if you're verifiably RL'ing a model where you don't intend to hide the chain of thought, what's even the point of <think> tags? why not let the model structure the output totally to its own…

Posters: voooooogel. voooooogel

4 likes recorded by the source corpus.

Topics: AI & LLMs

View Post: https://x.com/voooooogel/status/1890192332332597757