Neural Theorem Proving

LLMlean 2.0: Enhanced Lean 4 Integration with LLMs

Major update to LLMlean with improved tactic suggestions, better context understanding, and support for more LLM providers.

Release

LLMlean 2.0 brings significant improvements to the Lean 4 + LLM integration.

New Features

  • Better context: Improved proof state representation
  • Multi-provider: OpenAI, Anthropic, local models
  • Custom tactics: User-defined tactic templates
  • Performance: 2x faster inference

Improvements

  • More accurate tactic predictions
  • Better handling of complex proof states
  • Integration with mathlib4 latest
  • VS Code extension updates

Usage

import LLMlean

-- Get AI suggestions
llmstep ""

-- Custom prompt
llmstep "induction on n"

Community

Growing adoption in Lean community for proof assistance.