3 ms·
From the paper: > When producing full natural-language proofs on IMO-AG-30, however, GPT-4 has a success rate of 0% ...
by 1024core 3y ago
From the paper:
> When producing full natural-language proofs on IMO-AG-30, however, GPT-4 has a success rate of 0% ...