Skip to content

Dataset · ZefanCai

Typed decisions: 12 Open-Jev datasets

ZefanCai/Open-Jev

Task collections where a state and question require a typed decision: yes/no, class, number.

What's inside

Open-Jev publishes twelve separate frozen data configs for the project. Each example turns a state and question into a typed decision: yes/no probability, option distribution, independent label probabilities, or a discrete numerical/ordinal decision. Tasks are mostly synthetic and controlled; manifests, raw logs, code, and reconstruction instructions are provided. Languages include English, Chinese, and Turkish.

What you can do with it

  • Training and evaluating typed decision classifiers
  • Testing LLMs on probabilistic and ordinal answers
  • Yes/no calibration benchmark
Tasks
Text classification
Languages
English, Chinese, Turkish
Size
100K–1M rows
Downloads this month
3.3K
Likes
116

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("ZefanCai/Open-Jev")