Discover and Prove: An Open-source Agentic Framework for Hard Mode Automated Theorem Proving in Lean 4
About
Most ATP benchmarks embed the final answer within the formal statement -- a convention we call "Easy Mode" -- a design that simplifies the task relative to what human competitors face and may lead to optimistic estimates of model capability. We call the stricter, more realistic setting "Hard Mode": the system must independently discover the answer before constructing a formal proof. To enable Hard Mode research, we make two contributions. First, we release MiniF2F-Hard and FIMO-Hard, expert-reannotated Hard Mode variants of two widely-used ATP benchmarks. Second, we introduce Discover And Prove (DAP), an agentic framework that uses LLM natural-language reasoning with explicit self-reflection to discover answers, then rewrites Hard Mode statements into Easy Mode ones for existing ATP provers. DAP sets the state of the art: on CombiBench it raises solved problems from 7 (previous SOTA, Pass@16) to 10; on PutnamBench it is the first system to formally prove 36 theorems in Hard Mode -- while simultaneously revealing that state-of-the-art LLMs exceed 80% answer accuracy on the same problems where formal provers manage under 10%, exposing a substantial gap that Hard Mode benchmarks are uniquely suited to measure.
Related benchmarks
| Task | Dataset | Result | Rank | |
|---|---|---|---|---|
| Automated Theorem Proving | miniF2F Easy Mode (test) | -- | 5 | |
| Automated Theorem Proving | FIMO Easy Mode | -- | 5 | |
| Automated Theorem Proving | CombiBench Easy Mode | -- | 4 | |
| Automated Theorem Proving | CombiBench Hard Mode | Total Solved (Pass@32)10 | 3 | |
| Automated Theorem Proving | PutnamBench Easy Mode | -- | 3 | |
| Automated Theorem Proving | PutnamBench Hard Mode | Total Solved (Pass@32)36 | 2 | |
| Automated Theorem Proving | miniF2F Hard Mode (test) | Total Solved (Pass@32)204 | 2 | |
| Automated Theorem Proving | FIMO Hard Mode | Total Solved (Pass@32)3 | 2 |