Research map / Large language models / Retrieval and knowledge
Tools are under-documented: Simple Document Expansion Boosts Tool Retrieval
Xuan Lu, Haohang Huang, Rui Meng, Yaohui Jin, Wenjun Zeng, Xiaoyu Shen
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| train retriever on expanded tool documents instead of original (ablation) | Qwen3-Embedding-4B · Tool-DE | NDCG@10: +3.67 -> +6.69 over Qwen3-Embedding-0.6B baseline (4B tool-embed) | Improved |
| train reranker on expanded documents instead of original (ablation) | Qwen3-Reranker-4B · Tool-DE / ToolRet | NDCG@10: +2.00 -> +4.21 | Improved |
| evaluate BM25s on expanded documents | BM25s · Tool-DE vs ToolRet | NDCG@10: 36.41 -> 39.35 | Improved |
| evaluate e5-mistral-7b-instruct on expanded documents | e5-mistral-7b-instruct · Tool-DE vs ToolRet | NDCG@10: 40.02 -> 38.85 | Hurt |
| evaluate Qwen3-Embedding-4B/8B on expanded documents | Qwen3-Embedding-4B/8B · Tool-DE vs ToolRet | Recall@10: R@10 -1.23 and -0.69, N@10 slightly up | Mixed |
| remove example_usage field from expanded documents (ablation) | GritLM / BM25s · Tool-DE | NDCG@10 | Improved |
| remove function_description or tags from expanded documents (ablation) | GritLM / BM25s · Tool-DE | NDCG@10 | Hurt |
| add example_usage field alone to original document (ablation) | GritLM · Tool-DE | NDCG@10 | Hurt |
| add reranker on original (non-expanded) ToolRet documents | jina-reranker-m0 · ToolRet | NDCG@10: -0.08 | No clear effect |
| add reranker on expanded documents | Qwen3-Reranker-4B · Tool-DE vs ToolRet | NDCG@10: +0.50 -> +1.20 | Improved |
| document expansion at evaluation time lowers similarity scores asymmetrically | GritLM · Tool-DE | mean similarity: negatives Web -0.0055 vs positives -0.0014 | Improved |
| apply document expansion to tool documents | LLM-audited documents · 300 sampled docs | incomplete-doc rate: 41.6% -> 23.5% | Improved |
Closest papers by meaning
- From Exploration to Mastery: Enabling LLMs to Master Tools via Self-Driven Interactions (ICLR 2025 · similarity 0.87)
- Can Language Models Actually Retrieve In-Context? Drowning in Documents at Million Token Scale (NeurIPS 2026 · similarity 0.86)
- A Systematic Exploration of Knowledge Graph Alignment with Large Language Models in Retrieval Augmented Generation (AAAI 2025 · similarity 0.86)
- MetaTool Benchmark for Large Language Models: Deciding Whether to Use Tools and Which to Use (ICLR 2024 · similarity 0.85)
- Docopilot: Improving Multimodal Models for Document-Level Understanding (CVPR 2025 · similarity 0.85)
- Beyond Content Relevance: Evaluating Instruction Following in Retrieval Models (ICLR 2025 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
