s1lv3rj1nx/openjev-mixture
OpenJev training mixture 279 classification and multiple-choice tasks, 323,466 rows, normalised into one typed-decision format so a single model can be trained across all of them. Assembled from tasksource plus two curated ordinal datasets. Built for OpenJev. The format is plain JSONL, so nothing here requires that code. Composition primitive tasks what it is choice 234 pick one of K options noul 34 yes/no score 11 one level on an ordered scale… See the full description on the dataset page: https://huggingface.co/datasets/s1lv3rj1nx/openjev-mixture.
0191
