Comment by BenzeneDream

You actually think recent chats were used in training of these models to THEN complete the proofs? This seems incredibly far-fetched and almost impossible based on how LLMs train. Occam's Razor says that simply the models are getting increasingly more capable in these domains.