Viewing a single comment thread. View all comments

MrEloi t1_j9jneec wrote

Hold your horses!

Research has shown that excessively large contexts disrupt the model performance.

That said, I have no idea where the token size boundary between amazing and broken is.

5