ShareGPT4Video: benchmark results
Provider: LLaVA Team.
Aggregate Rating 1415 ± 43, rank #5432 of 10916 rated models, from 16 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| CG-Bench - Clue Grounding mIoU | 2.07 | Temporal Grounding mIoU (%) | 38.1 |
| VidPair-Halluc - Multi-Choice F1 | 38.5 | Multiple-choice F1 (%) for assessing and ranking clip-level | 35.7 |
| CG-Bench | 27.1 | Long-Video MCQ Accuracy (%) | 28.6 |
| CG-Bench - Accuracy@IoU | 1.02 | Accuracy@IoU (%) | 23.8 |
| CG-Bench - Clue MCQ | 40.9 | Clue-Interval MCQ Accuracy (%) | 23.8 |
| CG-Bench - Open-Ended | 21.5 | Open-Ended Accuracy (%) | 23.8 |
| VidPair-Halluc | 19.53 | Binary-QA weighted accuracy wAcc (%): question-pair and vide | 21.4 |
| LongVideoBench | 41.8 | Test Total Accuracy (%) | 17.6 |
| VidPair-Halluc - Open-Ended Description | 34.27 | Open-ended description score (%) judged by GPT-4o on VidPair | 14.3 |
| Video-MME - Short Videos (w/ subs) | 53.6 | Accuracy (%) | 12 |
| Video-MME - Short Videos (w/o subs) | 48.3 | Accuracy (%) | 9 |
| VideoMME w sub. | 43.6 | Score (imported) | 5.5 |
Interactive version: theaggregate.ai/model?slug=sharegpt4video · How It Works · Data refreshed daily, snapshot 2026-11-08.