Benchmark
PIQA
reasoning
physics
general
text
PIQA (Physical Interaction: Question Answering) is a benchmark dataset for physical commonsense reasoning in natural language. It tests AI systems' ability to answer questions requiring physical world knowledge through multiple choice questions with everyday situations, focusing on atypical solutions inspired by instructables.com. The dataset contains 21,000 multiple choice questions where models must choose the most appropriate solution for physical interactions.
语言EN
满分1
参评模型9
模型排名
| 名次 | 模型 | 机构 | 分数 | 来源 |
|---|---|---|---|---|
| 1 | Phi-3.5-MoE-instruct | Microsoft | 88.6 | 来源 ↗ |
| 2 | Gemma 2 27B | 83.2 | 来源 ↗ | |
| 3 | Gemma 2 9B | 81.7 | 来源 ↗ | |
| 4 | Gemma 3n E4B | 81.0 | 来源 ↗ | |
| 5 | Gemma 3n E4B Instructed LiteRT Preview | 81.0 | 来源 ↗ | |
| 6 | Phi-3.5-mini-instruct | Microsoft | 81.0 | 来源 ↗ |
| 7 | Gemma 3n E2B | 78.9 | 来源 ↗ | |
| 8 | Gemma 3n E2B Instructed LiteRT (Preview) | 78.9 | 来源 ↗ | |
| 9 | Phi 4 Mini | Microsoft | 77.6 | 来源 ↗ |