---
title: "面向 LLM Agent 的最优技能选择：首个可证明双准则保证算法"
source: "arXiv AI/CL/LG（论文）"
source_tier: "t2"
url: "https://arxiv.org/abs/2608.19993v1"
permalink: "https://feed.astralor.com/items/itm_b2c60c46880b08c89276a5cd"
published: "2026-08-20T13:08:17.000Z"
score: 77
channel: "ai"
tags: ["BigCodeBench","BPS","LLM Agent","子模优化","技能选择","近似算法"]
---

# 面向 LLM Agent 的最优技能选择：首个可证明双准则保证算法

## AI 点评

> 这篇论文把 Agent 技能路由从‘拼感觉’变成了‘有边界’的优化问题，对做 Agent 基础设施和推理成本优化的人来说是必读。真正值得关注的不只是那个近似比，而是它证明了独立相关性排序在集合层面可能天然低效。

## 速览

本论文首次将 LLM Agent 的技能选择建模为带硬 token 预算的优化问题，目标是在最大化单调子模收益与上下文惩罚之间取得平衡。作者提出多项式时间算法 BPS，并证明其具有 (1-1/e,1) 双准则近似保证，且收益系数在多项式时间内最优。在受污染控制的 BigCodeBench 变体上，BPS 以比最强发布路由器少 28% 的 token 达到 0.73 的任务成功率，远超多种基线。

## 关键点

- 将技能选择形式化为带 token 预算的单调子模收益扣除上下文惩罚的优化问题。
- 提出多项式时间算法 BPS，并证明首个 (1-1/e,1) 双准则近似保证。
- 在 BigCodeBench 变体上，BPS 以少 28% 的 token 达到 0.73 任务成功率，显著优于现有路由器。

## 全文

Loading reusable skill documents into a bounded context window is now the primary way large language model (LLM) agents acquire task-specific capabilities, which makes skill selection a first-order determinant of task performance and token cost. Yet current agents score skills independently by semantic relevance and assemble the set by top-$k$ or greedy packing, with no quality guarantee or cost awareness on the selected set. As a result, redundant or poorly chosen skills waste scarce context tokens and can even degrade performance. We give the first model of how the selected skill set shapes execution outcomes and cast skill selection as an optimization problem: choose a skill set under a hard token budget to maximize a monotone submodular benefit minus context penalty. For this problem, we develop Best Prefix Selection (BPS), a polynomial-time algorithm, and prove, to our knowledge, the first performance guarantee for skill selection: a bicriteria $(1-1/e,1)$ approximation whose benefit coefficient is optimal in polynomial time. On a contamination-controlled BigCodeBench variant, BPS outperforms all the baselines, reaching $0.73$ measured task success versus $0.20$--$0.52$ for released skill routers, text retrievers, and the executor's own selection, on $28\%$ fewer tokens than the strongest released router.

---

引用前请核对原文;本文由星际流动整理:https://feed.astralor.com/items/itm_b2c60c46880b08c89276a5cd