---
title: "生成式技能组合：为LLM智能体实现结构化技能规划"
source: "arXiv AI/CL/LG（论文）"
source_tier: "t2"
url: "https://arxiv.org/abs/2606.32025v1"
permalink: "https://feed.astralor.com/items/itm_f30b61d5534941f8a3cac8b3"
published: "2026-06-30T17:53:09.000Z"
score: 59
channel: "ai"
tags: ["LLM智能体","SkillComposer","技能组合","结构化规划","自回归解码"]
---

# 生成式技能组合：为LLM智能体实现结构化技能规划

## 速览

大型语言模型智能体通过技能库解决复杂任务，但如何从大量技能中自动选择并编排合适的技能序列是当前瓶颈。现有方法或暴露全部技能给推理过程，或基于检索排序，但都忽略了技能组合的结构性——即同时决定技能子集、数量和顺序。本文形式化定义了结构化技能组合问题，并提出SkillComposer，通过任务条件化的自回归解码器，将子集、数量和顺序联合预测为一个技能计划序列。在真实人工策展的技能库上构建训练集，实验表明SkillComposer在组合质量和下游任务成功率上显著优于检索基线，匹配黄金技能检索上界，且提示词开销更低。

## 关键点

- 形式化定义了结构化技能组合问题：给定任务和技能库，同时预测技能子集、数量和执行顺序
- 提出SkillComposer，使用任务条件化的自回归解码器联合生成技能计划，自然捕捉技能间依赖
- 在真实人工策展技能库上构建训练集，评估显示组合质量和下游成功率显著提升
- 在GPT-5.2-Codex和Gemini-3-Pro-Preview上，SkillComposer使通过率分别提升23.1%和18.2%，接近上限，并降低提示词开销

## 全文

Recent LLM agents benefit from skills for solving complex tasks. Skills encapsulate modular packages of procedural knowledge and instructions for performing specialized tasks, such as setting up a sandboxed environment, running a test suite, or refactoring a function across multiple files. As skill libraries grow and become reusable across tasks and domains, selecting an appropriate skill composition has emerged as a central bottleneck. Existing approaches fall into two categories. One exposes the agent's reasoning to the entire skill collection; the other performs skill retrieval via embeddings or LLM-based rerankers. Both provide useful insights; however, they miss the structural nature of skill composition, which is a joint decision over which skills, how many, and in what order -- three dimensions that cannot be decoupled. We formalize this as structured skill composition: given a task and a skill library, predict an executable skill plan that jointly specifies the activated subset, count, and execution order. We propose SkillComposer, which instantiates structured skill composition as task-conditioned skill sequence prediction. SkillComposer uses a constrained autoregressive decoder over skill identifiers, so subset, count, and order emerge jointly from a single decoding pass, and dependencies between successive skills are captured naturally. We build a training set of task-composition pairs from a real, human-curated skill library. We then evaluate SkillComposer along two axes: composition quality on a held-out test set, and downstream task success on SkillsBench across two production-grade coding agents. On GPT-5.2-Codex, Gemini-3-Pro-Preview, SkillComposer raises the pass rate by +23.1, +18.2pp over the no-skill baseline, surpassing top-3 retrieval and matching the gold-skill retrieval upper bound at lower prompt-token cost.

---

引用前请核对原文;本文由星际流动整理:https://feed.astralor.com/items/itm_f30b61d5534941f8a3cac8b3