Metraipolis › archive › dc_078db009c6bd
Scaling Discovery Requires Forgetting — arXiv:2610.20563
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2610.20563
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-15T20:02:30.353Z
- SHA-256 of the archived text
dfc4d856d402306f7ea121e2892bfe97dcd66879033d941960d1628d6822a45c- Stable URL of this page
https://metraipolis.com/documento/dc_078db009c6bd
How to cite
Scaling Discovery Requires Forgetting — arXiv:2610.20563 — arXiv (en), https://arxiv.org/abs/2610.20563 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_078db009c6bd (Metraipolis archive dc_078db009c6bd, sha256 dfc4d856d402306f7ea121e2892bfe97dcd66879033d941960d1628d6822a45c)
BibTeX
@misc{dc078db009c6bd,
title = {Scaling Discovery Requires Forgetting — arXiv:2610.20563},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2610.20563},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_078db009c6bd; sha256 dfc4d856d402306f7ea121e2892bfe97dcd66879033d941960d1628d6822a45c},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.20563. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Automated research with LLM agents has shown promising results in various domains. Through iterative refinement and scored verifiers, LLM agents are capable of discovering new ideas. There are different ways to scale automated research: let one agent iterate for longer, run several agents in parallel, or further let the parallel agents communicate and build on one another's candidates. However, we find that all three strategies eventually plateau as the agents accumulate history. Each agent becomes anchored to its own past attempts, making fewer and smaller edits. Beyond this point, additional budget brings little further improvement. In this paper, we show that the key to scaling discovery further is to forget. Periodically discarding the search state, including every agent's session and the candidate history, and restarting from the best candidate found so far, unblocks the search and pushes the same budget further. We implement forgetting in a team of agents with communication and evaluate it on four tasks spanning heuristic programming, kernel optimization and model training. The results show up to 15% improvement compared to parallel agents, and up to 39% compared to a single agent. Our results suggest that, for open-ended discovery, what an agent is allowed to remember matters as much as how much compute it is given. Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.AI, nuovi lavori del giorno). Categorie arXiv: Artificial Intelligence (cs.AI) ; Computation and Language (cs.CL). Pagina del lavoro: https://arxiv.org/abs/2610.20563 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito). Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina. Letto il 2026-10-15T19:35:48.308Z dalla pagina https://arxiv.org/list/cs.AI/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).