3 ms·

Didn’t OpenAI themselves publish a papers years ago that scaling parameters has diminishing returns?