LeanReasoner: Boosting Complex Logical Reasoning with Lean
This addresses logical inconsistencies in AI reasoning for tasks like theorem proving, though it is incremental as it builds on existing Lean methods.
The paper tackles the problem of large language models struggling with complex logical reasoning by formalizing reasoning problems into theorems using the Lean theorem proving framework, achieving state-of-the-art performance on the FOLIO dataset and near-state-of-the-art on ProofWriter with fine-tuning on fewer than 100 samples per dataset.
Large language models (LLMs) often struggle with complex logical reasoning due to logical inconsistencies and the inherent difficulty of such reasoning. We use Lean, a theorem proving framework, to address these challenges. By formalizing logical reasoning problems into theorems within Lean, we can solve them by proving or disproving the corresponding theorems. This method reduces the risk of logical inconsistencies with the help of Lean's symbolic solver. It also enhances our ability to treat complex reasoning tasks by using Lean's extensive library of theorem proofs. Our method achieves state-of-the-art performance on the FOLIO dataset and achieves performance near this level on ProofWriter. Notably, these results were accomplished by fine-tuning on fewer than 100 in-domain samples for each dataset.