Lean Finder: Semantic Search for Mathlib That Understands User Intents
Jialin Lu, Kye Emond, Kaiyu Yang, Swarat Chaudhuri, Weiran Sun, Wuyang Chen
cs.LG
Oct 8, 2025 · v1
TL;DR
Builds Lean Finder, a semantic search engine over Lean and Mathlib fine-tuned to match mathematicians' query intents.
Abstract
We present Lean Finder, a semantic search engine for Lean and mathlib that understands and aligns with the intents of mathematicians. Progress in formal theorem proving is often hindered by the difficulty of locating relevant theorems and the steep learning curve of the Lean 4 language, making advancement slow and labor-intensive. Existing Lean search engines, though helpful, rely primarily on informalizations (natural language translation of the formal statements), while largely overlooking the mismatch with real-world user queries. In contrast, we propose a user-centered semantic search tailored to the needs of mathematicians. Our approach begins by analyzing and clustering the semantics of public Lean discussions, then fine-tuning text embeddings on synthesized queries that emulate user intents. We further align Lean Finder with mathematicians' preferences using diverse feedback signals, encoding it with a rich awareness of their goals from multiple perspectives. Evaluations on real-world queries, informalized statements, and proof states demonstrate that our Lean Finder achieves over $30\%$ relative improvement compared to previous search engines and GPT-4o. In addition, Lean Finder is compatible with LLM-based theorem provers, bridging retrieval with formal reasoning. Lean Finder is available at:
https://leanfinder.github.io
Problem
Existing search engines for Lean's Mathlib rely primarily on informalizations of formal statements, which do not align well with real-world user queries by mathematicians.
Approach
Lean Finder is a semantic search engine for Lean and Mathlib that aligns with mathematicians' intents. The approach analyzes and clusters the semantics of public Lean discussions, fine-tunes text embeddings on synthesized queries emulating user intents, and further aligns with mathematicians' preferences using diverse feedback signals. It is compatible with LLM-based theorem provers.
Results
Lean Finder achieves over 30% relative improvement compared to previous search engines and GPT-4o. In evaluation with real user queries, users preferred Lean Finder in 81.6% of cases versus 56.9% for Lean Search and 54.1% for GPT-4o.