Open Models Almanac
Methodsdata curationdata filtering

single mention · data curation

Normalize training samples to their problem statements

Samples are reduced to problem statements and recurring dataset templates are stripped to avoid spurious benchmark matches.

The sources on this page only mention it in passing. The line above is the extraction's one-line gloss. See data filtering for the methods filed beside it.

Also called Normalization.

Filed under data curation › data filtering