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
{{ message }}
This repository has been archived by the owner on Oct 5, 2023. It is now read-only.
"I want to see examples that are lower-scoring than other examples of similar length"
The reason why this is important is because some metrics are sensitive to the length of the output, so if you select all low-scoring examples you might get all short examples, which is not super-interesting.
Generalizing this a bit, it could be interesting to think of a way to say "I want examples that are low-scoring, controlling for some other feature".
This is relatively easy when controlling for a categorical feature (you could just take all examples in the bottom nth percentile given any particular categorical feature value), but it is somewhat more complicated conceptually when the feature is a float or integer.
The text was updated successfully, but these errors were encountered:
One common use case that I have noticed is
The reason why this is important is because some metrics are sensitive to the length of the output, so if you select all low-scoring examples you might get all short examples, which is not super-interesting.
Generalizing this a bit, it could be interesting to think of a way to say "I want examples that are low-scoring, controlling for some other feature".
This is relatively easy when controlling for a categorical feature (you could just take all examples in the bottom nth percentile given any particular categorical feature value), but it is somewhat more complicated conceptually when the feature is a float or integer.
The text was updated successfully, but these errors were encountered: