AI Paper Atlas

Research map / Large language models / Agents, code and math

RefTool: Reference-Guided Tool Creation for Knowledge-Intensive Reasoning

Xiao Liu, Da Yin, Zirui Wu, Yansong Feng

ICLR 2026 · Large language models › Agents, code and math · approach: Mathematical and structural reasoning benchmarks · 2 citations (OpenAlex)

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 changedWhereMeasuredEffect
add RefTool tools to PoTPoT · QRData-causalaccuracy: 37.1 -> 41.6Improved
add RefTool tools to ReActReAct · QRData-causalaccuracy: 43.8 -> 46.3Improved
add RefTool tools to PoTPoT · TheoremQA-physicsaccuracy: 52.2 -> 54.8Improved
add RefTool tools to PoTPoT · SciBench-chemistryaccuracy: 55.0 -> 57.7Improved
add plain RAG over the same textbooks to PoTPoT · QRData-causalaccuracy: 37.1 -> 36.4No clear effect
add plain RAG over the same textbooks to PoTPoT · TheoremQA-physicsaccuracy: 52.2 -> 51.1Hurt
add plain RAG over the same textbooks to PoTPoT · SciBench-chemistryaccuracy: 55.0 -> 55.6No clear effect
add code-form tools in place of raw text segments (hierarchy kept) (ablation)PoT + RefTool · QRData-causalaccuracy: 39.0 -> 41.6Improved
add code-form tools in place of raw text segments (hierarchy kept) (ablation)PoT + RefTool · TheoremQA-physicsaccuracy: 52.2 -> 54.8Improved
add hierarchical selection in place of similarity-based retrieval (ablation)PoT + RefTool · QRData-causalaccuracy: 38.0 -> 41.6Improved
add hierarchical selection in place of similarity-based retrieval (ablation)PoT + RefTool · TheoremQA-physicsaccuracy: 51.8 -> 54.8Improved
add RefTool to Zhuang-Chinese translation promptingRule-by-Rule Retrieval w. Code · Zhuang-Chinese translation (ZhuangRules)BLEU: 43.9 -> 54.0Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy