Open Models Almanac
Methodsdata curationsynthetic data

single mention · data curation

Randomize requested instruction length during VLM relabeling

Samples a target word count to vary the length of generated task instructions.

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

Also called randomly sample a number.

Filed under data curation › synthetic data