Kangaroo: benchmark results
Provider: Other.
Aggregate Rating 1486 ± 38, rank #2586 of 10916 rated models, from 13 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| CG-Bench - Open-Ended | 25.9 | Open-Ended Accuracy (%) | 66.7 |
| CG-Bench | 31.2 | Long-Video MCQ Accuracy (%) | 61.9 |
| CG-Bench - Clue MCQ | 46.4 | Clue-Interval MCQ Accuracy (%) | 61.9 |
| LongVideoBench | 54.8 | Test Total Accuracy (%) | 58.8 |
| CG-Bench - Accuracy@IoU | 1.54 | Accuracy@IoU (%) | 57.1 |
| CG-Bench - Clue Grounding mIoU | 2.32 | Temporal Grounding mIoU (%) | 52.4 |
| Video-MME - Medium Videos (w/o subs) | 55.3 | Accuracy (%) | 34 |
| Video-MME - Short Videos (w/o subs) | 66.1 | Accuracy (%) | 30 |
| Video-MME - Short Videos (w/ subs) | 68 | Accuracy (%) | 28 |
| Video-MME - Long Videos (w/o subs) | 46.6 | Accuracy (%) | 24 |
| Video-MME - Medium Videos (w/ subs) | 55.4 | Accuracy (%) | 24 |
| VideoMME w sub. | 57.6 | Score (imported) | 23.6 |
Interactive version: theaggregate.ai/model?slug=kangaroo · How It Works · Data refreshed daily, snapshot 2026-11-08.