{"id":"arxiv:2609.14858","title":"Dream-RSI: Recursive Self-Improvement through Evolving Worlds","date":"2026-09-14","section":"Agents, open-endedness, AGI","quality":null,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/4c9ce361df4d9a5f89f3dc0a70a71e795eb66f2c4e3c262f73f04b2d117c8ea1/","source_url":"https://arxiv.org/abs/2609.14858","text_available":true,"authors":["Tong Zheng","Xidong Wu","Zheng Zhang","Zhankui He"],"outgoing":[{"source":"arxiv:2609.14858","target":"arxiv:2301.04104","evidence":"id","id":"arxiv:2301.04104","title":"Mastering Diverse Domains through World Models","url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/266d59e63d2f705efdc2eb7ad93f5126b7092b4450efd0d8c17099c51574b176/"},{"source":"arxiv:2609.14858","target":"arxiv:2505.22954","evidence":"title","id":"arxiv:2505.22954","title":"Darwin Godel Machine: Open-Ended Evolution of Self-Improving Agents","url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/c29f716871f80314868b5647692de387f26e12166b93bd362915fd272b6e081c/"},{"source":"arxiv:2609.14858","target":"arxiv:2601.16175","evidence":"id","id":"arxiv:2601.16175","title":"Learning to Discover at Test Time","url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/5d0b45cbbc72945765e689cbfb7ac7e4e92dfbb764861b27682ac0744ef05b70/"},{"source":"arxiv:2609.14858","target":"arxiv:2603.19461","evidence":"id","id":"arxiv:2603.19461","title":"Hyperagents","url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/17eec8a15debc212b5530f117dd8ba8bad339590729cbbc4f75a47755ebcb544/"},{"source":"arxiv:2609.14858","target":"arxiv:2603.28052","evidence":"id","id":"arxiv:2603.28052","title":"Meta-Harness: End-to-End Optimization of Model Harnesses","url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/8bb0b87384dfdd42c419cf26cc7e52f1f0d89f63afa5d5968e246b73a44c99cb/"},{"source":"arxiv:2609.14858","target":"arxiv:2604.19341","evidence":"id","id":"arxiv:2604.19341","title":"Structured Scaling of AI Discovery Across Diverse Scientific Domains","url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/33ad3375247f86197210840a19064548b997e8617475e606d7a7ac2cda64740e/"},{"source":"arxiv:2609.14858","target":"doi:10.1038/s41586-023-06924-6","evidence":"title","id":"doi:10.1038/s41586-023-06924-6","title":"Mathematical discoveries from program search with large language models","url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/70a6b427d42fd28c9d538648e148a97db2fb2614c7764f3231e53f4bebdce887/"},{"source":"arxiv:2609.14858","target":"url:sakana.ai/shinka-evolve","evidence":"title","id":"url:sakana.ai/shinka-evolve","title":"ShinkaEvolve: Towards Open-Ended And Sample-Efficient Program Evolution","url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/a13eace17f3407cb1634b579b2738f15f1e035897a8f1842305198c6d2bf35e1/"}],"incoming":[],"text_parts":[{"start":0,"end":15919,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/4c9ce361df4d9a5f89f3dc0a70a71e795eb66f2c4e3c262f73f04b2d117c8ea1/part-1.html"},{"start":15919,"end":31871,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/4c9ce361df4d9a5f89f3dc0a70a71e795eb66f2c4e3c262f73f04b2d117c8ea1/part-2.html"},{"start":31871,"end":39002,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/4c9ce361df4d9a5f89f3dc0a70a71e795eb66f2c4e3c262f73f04b2d117c8ea1/part-3.html"}],"contents":[{"title":"Dream-RSI: Recursive Self-Improvement through Evolving Worlds","offset":10,"level":1,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/4c9ce361df4d9a5f89f3dc0a70a71e795eb66f2c4e3c262f73f04b2d117c8ea1/part-1.html#text"},{"title":"Abstract","offset":75,"level":6,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/4c9ce361df4d9a5f89f3dc0a70a71e795eb66f2c4e3c262f73f04b2d117c8ea1/part-1.html#text"},{"title":"1 Introduction","offset":1639,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/4c9ce361df4d9a5f89f3dc0a70a71e795eb66f2c4e3c262f73f04b2d117c8ea1/part-1.html#text"},{"title":"2 Motivation: Discovery History as a Replay Simulator","offset":8213,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/4c9ce361df4d9a5f89f3dc0a70a71e795eb66f2c4e3c262f73f04b2d117c8ea1/part-1.html#text"},{"title":"3 Dream-RSI : Recursive Self-Improvement through Evolving Worlds","offset":10649,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/4c9ce361df4d9a5f89f3dc0a70a71e795eb66f2c4e3c262f73f04b2d117c8ea1/part-1.html#text"},{"title":"Discovery trees and the shared decision interface.","offset":11468,"level":4,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/4c9ce361df4d9a5f89f3dc0a70a71e795eb66f2c4e3c262f73f04b2d117c8ea1/part-1.html#text"},{"title":"Online rollout.","offset":13279,"level":4,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/4c9ce361df4d9a5f89f3dc0a70a71e795eb66f2c4e3c262f73f04b2d117c8ea1/part-1.html#text"},{"title":"Offline evaluation.","offset":15185,"level":4,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/4c9ce361df4d9a5f89f3dc0a70a71e795eb66f2c4e3c262f73f04b2d117c8ea1/part-1.html#text"},{"title":"Replay objective.","offset":18704,"level":4,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/4c9ce361df4d9a5f89f3dc0a70a71e795eb66f2c4e3c262f73f04b2d117c8ea1/part-2.html#text"},{"title":"Policy improvement and selection.","offset":19908,"level":4,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/4c9ce361df4d9a5f89f3dc0a70a71e795eb66f2c4e3c262f73f04b2d117c8ea1/part-2.html#text"},{"title":"4 Experiments","offset":21361,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/4c9ce361df4d9a5f89f3dc0a70a71e795eb66f2c4e3c262f73f04b2d117c8ea1/part-2.html#text"},{"title":"4.1 Algorithm Engineering","offset":23553,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/4c9ce361df4d9a5f89f3dc0a70a71e795eb66f2c4e3c262f73f04b2d117c8ea1/part-2.html#text"},{"title":"Baselines and Setup.","offset":24451,"level":4,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/4c9ce361df4d9a5f89f3dc0a70a71e795eb66f2c4e3c262f73f04b2d117c8ea1/part-2.html#text"},{"title":"Main Results.","offset":24870,"level":4,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/4c9ce361df4d9a5f89f3dc0a70a71e795eb66f2c4e3c262f73f04b2d117c8ea1/part-2.html#text"},{"title":"Recursive Discovery Dynamics.","offset":26011,"level":4,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/4c9ce361df4d9a5f89f3dc0a70a71e795eb66f2c4e3c262f73f04b2d117c8ea1/part-2.html#text"},{"title":"Discovered Solver Analysis.","offset":26692,"level":4,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/4c9ce361df4d9a5f89f3dc0a70a71e795eb66f2c4e3c262f73f04b2d117c8ea1/part-2.html#text"},{"title":"4.2 Mathematics Optimization","offset":27399,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/4c9ce361df4d9a5f89f3dc0a70a71e795eb66f2c4e3c262f73f04b2d117c8ea1/part-2.html#text"},{"title":"Results.","offset":28600,"level":4,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/4c9ce361df4d9a5f89f3dc0a70a71e795eb66f2c4e3c262f73f04b2d117c8ea1/part-2.html#text"},{"title":"4.3 Kernel Engineering","offset":29341,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/4c9ce361df4d9a5f89f3dc0a70a71e795eb66f2c4e3c262f73f04b2d117c8ea1/part-2.html#text"},{"title":"Results.","offset":30341,"level":4,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/4c9ce361df4d9a5f89f3dc0a70a71e795eb66f2c4e3c262f73f04b2d117c8ea1/part-2.html#text"},{"title":"5 Further Analysis","offset":30894,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/4c9ce361df4d9a5f89f3dc0a70a71e795eb66f2c4e3c262f73f04b2d117c8ea1/part-2.html#text"},{"title":"5.1 Analysis of Historical Inductive Biases in Long-Horizon Discovery","offset":30917,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/4c9ce361df4d9a5f89f3dc0a70a71e795eb66f2c4e3c262f73f04b2d117c8ea1/part-2.html#text"},{"title":"5.2 Analysis of Evolution of Exploration Behavior","offset":31871,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/4c9ce361df4d9a5f89f3dc0a70a71e795eb66f2c4e3c262f73f04b2d117c8ea1/part-3.html#text"},{"title":"6 Related Work","offset":32399,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/4c9ce361df4d9a5f89f3dc0a70a71e795eb66f2c4e3c262f73f04b2d117c8ea1/part-3.html#text"},{"title":"AI-Driven Scientific and Algorithmic Discovery.","offset":32418,"level":4,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/4c9ce361df4d9a5f89f3dc0a70a71e795eb66f2c4e3c262f73f04b2d117c8ea1/part-3.html#text"},{"title":"Self-Evolving Agents.","offset":33415,"level":4,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/4c9ce361df4d9a5f89f3dc0a70a71e795eb66f2c4e3c262f73f04b2d117c8ea1/part-3.html#text"},{"title":"Memory, History, and Experience Reuse.","offset":34769,"level":4,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/4c9ce361df4d9a5f89f3dc0a70a71e795eb66f2c4e3c262f73f04b2d117c8ea1/part-3.html#text"},{"title":"7 Conclusion","offset":35808,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/4c9ce361df4d9a5f89f3dc0a70a71e795eb66f2c4e3c262f73f04b2d117c8ea1/part-3.html#text"},{"title":"Appendix A Detailed Task Description","offset":36739,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/4c9ce361df4d9a5f89f3dc0a70a71e795eb66f2c4e3c262f73f04b2d117c8ea1/part-3.html#text"},{"title":"Appendix B Prompts","offset":36780,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/4c9ce361df4d9a5f89f3dc0a70a71e795eb66f2c4e3c262f73f04b2d117c8ea1/part-3.html#text"},{"title":"B.1 Exploration Prompt","offset":37027,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/4c9ce361df4d9a5f89f3dc0a70a71e795eb66f2c4e3c262f73f04b2d117c8ea1/part-3.html#text"},{"title":"B.2 Replay-Based Policy Improvement Prompt","offset":37372,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/4c9ce361df4d9a5f89f3dc0a70a71e795eb66f2c4e3c262f73f04b2d117c8ea1/part-3.html#text"},{"title":"Appendix C Discovered Programs","offset":37700,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/4c9ce361df4d9a5f89f3dc0a70a71e795eb66f2c4e3c262f73f04b2d117c8ea1/part-3.html#text"},{"title":"Instructions for reporting errors","offset":38038,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/4c9ce361df4d9a5f89f3dc0a70a71e795eb66f2c4e3c262f73f04b2d117c8ea1/part-3.html#text"}],"text_code_points":39002,"source_markdown":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/4c9ce361df4d9a5f89f3dc0a70a71e795eb66f2c4e3c262f73f04b2d117c8ea1/source.md"}