A no-fine-tuning ChatGPT pipeline with breadth-first and depth-first tactic searches achieves a 31.15% pass rate on miniF2F in Lean, surpassing most but not all published baselines.
In: Conference on Neural Information Processing Systems, vol
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LO 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Simplifying Formal Proof-Generating Models with ChatGPT and Basic Searching Techniques
A no-fine-tuning ChatGPT pipeline with breadth-first and depth-first tactic searches achieves a 31.15% pass rate on miniF2F in Lean, surpassing most but not all published baselines.