Why am I excited about IMO results we just published: - we did very little IMO-specific work, we just keep training general models - all natural language proofs - no evaluation harness We needed a new research breakthrough and @alexwei_ and team delivered
174,95K