研究库 论文知识库

内容库 / 论文

arXiv:2609.05820 · 工程化

Online Learning with LLM Experts from Limited Feedback

Online Learning with LLM Experts from Limited Feedback

类型：arxiv
标识：2609.05820
链接：https://arxiv.org/abs/2609.05820
主分类：engineering
形态：method
TLDR：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 O(d T / m), while in the bandit setting we achieve O(d T K / m), where m ll T is a budget on feedback. Our experiments show that we efficiently learn high-quality routing
待LLM分类：是
来源文件：
/inbox/tom/_candidates/2026-09-14-agent-rag-longcontext-candidates.json

相关 · 按语义匹配

笔记精读与批判 · From 128K to 4M: Efficient Training of Ultra-Long Context LLMs

仓库 Dusto-AI/llm-situational-awareness-paper-list

笔记 Cool Papers · cs.CL (papers.cool) · RSS 摘要

笔记 📚 学术研究知识库草稿 · Jay · 2026-06-22

笔记 Context Length Alone Hurts LLM Performance Despite Perfect Retrieval — 精读与批判

仓库 AIDASLab/Awesome-LLM-Values-and-Pluralistic-Alignment

arXiv

2609.05820

TLDR

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 O(d T / m), while in the bandit setting we achieve O(d T K / m), where m ll T is a budget on feedback. Our experiments show that we efficiently learn high-quality routing

arXiv ↗ 中文全文