ImProver 2: Iteratively Self-Improving LMs for Neurosymbolic Proof Optimization

In three linesImProver 2 is a neurosymbolic framework for automated proof optimization in Lean 4. A 7B-parameter model trained outperforms orders-of-magnitude larger models and is competitive with mid-tier frontier models. The scaffold exposes formal structure alongside lightweight informal abstractions.

Summary generated by Claude — human-verified