If there is consensus on how to attribute credit for LLM-solved human-steered (with no mathematical human input) proofs, I am happy to follow that consensus. Do you have a concrete alternative recommendation? What should I change?