{"id":"arxiv:2603.19461","title":"Hyperagents","date":"2026-03-19","section":"Agents, open-endedness, AGI","quality":0.02014,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/17eec8a15debc212b5530f117dd8ba8bad339590729cbbc4f75a47755ebcb544/","source_url":"https://arxiv.org/abs/2603.19461","text_available":true,"authors":["Jenny Zhang","Bingchen Zhao","Wannan Yang","Jakob Foerster"],"outgoing":[{"source":"arxiv:2603.19461","target":"arxiv:1905.10985","evidence":"id","id":"arxiv:1905.10985","title":"AI-GAs: AI-generating algorithms, an alternate paradigm for producing general artificial intelligence","url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/089142f7297bed414d01b82c78dc888e4018fecff7efbe5cae66442e03f87c4b/"},{"source":"arxiv:2603.19461","target":"arxiv:2309.16797","evidence":"id","id":"arxiv:2309.16797","title":"Promptbreeder: Self-Referential Self-Improvement Via Prompt Evolution","url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/daac3bcfcd7f39084d995905e9ef96ae01c5aab7bfe8e565bd7622b4e44088f7/"},{"source":"arxiv:2603.19461","target":"arxiv:2402.16822","evidence":"title","id":"arxiv:2402.16822","title":"Rainbow Teaming: Open-Ended Generation of Diverse Adversarial Prompts","url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/d4803e891a4b9c929eca007e3c3f003c330d2fc741109975c6ddd7d2c32828fc/"},{"source":"arxiv:2603.19461","target":"arxiv:2402.16823","evidence":"title","id":"arxiv:2402.16823","title":"Language Agents as Optimizable Graphs","url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/750bd3807bf9c9a9b5cd2d60cc14387431aec3fe8a3f942916ca5d8922b2a5b0/"},{"source":"arxiv:2603.19461","target":"arxiv:2406.04268","evidence":"id","id":"arxiv:2406.04268","title":"Open-Endedness is Essential for Artificial Superhuman Intelligence","url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/13964b2db9a9f793d1fcae07ca83fbe09f65454214269cac56777cb8be423a4e/"},{"source":"arxiv:2603.19461","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:2603.19461","target":"arxiv:2505.22954","evidence":"id","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:2603.19461","target":"arxiv:2506.10139","evidence":"id","id":"arxiv:2506.10139","title":"Unsupervised Elicitation of Language Models","url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/f27ae3b426a87310071610dedea4c144b50b1be183434e32486b31562f9a3da9/"},{"source":"arxiv:2603.19461","target":"arxiv:2511.15593","evidence":"id","id":"arxiv:2511.15593","title":"What Does It Take to Be a Good AI Research Agent? Studying the Role of Ideation Diversity","url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/be2cc7f3c48da553c3ae6c2de2a5474a30deb993704a409d471ac0570b91dcab/"},{"source":"arxiv:2603.19461","target":"arxiv:2512.18552","evidence":"id","id":"arxiv:2512.18552","title":"Toward Training Superintelligent Software Agents through Self-Play SWE-RL","url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/da99b397e01b725186243c38a2dc6072d3a5c81bd8f759e898f04ad09aa0f4b1/"},{"source":"arxiv:2603.19461","target":"arxiv:2512.24601","evidence":"id","id":"arxiv:2512.24601","title":"Recursive Language Models","url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/ac41983c934a3432ce7b6531e367b9e639bddbed93d4c6513781e76407d9a8bd/"},{"source":"arxiv:2603.19461","target":"arxiv:2601.21557","evidence":"id","id":"arxiv:2601.21557","title":"Meta Context Engineering via Agentic Skill Evolution","url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/bedaeecadf58846f4728c50c03396fc96ba18429ecf14052b4c5e177394a246e/"},{"source":"arxiv:2603.19461","target":"arxiv:2602.07755","evidence":"id","id":"arxiv:2602.07755","title":"Learning to Continually Learn via Meta-learning Agentic Memory Designs","url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/4d0a0ea160495ea50583e0e24027b6085e2f430a44195e3053ba69cda8cc82ea/"},{"source":"arxiv:2603.19461","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/"}],"incoming":[{"source":"arxiv:2605.13821","target":"arxiv:2603.19461","evidence":"id","id":"arxiv:2605.13821","title":"Harnessing Agentic Evolution","url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/1f1060a507e2bf439d2934d610827d1b3177d85c90c1a2cab483608a56de5138/"},{"source":"arxiv:2608.19197","target":"arxiv:2603.19461","evidence":"id","id":"arxiv:2608.19197","title":"SPADE: Self-Play in Adaptive Synthetic Executable Environments","url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/c524376d6034b5d79661d6f1a444591404f4a250bcb0790d802c05e02f8249df/"},{"source":"arxiv:2608.23875","target":"arxiv:2603.19461","evidence":"id","id":"arxiv:2608.23875","title":"AI Finds A Way","url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/afe34dd5ee58c2e9739f9651349c4c9180d95dabff51994a13339f48dfd0bb27/"},{"source":"arxiv:2609.14858","target":"arxiv:2603.19461","evidence":"id","id":"arxiv:2609.14858","title":"Dream-RSI: Recursive Self-Improvement through Evolving Worlds","url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/4c9ce361df4d9a5f89f3dc0a70a71e795eb66f2c4e3c262f73f04b2d117c8ea1/"},{"source":"url:lilianweng.github.io/posts/2026-07-04-harness","target":"arxiv:2603.19461","evidence":"id","id":"url:lilianweng.github.io/posts/2026-07-04-harness","title":"Harnesses for models (Lilian Weng, 2026)","url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/c1ea249334ba128a9fff468e25fa4f8d13d4cccd13661ee130bec020ebb6463b/"}],"text_parts":[{"start":0,"end":15736,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/17eec8a15debc212b5530f117dd8ba8bad339590729cbbc4f75a47755ebcb544/part-1.html"},{"start":15736,"end":31044,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/17eec8a15debc212b5530f117dd8ba8bad339590729cbbc4f75a47755ebcb544/part-2.html"},{"start":31044,"end":46692,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/17eec8a15debc212b5530f117dd8ba8bad339590729cbbc4f75a47755ebcb544/part-3.html"},{"start":46692,"end":61860,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/17eec8a15debc212b5530f117dd8ba8bad339590729cbbc4f75a47755ebcb544/part-4.html"},{"start":61860,"end":77547,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/17eec8a15debc212b5530f117dd8ba8bad339590729cbbc4f75a47755ebcb544/part-5.html"},{"start":77547,"end":93356,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/17eec8a15debc212b5530f117dd8ba8bad339590729cbbc4f75a47755ebcb544/part-6.html"},{"start":93356,"end":98124,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/17eec8a15debc212b5530f117dd8ba8bad339590729cbbc4f75a47755ebcb544/part-7.html"}],"contents":[{"title":"HyperAgents","offset":97,"level":1,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/17eec8a15debc212b5530f117dd8ba8bad339590729cbbc4f75a47755ebcb544/part-1.html#text"},{"title":"Abstract","offset":112,"level":6,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/17eec8a15debc212b5530f117dd8ba8bad339590729cbbc4f75a47755ebcb544/part-1.html#text"},{"title":"1 Introduction","offset":2563,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/17eec8a15debc212b5530f117dd8ba8bad339590729cbbc4f75a47755ebcb544/part-1.html#text"},{"title":"2 Related Work","offset":8652,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/17eec8a15debc212b5530f117dd8ba8bad339590729cbbc4f75a47755ebcb544/part-1.html#text"},{"title":"3 Methods","offset":14798,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/17eec8a15debc212b5530f117dd8ba8bad339590729cbbc4f75a47755ebcb544/part-1.html#text"},{"title":"4 Experiment Setup","offset":20413,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/17eec8a15debc212b5530f117dd8ba8bad339590729cbbc4f75a47755ebcb544/part-2.html#text"},{"title":"4.1 Baselines","offset":21570,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/17eec8a15debc212b5530f117dd8ba8bad339590729cbbc4f75a47755ebcb544/part-2.html#text"},{"title":"4.2 Domains","offset":23925,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/17eec8a15debc212b5530f117dd8ba8bad339590729cbbc4f75a47755ebcb544/part-2.html#text"},{"title":"5 Results","offset":29829,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/17eec8a15debc212b5530f117dd8ba8bad339590729cbbc4f75a47755ebcb544/part-2.html#text"},{"title":"5.1 Improving Task Performance","offset":31008,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/17eec8a15debc212b5530f117dd8ba8bad339590729cbbc4f75a47755ebcb544/part-2.html#text"},{"title":"5.2 Improving the Ability to Improve","offset":38553,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/17eec8a15debc212b5530f117dd8ba8bad339590729cbbc4f75a47755ebcb544/part-3.html#text"},{"title":"5.3 Compounding Self-Improvements","offset":47888,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/17eec8a15debc212b5530f117dd8ba8bad339590729cbbc4f75a47755ebcb544/part-4.html#text"},{"title":"6 Safety Discussion","offset":50448,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/17eec8a15debc212b5530f117dd8ba8bad339590729cbbc4f75a47755ebcb544/part-4.html#text"},{"title":"7 Limitations and Conclusion","offset":52829,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/17eec8a15debc212b5530f117dd8ba8bad339590729cbbc4f75a47755ebcb544/part-4.html#text"},{"title":"Acknowledgments","offset":54858,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/17eec8a15debc212b5530f117dd8ba8bad339590729cbbc4f75a47755ebcb544/part-4.html#text"},{"title":"Author Contributions","offset":55263,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/17eec8a15debc212b5530f117dd8ba8bad339590729cbbc4f75a47755ebcb544/part-4.html#text"},{"title":"Appendix A Algorithmic details","offset":55657,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/17eec8a15debc212b5530f117dd8ba8bad339590729cbbc4f75a47755ebcb544/part-4.html#text"},{"title":"A.1 Initial Agent","offset":56192,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/17eec8a15debc212b5530f117dd8ba8bad339590729cbbc4f75a47755ebcb544/part-4.html#text"},{"title":"A.2 Parent Selection","offset":56457,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/17eec8a15debc212b5530f117dd8ba8bad339590729cbbc4f75a47755ebcb544/part-4.html#text"},{"title":"A.3 Pseudocode","offset":59551,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/17eec8a15debc212b5530f117dd8ba8bad339590729cbbc4f75a47755ebcb544/part-4.html#text"},{"title":"A.4 Multi-domain Optimization","offset":59634,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/17eec8a15debc212b5530f117dd8ba8bad339590729cbbc4f75a47755ebcb544/part-4.html#text"},{"title":"Appendix B Baseline Details","offset":60443,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/17eec8a15debc212b5530f117dd8ba8bad339590729cbbc4f75a47755ebcb544/part-4.html#text"},{"title":"Appendix C Domain Details","offset":61431,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/17eec8a15debc212b5530f117dd8ba8bad339590729cbbc4f75a47755ebcb544/part-4.html#text"},{"title":"C.1 Polyglot","offset":61842,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/17eec8a15debc212b5530f117dd8ba8bad339590729cbbc4f75a47755ebcb544/part-4.html#text"},{"title":"C.2 Paper Review","offset":64543,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/17eec8a15debc212b5530f117dd8ba8bad339590729cbbc4f75a47755ebcb544/part-5.html#text"},{"title":"C.3 Robotics Reward Design","offset":65572,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/17eec8a15debc212b5530f117dd8ba8bad339590729cbbc4f75a47755ebcb544/part-5.html#text"},{"title":"C.4 Olympiad-level Math Grading","offset":67844,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/17eec8a15debc212b5530f117dd8ba8bad339590729cbbc4f75a47755ebcb544/part-5.html#text"},{"title":"Appendix D Experiment Details","offset":68804,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/17eec8a15debc212b5530f117dd8ba8bad339590729cbbc4f75a47755ebcb544/part-5.html#text"},{"title":"D.1 Hyperparameters for FMs","offset":69431,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/17eec8a15debc212b5530f117dd8ba8bad339590729cbbc4f75a47755ebcb544/part-5.html#text"},{"title":"D.2 Cost Estimate","offset":70296,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/17eec8a15debc212b5530f117dd8ba8bad339590729cbbc4f75a47755ebcb544/part-5.html#text"},{"title":"D.3 Improvement@k Metric","offset":71046,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/17eec8a15debc212b5530f117dd8ba8bad339590729cbbc4f75a47755ebcb544/part-5.html#text"},{"title":"D.4 Transfer Agent Selection","offset":73316,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/17eec8a15debc212b5530f117dd8ba8bad339590729cbbc4f75a47755ebcb544/part-5.html#text"},{"title":"Appendix E Additional Results","offset":75064,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/17eec8a15debc212b5530f117dd8ba8bad339590729cbbc4f75a47755ebcb544/part-5.html#text"},{"title":"E.1 Best Discovered Task Agents","offset":75869,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/17eec8a15debc212b5530f117dd8ba8bad339590729cbbc4f75a47755ebcb544/part-5.html#text"},{"title":"E.1.1 Paper Review","offset":76066,"level":4,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/17eec8a15debc212b5530f117dd8ba8bad339590729cbbc4f75a47755ebcb544/part-5.html#text"},{"title":"E.1.2 Robotics Reward Design","offset":76213,"level":4,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/17eec8a15debc212b5530f117dd8ba8bad339590729cbbc4f75a47755ebcb544/part-5.html#text"},{"title":"E.1.3 Olympiad-level Math Grading","offset":76377,"level":4,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/17eec8a15debc212b5530f117dd8ba8bad339590729cbbc4f75a47755ebcb544/part-5.html#text"},{"title":"E.2 Qualitative: Improving Task Performance","offset":76648,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/17eec8a15debc212b5530f117dd8ba8bad339590729cbbc4f75a47755ebcb544/part-5.html#text"},{"title":"E.3 Qualitative: Improving the Ability to Improve","offset":81032,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/17eec8a15debc212b5530f117dd8ba8bad339590729cbbc4f75a47755ebcb544/part-6.html#text"},{"title":"E.3.1 Better Evaluation Analysis","offset":82110,"level":4,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/17eec8a15debc212b5530f117dd8ba8bad339590729cbbc4f75a47755ebcb544/part-6.html#text"},{"title":"E.3.2 Compute-Aware Strategic Planning","offset":82882,"level":4,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/17eec8a15debc212b5530f117dd8ba8bad339590729cbbc4f75a47755ebcb544/part-6.html#text"},{"title":"E.3.3 Automated Bias Detection and Correction","offset":84494,"level":4,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/17eec8a15debc212b5530f117dd8ba8bad339590729cbbc4f75a47755ebcb544/part-6.html#text"},{"title":"E.3.4 Performance Tracking Infrastructure","offset":85546,"level":4,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/17eec8a15debc212b5530f117dd8ba8bad339590729cbbc4f75a47755ebcb544/part-6.html#text"},{"title":"E.3.5 Prompt Template Systems","offset":86065,"level":4,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/17eec8a15debc212b5530f117dd8ba8bad339590729cbbc4f75a47755ebcb544/part-6.html#text"},{"title":"E.3.6 Refinement Based on Multi-Generation Performance","offset":86660,"level":4,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/17eec8a15debc212b5530f117dd8ba8bad339590729cbbc4f75a47755ebcb544/part-6.html#text"},{"title":"E.3.7 Persistent Memory","offset":87847,"level":4,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/17eec8a15debc212b5530f117dd8ba8bad339590729cbbc4f75a47755ebcb544/part-6.html#text"},{"title":"E.4 Olympiad-level Math Graders","offset":89714,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/17eec8a15debc212b5530f117dd8ba8bad339590729cbbc4f75a47755ebcb544/part-6.html#text"},{"title":"E.5 Modifying Parent Selection","offset":91088,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/17eec8a15debc212b5530f117dd8ba8bad339590729cbbc4f75a47755ebcb544/part-6.html#text"},{"title":"Appendix F Additional Safety Discussion","offset":94651,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/17eec8a15debc212b5530f117dd8ba8bad339590729cbbc4f75a47755ebcb544/part-7.html#text"},{"title":"Instructions for reporting errors","offset":97160,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/17eec8a15debc212b5530f117dd8ba8bad339590729cbbc4f75a47755ebcb544/part-7.html#text"}],"text_code_points":98124,"source_markdown":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/17eec8a15debc212b5530f117dd8ba8bad339590729cbbc4f75a47755ebcb544/source.md"}