Open Models Almanac
Methodsdata curationsynthetic data

single mention · data curation

Two-stage GitHub-derived software task generation and RL refinement

Uses a two-stage process to turn candidate software tasks into runnable environments and refine them.

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 two-stage approach.

Filed under data curation › synthetic data