13c594dd06b","name":"Maheswaran Sathiamoorthy","hidden":false},{"_id":"68417762d777f13c594dd06c","name":"Alexandros G. Dimakis","hidden":false},{"_id":"68417762d777f13c594dd06d","name":"Ludwig Schmidt","hidden":false}],"mediaUrls":["https://cdn-uploads.huggingface.co/production/uploads/65a8c9dff84e045590476618/d9LCy-Rdi_X-7Brofjm1h.png","https://cdn-uploads.huggingface.co/production/uploads/65a8c9dff84e045590476618/V1vDfV2EktH_CjqbdSxLv.png"],"publishedAt":"2025-06-04T17:25:39.000Z","submittedOnDailyAt":"2025-06-05T15:46:43.237Z","title":"OpenThoughts: Data Recipes for Reasoning Models","submittedOnDailyBy":{"_id":"65a8c9dff84e045590476618","avatarUrl":"/avatars/9caef727784cda09c7bbcd373cbe3053.svg","isPro":false,"fullname":"Etash Guha","user":"EtashGuha","type":"user"},"summary":"Reasoning models have made rapid progress on many benchmarks involving math,\ncode, and science. Yet, there are still many open questions about the best\ntraining recipes for reasoning since state-of-the-art models often rely on\nproprietary datasets with little to no public information available. To address\nthis, the goal of the OpenThoughts project is to create open-source datasets\nfor training reasoning models. After initial explorations, our OpenThoughts2-1M\ndataset led to OpenThinker2-32B, the first model trained on public reasoning\ndata to match DeepSeek-R1-Distill-32B on standard reasoning benchmarks such as\nAIME and LiveCodeBench. We then improve our dataset further by systematically\ninvestigating each step of our data generation pipeline with 1,000+ controlled\nexperiments, which led to OpenThoughts3. Scaling the pipeline to 1.2M examples\nand using QwQ-32B as teacher yields our OpenThinker3-7B model, which achieves\nstate-of-the-art results: 53% on AIME 2025, 51% on LiveCodeBench 06/24-01/25,\nand 54% on GPQA Diamond. All of our datasets and models are available on\nhttps://openthoughts.ai.","upvotes":52,"discussionId":"68417763d777f13c594dd0af","projectPage":"https://openthoughts.ai","githubRepo":"https://github.com/open-thoughts/open-thoughts","githubRepoAddedBy":"auto","ai_summary":"The OpenThoughts project created open-source datasets leading to reasoning models that match or exceed state-of-the-art benchmarks in math, code, and science.","ai_keywords":["reasoning models","OpenThoughts project","OpenThoughts2-1M","OpenThinker2-32B","DeepSeek-R1-Distill-32B","standard reasoning benchmarks","AIME","LiveCodeBench","OpenThoughts3","OpenThinker3-7B","GPQA Diamond"],"githubStars":2209},"canReadDatabase":false,"canManagePapers":false,"canSubmit":false,"hasHfLevelAccess":false,"upvoted":false,"upvoters":[{"_id":"65a8c9dff84e045590476618","avatarUrl":"/avatars/9caef727784cda09c7bbcd373cbe3053.svg","isPro":false,"fullname":"Etash Guha","user":"EtashGuha","type":"user"},{"_id":"6365c1158b6bd0d9f6605d3d","avatarUrl":"/avatars/1e3e6573e0cb2bb1b9d8043045120ee6.svg","isPro":false,"fullname":"Sedrick Keh","user":"sedrickkeh","type":"user"},{"_id":"62be399f01dc22b4d22fc990","avatarUrl":"/avatars/3e1f3ed30773cbb410b8a9d3c854cac7.svg","isPro":false,"fullname":"Jean Mercat","user":"jmercat","type":"user"},{"_id":"63c47046c7d7f4c63a5dbb41","avatarUrl":"/avatars/57bc0787d6e0fc592f96fe099f0fbd5d.svg","isPro":false,"fullname":"Georgios Smyrnis","user":"gsmyrnis","type":"user"},{"_id":"63732cef759c1ca82047ba83","avatarUrl":"/avatars/6f31a48fc216b7117c722e79402073a2.svg","isPro":false,"fullname":"Negin Raoof","user":"neginr","type":"user"},{"_id":"62f7f4efe7c1c9bf10c81465","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/62f7f4efe7c1c9bf10c81465/AYlOg0fkP1o4GAP-8Y3xt.jpeg","isPro":true,"fullname":"Benjamin Feuer","user":"penfever","type":"user"},{"_id":"626c182a030a6e7363b6fe0a","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/626c182a030a6e7363b6fe0a/NaMcIOPVgJ6JFt8fojZWU.jpeg","isPro":false,"fullname":"Ryan Marten","user":"ryanmarten","type":"user"},{"_id":"647f5faa35bc6d6aa5fbfa56","avatarUrl":"/avatars/b5b2f8277872ebcad87c186eec70e282.svg","isPro":false,"fullname":"a universe of atoms","user":"atom-in-the-universe","type":"user"},{"_id":"657e03daf4f72f2c4c1dd455","avatarUrl":"/avatars/f879274d830359ea0469696c6a9437ed.svg","isPro":false,"fullname":"Ben Newman","user":"blnewman-uw","type":"user"},{"_id":"61703fa3dff0ef663e421ab5","avatarUrl":"/avatars/96172e2782e218bbbddfdf47f96c1ad4.svg","isPro":false,"fullname":"Jaehun Jung","user":"Jaehun","type":"user"},{"_id":"661595d1b3d0b21da55cde7d","avatarUrl":"/avatars/ba3fa065536518637d21a5c46cee5dd1.svg","isPro":false,"fullname":"Tatsu Hashimoto","user":"thashim","type":"user"},{"_id":"611a7ec4289467cafea62d13","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/611a7ec4289467cafea62d13/pck-0fmPQkoU7yzh6-WoL.jpeg","isPro":false,"fullname":"Alon Albalak","user":"alon-albalak","type":"user"}],"acceptLanguages":["*"],"dailyPaperRank":2}">

Papers

arxiv:2506.04178

OpenThoughts: Data Recipes for Reasoning Models

Published on Jun 4, 2025

· Submitted by

Etash Guha on Jun 5, 2025

#2 Paper of the day

Upvote

Authors:

Ryan Marten ,

Marianna Nezhurina ,

Trung Vu ,

Ashima Suvarna ,

Benjamin Feuer ,

Shreyas Pimpalgaonkar

Abstract

The OpenThoughts project created open-source datasets leading to reasoning models that match or exceed state-of-the-art benchmarks in math, code, and science.

AI-generated summary

Reasoning models have made rapid progress on many benchmarks involving math, code, and science. Yet, there are still many open questions about the best training recipes for reasoning since state-of-the-art models often rely on proprietary datasets with little to no public information available. To address this, the goal of the OpenThoughts project is to create open-source datasets for training reasoning models. After initial explorations, our OpenThoughts2-1M dataset led to OpenThinker2-32B, the first model trained on public reasoning data to match DeepSeek-R1-Distill-32B on standard reasoning benchmarks such as AIME and LiveCodeBench. We then improve our dataset further by systematically investigating each step of our data generation pipeline with 1,000+ controlled experiments, which led to OpenThoughts3. Scaling the pipeline to 1.2M examples and using QwQ-32B as teacher yields our OpenThinker3-7B model, which achieves state-of-the-art results: 53% on AIME 2025, 51% on LiveCodeBench 06/24-01/25, and 54% on GPQA Diamond. All of our datasets and models are available on https://openthoughts.ai.

View arXiv page View PDF Project page GitHub 2.21k auto Add to collection

Community

EtashGuha

Paper submitter Jun 5, 2025

Reasoning models have made rapid progress on many benchmarks involving math, code, and science. Yet, there are still many open questions about the best training recipes for reasoning since state-of-the-art models often rely on proprietary datasets with little to no public information available. To address this, the goal of the OpenThoughts project is to create open-source datasets for training reasoning models. After initial explorations, our OpenThoughts2-1M dataset led to OpenThinker2-32B, the first model trained on public reasoning data to match DeepSeek-R1-Distill-32B on standard reasoning benchmarks such as AIME and LiveCodeBench. We then improve our dataset further by systematically investigating each step of our data generation pipeline with 1,000+ experiments, which led to OpenThoughts3. Scaling the pipeline to 1.2M examples and using QwQ-32B as teacher yields our OpenThinker3-7B model, which achieves state-of-the-art results: 53% on AIME 2025, 51% on LiveCodeBench 06/24-01/25, and 54% on GPQA Diamond. All of our datasets and models are available on openthoughts.ai.