VideoChat2: benchmark results
Provider: Other.
Aggregate Rating 1484 ± 31, rank #2601 of 10916 rated models, from 49 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| SceneBench - Scene QA | 31.5 | Accuracy (%) on SceneBench's 3,311 scene qa items (questions | 88.9 |
| SceneBench - Scene QA (Audio) | 36.1 | Accuracy (%) on SceneBench's 4,097 scene qa (audio) items (s | 88.9 |
| MuseBench - Game Arts (Single-Select) | 10.45 | Game-arts chance-adjusted accuracy (%): raw single-select ac | 63 |
| SceneBench - Title Prediction | 98.3 | Accuracy (%) on SceneBench's 345 title prediction items (cho | 55.6 |
| E3mo-Bench (Evoked) - Open-Vocabulary Recognition (Emotion Wheel) | 57.17 | Emotion-wheel set F-score (%; the OV-MER semantic emotion wh | 53.3 |
| VidPair-Halluc - Open-Ended Description | 42.25 | Open-ended description score (%) judged by GPT-4o on VidPair | 50 |
| E3mo-Bench (Evoked) - Open-Vocabulary Recognition (VAD Similarity) | 44.66 | Redundancy-aware VAD set similarity F-VAD (%; predicted and | 46.7 |
| SceneBench - Comment Prediction | 72.5 | Accuracy (%) on SceneBench's 297 comment prediction items (c | 44.4 |
| SceneBench | 52.8 | Average accuracy (%) over SceneBench's six tasks (title and | 38.9 |
| E3mo-Bench (Evoked) - Video-Pair Perception (Yes-or-No) | 50.42 | Accuracy (%; yes-or-no questions comparing the emotion, vale | 33.3 |
| V-STaR | 18.64 | Average All Score | 30.8 |
| V-STaR - All mAM | 17.02 | Score | 30.8 |
Interactive version: theaggregate.ai/model?slug=videochat2 · How It Works · Data refreshed daily, snapshot 2026-11-08.