Metraipolis › archive › dc_294864582e70
Online Learning with LLM Experts from Limited Feedback — arXiv:2609.05820
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2609.05820
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-16T13:56:21.270Z
- SHA-256 of the archived text
dc947e1116ab101ed3302c1264798355e240c5bd04070977aa71d5e944066ecc- Stable URL of this page
https://metraipolis.com/documento/dc_294864582e70
How to cite
Online Learning with LLM Experts from Limited Feedback — arXiv:2609.05820 — arXiv (en), https://arxiv.org/abs/2609.05820 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_294864582e70 (Metraipolis archive dc_294864582e70, sha256 dc947e1116ab101ed3302c1264798355e240c5bd04070977aa71d5e944066ecc)
BibTeX
@misc{dc294864582e70,
title = {Online Learning with LLM Experts from Limited Feedback — arXiv:2609.05820},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2609.05820},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_294864582e70; sha256 dc947e1116ab101ed3302c1264798355e240c5bd04070977aa71d5e944066ecc},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.05820. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
We study adaptive routing of prompts to large language model (LLM) experts to maximize response quality in an online setting with limited feedback. We formulate it as a bandit problem with $K$ actions that represent experts and $d$ features that encode prompts, over a horizon of $T$ rounds. We propose algorithms that strategically select and observe rewards to minimize regret. In the full-information setting, we achieve a regret of $\tilde{O}(d T / \sqrt{m})$, while in the bandit setting we achieve $\tilde{O}(d T \sqrt{K / m})$, where $m \ll T$ is a budget on feedback. Our experiments show that we efficiently learn high-quality routing strategies across diverse LLMs from limited feedback.
Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-09 in cs.LG).
Categorie arXiv: Machine Learning (cs.LG) ; Machine Learning (stat.ML).
Pagina del lavoro: https://arxiv.org/abs/2609.05820 — 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 2026-10-16T13:39:42.811Z dalla pagina https://arxiv.org/catchup/cs.LG/2026-09-09?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).