A 4B model post-trained with SFT, RL, and a reasoning cache surpasses larger open models and approaches proprietary ones on Olympiad proof generation.
</errors> Input data •Problem Statement:{problem} •Marking Scheme:{marking scheme} •Proof Solution:{solution} ProofBench Grader Prompt You are anexpert math proof grader
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.AI 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
QED-Nano: Teaching a Tiny Model to Prove Hard Theorems
A 4B model post-trained with SFT, RL, and a reasoning cache surpasses larger open models and approaches proprietary ones on Olympiad proof generation.