You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
As far as I know, we have had discussions about this, and in cases, where the training data can't be disclosed, we take the model maker's word on whether they have trained on any of the MTEB task datasets.
Voyage hasn't told us specifically what they trained on, but they have told us that they did not train on these datasets, while with voyage-3-m-exp they told us that they did.
I think this is fair play as long as the companies stay transparent and accountable.
If the training data is unknown, there should probably be no checkmark for zero-shot?
The text was updated successfully, but these errors were encountered: