{"id":"arxiv:2601.14525","title":"Towards Execution-Grounded Automated AI Research","date":"2026-01-20","section":"Post-training","quality":0.042877,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/895006e818735c9a565dbe99ae31404366eefd6bdc2278612271c65a8493e4a5/","source_url":"https://arxiv.org/abs/2601.14525","text_available":true,"authors":["Chenglei Si","Zitong Yang","Yejin Choi"],"outgoing":[{"source":"arxiv:2601.14525","target":"arxiv:2302.06675","evidence":"title","id":"arxiv:2302.06675","title":"Symbolic Discovery of Optimization Algorithms","url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/3438564652cae2c199823d6d3456a9bf87703e715521d9faaf396c18621f4a5a/"},{"source":"arxiv:2601.14525","target":"arxiv:2402.03300","evidence":"title","id":"arxiv:2402.03300","title":"DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models","url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/636fb871904bc06f742cfdea069796f9e9665f95bd58dc63260eaff5154ef985/"},{"source":"arxiv:2601.14525","target":"arxiv:2408.08435","evidence":"title","id":"arxiv:2408.08435","title":"Automated Design of Agentic Systems","url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/c9890d5668d1c1b33e29569b74626fd586f456af6a97c657b8c99895c90f59a5/"},{"source":"arxiv:2601.14525","target":"arxiv:2410.07095","evidence":"title","id":"arxiv:2410.07095","title":"MLE-bench: Evaluating Machine Learning Agents on Machine Learning Engineering","url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f0a0c6321ec079098a36ad19b3900d77197a3af1f6bbf6eee8a083e5eb03d42e/"},{"source":"arxiv:2601.14525","target":"arxiv:2504.13837","evidence":"title","id":"arxiv:2504.13837","title":"Does Reinforcement Learning Really Incentivize Reasoning Capacity in LLMs Beyond the Base Model?","url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/b90f70269aeb02311d74ffaadf47e139683424a47118dc18619000ba72174d17/"},{"source":"arxiv:2601.14525","target":"arxiv:2504.20571","evidence":"title","id":"arxiv:2504.20571","title":"Reinforcement Learning for Reasoning in Large Language Models with One Training Example","url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/b2b21c3df49619588c371411dba9467a41f81b3fd4da5441898b5145bb0de120/"},{"source":"arxiv:2601.14525","target":"arxiv:2507.18074","evidence":"title","id":"arxiv:2507.18074","title":"AlphaGo Moment for Model Architecture Discovery","url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/1c22da54910a26f1f81b276f5cae3a81e22f7db024f7d1a5fa89ccd4f291ffa8/"}],"incoming":[],"text_parts":[{"start":0,"end":15323,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/895006e818735c9a565dbe99ae31404366eefd6bdc2278612271c65a8493e4a5/part-1.html"},{"start":15323,"end":31065,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/895006e818735c9a565dbe99ae31404366eefd6bdc2278612271c65a8493e4a5/part-2.html"},{"start":31065,"end":47021,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/895006e818735c9a565dbe99ae31404366eefd6bdc2278612271c65a8493e4a5/part-3.html"},{"start":47021,"end":48825,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/895006e818735c9a565dbe99ae31404366eefd6bdc2278612271c65a8493e4a5/part-4.html"}],"contents":[{"title":"Towards Execution-Grounded Automated AI Research","offset":0,"level":1,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/895006e818735c9a565dbe99ae31404366eefd6bdc2278612271c65a8493e4a5/part-1.html#text"},{"title":"Abstract","offset":52,"level":6,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/895006e818735c9a565dbe99ae31404366eefd6bdc2278612271c65a8493e4a5/part-1.html#text"},{"title":"Keywords:","offset":1700,"level":6,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/895006e818735c9a565dbe99ae31404366eefd6bdc2278612271c65a8493e4a5/part-1.html#text"},{"title":"1 Introduction","offset":1718,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/895006e818735c9a565dbe99ae31404366eefd6bdc2278612271c65a8493e4a5/part-1.html#text"},{"title":"2 Automated Idea Executor","offset":6122,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/895006e818735c9a565dbe99ae31404366eefd6bdc2278612271c65a8493e4a5/part-1.html#text"},{"title":"2.1 Research Environments for Ideation","offset":6427,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/895006e818735c9a565dbe99ae31404366eefd6bdc2278612271c65a8493e4a5/part-1.html#text"},{"title":"2.2 System design","offset":9697,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/895006e818735c9a565dbe99ae31404366eefd6bdc2278612271c65a8493e4a5/part-1.html#text"},{"title":"Implementer","offset":10256,"level":4,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/895006e818735c9a565dbe99ae31404366eefd6bdc2278612271c65a8493e4a5/part-1.html#text"},{"title":"Scheduler","offset":11144,"level":4,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/895006e818735c9a565dbe99ae31404366eefd6bdc2278612271c65a8493e4a5/part-1.html#text"},{"title":"Worker","offset":11419,"level":4,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/895006e818735c9a565dbe99ae31404366eefd6bdc2278612271c65a8493e4a5/part-1.html#text"},{"title":"3 Benchmarking LLM Ideators and Executors","offset":12082,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/895006e818735c9a565dbe99ae31404366eefd6bdc2278612271c65a8493e4a5/part-1.html#text"},{"title":"3.1 End-to-End Ideation and Execution","offset":12433,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/895006e818735c9a565dbe99ae31404366eefd6bdc2278612271c65a8493e4a5/part-1.html#text"},{"title":"3.2 Comparing Ideators with the Same Executor","offset":13650,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/895006e818735c9a565dbe99ae31404366eefd6bdc2278612271c65a8493e4a5/part-1.html#text"},{"title":"4 Execution-Guided Evolutionary Search","offset":14780,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/895006e818735c9a565dbe99ae31404366eefd6bdc2278612271c65a8493e4a5/part-1.html#text"},{"title":"4.1 Search Scaffold","offset":15298,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/895006e818735c9a565dbe99ae31404366eefd6bdc2278612271c65a8493e4a5/part-1.html#text"},{"title":"4.2 Experiment Results","offset":16365,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/895006e818735c9a565dbe99ae31404366eefd6bdc2278612271c65a8493e4a5/part-2.html#text"},{"title":"4.3 Comparison with Best-of-N","offset":18933,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/895006e818735c9a565dbe99ae31404366eefd6bdc2278612271c65a8493e4a5/part-2.html#text"},{"title":"4.4 Analysis of Generated Ideas","offset":19732,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/895006e818735c9a565dbe99ae31404366eefd6bdc2278612271c65a8493e4a5/part-2.html#text"},{"title":"5 Reinforcement Learning from Execution Reward","offset":22561,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/895006e818735c9a565dbe99ae31404366eefd6bdc2278612271c65a8493e4a5/part-2.html#text"},{"title":"5.1 Reward Design and Experiment Setup","offset":23180,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/895006e818735c9a565dbe99ae31404366eefd6bdc2278612271c65a8493e4a5/part-2.html#text"},{"title":"5.2 Experiment Results","offset":24969,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/895006e818735c9a565dbe99ae31404366eefd6bdc2278612271c65a8493e4a5/part-2.html#text"},{"title":"Positive Training Curves for Average Reward","offset":24997,"level":4,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/895006e818735c9a565dbe99ae31404366eefd6bdc2278612271c65a8493e4a5/part-2.html#text"},{"title":"The Case of Max Reward","offset":25850,"level":4,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/895006e818735c9a565dbe99ae31404366eefd6bdc2278612271c65a8493e4a5/part-2.html#text"},{"title":"5.3 Analysis of Training Dynamics","offset":26768,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/895006e818735c9a565dbe99ae31404366eefd6bdc2278612271c65a8493e4a5/part-2.html#text"},{"title":"Thinking Length","offset":26807,"level":4,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/895006e818735c9a565dbe99ae31404366eefd6bdc2278612271c65a8493e4a5/part-2.html#text"},{"title":"Diversity Collapse","offset":27897,"level":4,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/895006e818735c9a565dbe99ae31404366eefd6bdc2278612271c65a8493e4a5/part-2.html#text"},{"title":"6 Related Work","offset":29748,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/895006e818735c9a565dbe99ae31404366eefd6bdc2278612271c65a8493e4a5/part-2.html#text"},{"title":"AutoML","offset":29767,"level":4,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/895006e818735c9a565dbe99ae31404366eefd6bdc2278612271c65a8493e4a5/part-2.html#text"},{"title":"LLM-based Research Agents","offset":31033,"level":4,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/895006e818735c9a565dbe99ae31404366eefd6bdc2278612271c65a8493e4a5/part-2.html#text"},{"title":"AI for Research","offset":32318,"level":4,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/895006e818735c9a565dbe99ae31404366eefd6bdc2278612271c65a8493e4a5/part-3.html#text"},{"title":"Execution Grounding for Code","offset":33108,"level":4,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/895006e818735c9a565dbe99ae31404366eefd6bdc2278612271c65a8493e4a5/part-3.html#text"},{"title":"7 Conclusion","offset":33712,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/895006e818735c9a565dbe99ae31404366eefd6bdc2278612271c65a8493e4a5/part-3.html#text"},{"title":"8 Discussion","offset":34710,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/895006e818735c9a565dbe99ae31404366eefd6bdc2278612271c65a8493e4a5/part-3.html#text"},{"title":"Acknowledgment","offset":36830,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/895006e818735c9a565dbe99ae31404366eefd6bdc2278612271c65a8493e4a5/part-3.html#text"},{"title":"Appendix A Appendix","offset":37699,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/895006e818735c9a565dbe99ae31404366eefd6bdc2278612271c65a8493e4a5/part-3.html#text"},{"title":"A.1 Other RL Attempts","offset":37723,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/895006e818735c9a565dbe99ae31404366eefd6bdc2278612271c65a8493e4a5/part-3.html#text"},{"title":"Attempt 1: Dynamic Prompt","offset":37831,"level":4,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/895006e818735c9a565dbe99ae31404366eefd6bdc2278612271c65a8493e4a5/part-3.html#text"},{"title":"Attempt 2: Length Reward","offset":38296,"level":4,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/895006e818735c9a565dbe99ae31404366eefd6bdc2278612271c65a8493e4a5/part-3.html#text"},{"title":"Attempt 3: Diversity Reward","offset":38927,"level":4,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/895006e818735c9a565dbe99ae31404366eefd6bdc2278612271c65a8493e4a5/part-3.html#text"},{"title":"A.2 Additional Idea Examples","offset":39791,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/895006e818735c9a565dbe99ae31404366eefd6bdc2278612271c65a8493e4a5/part-3.html#text"},{"title":"A.3 Code Execution Examples","offset":43683,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/895006e818735c9a565dbe99ae31404366eefd6bdc2278612271c65a8493e4a5/part-3.html#text"},{"title":"Instructions for reporting errors","offset":47861,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/895006e818735c9a565dbe99ae31404366eefd6bdc2278612271c65a8493e4a5/part-4.html#text"}],"text_code_points":48825,"source_markdown":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/895006e818735c9a565dbe99ae31404366eefd6bdc2278612271c65a8493e4a5/source.md"}