{"id":"arxiv:2501.04519","title":"rStar-Math: Small LLMs Can Master Math Reasoning with Self-Evolved Deep Thinking","date":"2025-01-08","section":"Reasoning and the \"physics\" of language models","quality":0.393643,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/473ad7b5a354658a1d241813741e15a8ab51e7c2ea34f0c3d22eb06884864a46/","source_url":"https://arxiv.org/abs/2501.04519","text_available":true,"authors":["Xinyu Guan","Li Lyna Zhang","Yifei Liu","Ning Shang"],"outgoing":[{"source":"arxiv:2501.04519","target":"arxiv:2203.11171","evidence":"title","id":"arxiv:2203.11171","title":"Self-Consistency Improves Chain of Thought Reasoning in Language Models","url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/02a15b098d96a94d463bb470f59a34723abc34e59196f35c539f9796d8f3c3df/"},{"source":"arxiv:2501.04519","target":"arxiv:2407.21783","evidence":"id","id":"arxiv:2407.21783","title":"The Llama 3 Herd of Models","url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/de3cebb4c28bc5d41a843279f219dca14476284991ea5c7c0aeab7068a40e306/"}],"incoming":[{"source":"arxiv:2506.01939","target":"arxiv:2501.04519","evidence":"id","id":"arxiv:2506.01939","title":"Beyond the 80/20 Rule: High-Entropy Minority Tokens Drive Effective Reinforcement Learning for LLM Reasoning","url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/e0f4a55bca612bce5895b6b98f89b0efc14cbd8cf45946a7743347cf6daee1a1/"},{"source":"arxiv:2606.18810","target":"arxiv:2501.04519","evidence":"title","id":"arxiv:2606.18810","title":"Learning from Own Solutions: Self-Conditioned Credit Assignment for Reinforcement Learning with Verifiable Rewards","url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/9c11d77c6d08392f24aab79c8b45a3a3b0f4b743dd35e150a83ee7f6fa315248/"}],"text_parts":[{"start":0,"end":15428,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/473ad7b5a354658a1d241813741e15a8ab51e7c2ea34f0c3d22eb06884864a46/part-1.html"},{"start":15428,"end":31179,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/473ad7b5a354658a1d241813741e15a8ab51e7c2ea34f0c3d22eb06884864a46/part-2.html"},{"start":31179,"end":46839,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/473ad7b5a354658a1d241813741e15a8ab51e7c2ea34f0c3d22eb06884864a46/part-3.html"},{"start":46839,"end":53567,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/473ad7b5a354658a1d241813741e15a8ab51e7c2ea34f0c3d22eb06884864a46/part-4.html"}],"contents":[{"title":"rStar-Math: Small LLMs Can Master Math Reasoning with Self-Evolved Deep Thinking","offset":0,"level":1,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/473ad7b5a354658a1d241813741e15a8ab51e7c2ea34f0c3d22eb06884864a46/part-1.html#text"},{"title":"Abstract","offset":84,"level":6,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/473ad7b5a354658a1d241813741e15a8ab51e7c2ea34f0c3d22eb06884864a46/part-1.html#text"},{"title":"1 Introduction","offset":1592,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/473ad7b5a354658a1d241813741e15a8ab51e7c2ea34f0c3d22eb06884864a46/part-1.html#text"},{"title":"2 Related Works","offset":8384,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/473ad7b5a354658a1d241813741e15a8ab51e7c2ea34f0c3d22eb06884864a46/part-1.html#text"},{"title":"3 Methodology","offset":11243,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/473ad7b5a354658a1d241813741e15a8ab51e7c2ea34f0c3d22eb06884864a46/part-1.html#text"},{"title":"3.1 Design Choices","offset":11261,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/473ad7b5a354658a1d241813741e15a8ab51e7c2ea34f0c3d22eb06884864a46/part-1.html#text"},{"title":"3.2 Step-by-Step Verified Reasoning Trajectory","offset":14262,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/473ad7b5a354658a1d241813741e15a8ab51e7c2ea34f0c3d22eb06884864a46/part-1.html#text"},{"title":"3.3 Process Preference Model","offset":20603,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/473ad7b5a354658a1d241813741e15a8ab51e7c2ea34f0c3d22eb06884864a46/part-2.html#text"},{"title":"3.4 Self-Evolved Deep Thinking","offset":23976,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/473ad7b5a354658a1d241813741e15a8ab51e7c2ea34f0c3d22eb06884864a46/part-2.html#text"},{"title":"3.4.1 Training with Step-by-Step Verified Reasoning Trajectory","offset":24012,"level":4,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/473ad7b5a354658a1d241813741e15a8ab51e7c2ea34f0c3d22eb06884864a46/part-2.html#text"},{"title":"3.4.2 Recipe for Self-Evolution","offset":26879,"level":4,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/473ad7b5a354658a1d241813741e15a8ab51e7c2ea34f0c3d22eb06884864a46/part-2.html#text"},{"title":"4 Evaluation","offset":30396,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/473ad7b5a354658a1d241813741e15a8ab51e7c2ea34f0c3d22eb06884864a46/part-2.html#text"},{"title":"4.1 Setup","offset":30413,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/473ad7b5a354658a1d241813741e15a8ab51e7c2ea34f0c3d22eb06884864a46/part-2.html#text"},{"title":"4.2 Main Results","offset":33744,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/473ad7b5a354658a1d241813741e15a8ab51e7c2ea34f0c3d22eb06884864a46/part-3.html#text"},{"title":"4.3 Ablation Study and Analysis","offset":36804,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/473ad7b5a354658a1d241813741e15a8ab51e7c2ea34f0c3d22eb06884864a46/part-3.html#text"},{"title":"5 Findings and Discussions","offset":41037,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/473ad7b5a354658a1d241813741e15a8ab51e7c2ea34f0c3d22eb06884864a46/part-3.html#text"},{"title":"6 Conclusion","offset":45339,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/473ad7b5a354658a1d241813741e15a8ab51e7c2ea34f0c3d22eb06884864a46/part-3.html#text"},{"title":"Acknowledgement","offset":46234,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/473ad7b5a354658a1d241813741e15a8ab51e7c2ea34f0c3d22eb06884864a46/part-3.html#text"},{"title":"Appendix A Appendix","offset":46771,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/473ad7b5a354658a1d241813741e15a8ab51e7c2ea34f0c3d22eb06884864a46/part-3.html#text"},{"title":"A.1 Additional Experiments and Details","offset":46795,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/473ad7b5a354658a1d241813741e15a8ab51e7c2ea34f0c3d22eb06884864a46/part-3.html#text"},{"title":"A.2 Examples of Reasoning Trajectories","offset":52534,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/473ad7b5a354658a1d241813741e15a8ab51e7c2ea34f0c3d22eb06884864a46/part-4.html#text"},{"title":"A.3 Prompt Examples","offset":52578,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/473ad7b5a354658a1d241813741e15a8ab51e7c2ea34f0c3d22eb06884864a46/part-4.html#text"},{"title":"Instructions for reporting errors","offset":52603,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/473ad7b5a354658a1d241813741e15a8ab51e7c2ea34f0c3d22eb06884864a46/part-4.html#text"}],"text_code_points":53567,"source_markdown":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/473ad7b5a354658a1d241813741e15a8ab51e7c2ea34f0c3d22eb06884864a46/source.md"}