SkillSight: Seeing Through Shared Descriptions for Accurate Skill Retrieval
作者: Jinying Xiao, Bin Ji, Shasha Li, Xiaodong Liu, Ma Jun, Jiacheng Jie, Chao Wang, Nyima Tashi, Jie Yu
分类: cs.AI
发布日期: 2026-07-21
备注: 9 pages, 4 figures
🔗 代码/项目: GITHUB
💡 一句话要点
提出SkillSight以解决技能检索中的共享描述偏差问题
🎯 匹配领域: 支柱九：具身大模型 (Embodied Foundation Models)
关键词: 技能检索 共享描述 语义校准 词汇证据 无训练框架 大型语言模型 自动化系统
📋 核心要点
- 现有技能检索方法未能有效处理技能描述中的共享背景，导致检索准确性下降。
- SkillSight通过语义背景校准和词汇证据校准，减少共享描述模式带来的相似性，从而提高检索效果。
- 在SRA-Bench和SkillBench-Supp上的实验显示，SkillSight在检索指标上均有显著提升，尤其在Recall@10上提高了20.21个百分点。
📝 摘要（中文）
随着大型语言模型代理访问越来越大的技能库，准确检索所需技能变得至关重要。现有的检索方法往往将技能描述视为普通文档，忽视了其高度规律的结构。本文提出SkillSight，一个无训练的检索框架，通过语义和词汇空间校准共享背景，从而提高技能检索的准确性。实验结果表明，SkillSight在多个检索指标上均有显著提升，Recall@10提高了20.21个百分点，并且在端到端评估中表现最佳，速度比Dense + Reranker基线快多达1248倍。
🔬 方法详解
问题定义：本文旨在解决技能检索中由于共享描述背景导致的偏差问题。现有方法将技能描述视为普通文档，未能有效区分所需能力，造成检索效果不佳。
核心思路：SkillSight的核心思路是通过校准共享背景来提高检索的准确性，具体包括语义背景校准和词汇证据校准，以减少共享描述模式的影响。
技术框架：SkillSight的整体架构包括两个主要模块：语义背景校准和词汇证据校准。前者通过识别通用标记来估计背景子空间，后者则通过降低共享背景标记的权重来恢复区分性证据。
关键创新：SkillSight的主要创新在于其无训练的检索框架，明确校准共享背景，从而有效提升技能选择的准确性和效率。这一方法与现有依赖训练的检索方法本质上不同。
关键设计：在设计中，使用了逆文档频率（IDF）来识别通用标记，并通过调整相似性计算来减少共享描述模式的影响。
🖼️ 关键图片
📊 实验亮点
SkillSight在SRA-Bench和SkillBench-Supp上的实验结果显示，Recall@10提升了20.21个百分点，且在端到端评估中超越了三种代理模型，表现最佳。此外，SkillSight的速度比Dense + Reranker基线快多达1248倍，展现了其在效率上的显著优势。
🎯 应用场景
SkillSight的研究成果在多个领域具有潜在应用价值，尤其是在智能助手、自动化系统和人机交互等场景中。通过提高技能检索的准确性，能够显著提升系统的响应效率和用户体验，未来可能推动更智能的自动化决策系统的发展。
📄 摘要（原文）
As large language model agents gain access to increasingly large skill libraries, retrieving the right skill becomes critical to reliable capability selection and execution. Existing retrievers often treat skill descriptions as ordinary documents, overlooking their highly regular structure: shared descriptive patterns recur across many skills while providing little evidence for distinguishing the required capability. We show that this shared descriptive background systematically contributes to dense relevance scores, induces a pronounced energy gap between queries and skill documents, and obscures task-relevant signals. Based on this observation, we propose SkillSight, a training-free retrieval framework that calibrates shared background in both semantic and lexical spaces. Semantic Background Calibration estimates a background subspace from generic tokens identified by IDF, reducing similarity induced by shared descriptive patterns, while Lexical Evidence Calibration downweights shared background tokens to recover discriminative token-level evidence. Experiments on SRA-Bench and SkillBench-Supp demonstrate consistent improvements across retrieval metrics, with SkillSight improving Recall@10 by up to 20.21 percentage points over the original dense retriever. In end-to-end evaluation, SkillSight achieves the best overall performance across three agent models and outperforms LLM Selection by up to 4.97 percentage points. It is also up to 1,248 times faster than the Dense + Reranker baseline. These results identify shared descriptive background as a key source of bias in skill retrieval and demonstrate that explicitly calibrating it enables accurate and efficient skill selection without additional training. Our code is available at https://github.com/xiaojinying/SkillSight.