HellaSwag
Commonsense ReasoningEnglishBenchmark
HellaSwag is a commonsense reasoning benchmark dataset in English from Zellers et al. with 70 records in JSON format.
📊 This dataset is used as an LLM benchmark. See model leaderboards →
About HellaSwag
Dataset for studying grounded commonsense inference. It consists of 70k multiple choice questions about grounded situations: each question comes from one of two domains -- activitynet or wikihow -- with four answer choices about what might happen next in the scene.
Details
- Task
- Commonsense Reasoning
- Language
- English
- Format
- JSON
- Rows / instances
- 70
- Creator
- Zellers et al.
- Year
- 2019