Open Models Almanac
Methodsdata curationsynthetic data

single mention · data curation

Iterative reward-guided training for task construction

Trains a task-construction model iteratively using difficulty and correctness as reward signals.

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 model to construct better tasks.

Filed under data curation › synthetic data