DAP achieves SOTA on Hard Mode ATP by having LLMs discover answers then prove them formally, solving 10 CombiBench and 36 PutnamBench problems while exposing that LLMs exceed 80% answer accuracy where formal provers stay under 10%.
Ensure the syntax is correct for Lean 4 (e.g., `π` becomes `Real.pi`)
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.AI 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Discover and Prove: An Open-source Agentic Framework for Hard Mode Automated Theorem Proving in Lean 4
DAP achieves SOTA on Hard Mode ATP by having LLMs discover answers then prove them formally, solving 10 CombiBench and 36 PutnamBench problems while exposing that LLMs exceed 80% answer accuracy where formal provers stay under 10%.