EgoNormia
Can large models make normative decisions in physical-social embodied situations?
Verified Split: A high-quality subset of 200 videos with full agreement on the correct answers among 5 independent annotators.
Input Modality Types:
- Video: Models receive both video input (1 fps, concatenated into a single image) and questions
Model | Modality | Both | Act | Jus | Sen | Date |
|---|---|---|---|---|---|---|
  | Video | 67.8 | 74.4 | 68.9 | 56.7 | 2025-05-20 |
| Video | 58.4 | 69.7 | 59.6 | 58.9 | 2025-05-20 |
  | Video | 58.3 | 66.7 | 66.7 | 64.6 | 2025-05-20 |
| Video | 50.0 | 70.6 | 50.0 | 56.1 | 2025-05-20 |
| Video | 49.0 | 56.5 | 50.5 | 61.8 | 2025-05-20 |
| Video | 48.0 | 53.0 | 50.5 | 56.8 | 2025-05-20 |
  | Video | 47.0 | 57.5 | 48.0 | 68.2 | 2025-05-20 |
| Video | 46.4 | 50.0 | 50.0 | 57.7 | 2025-05-20 |
| Video | 45.5 | 53.0 | 50.0 | 62.7 | 2025-05-20 |
| Video | 37.5 | 37.5 | 37.5 | 39.6 | 2025-05-20 |
  | Video | 33.3 | 40.0 | 41.7 | 40.8 | 2025-05-20 |
| Video | 22.7 | 27.3 | 27.3 | 47.7 | 2025-05-20 |
S InternVL 2.5 | Video | 13.0 | 16.5 | 15.0 | 52.1 | 2025-05-20 |
  | Video | 4.0 | 18.0 | 10.5 | 55.6 | 2025-05-20 |