Video-LLaVA-7B: benchmark results
Provider: Other.
Aggregate Rating 1473 ± 24, rank #2790 of 10916 rated models, from 66 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Flat-Pack Bench - Temporal Localization | 35.92 | TLoc: identify the events just before or after the state sho | 75.9 |
| TOC-Bench - Event Existence | 51.1 | Event Existence accuracy (%) on TOC-Bench (2,323 human-verif | 75 |
| TOC-Bench - Reappearance or Disappearance | 59.7 | Reappearance or Disappearance accuracy (%) on TOC-Bench (2,3 | 75 |
| Flat-Pack Bench - Part Mating | 51.72 | Mate: decide whether two parts are connected in the final as | 74.1 |
| MLVU-Test - Sports QA | 38.9 | Multiple-choice accuracy (%) | 61.8 |
| MLVU-Test - Action Count | 21.7 | Multiple-choice accuracy (%) | 60.3 |
| MLVU-Test - Action Order | 20 | Multiple-choice accuracy (%) | 50 |
| MLVU-Test - Generation | 3.68 | Generation score, GPT-4 judged (0-10) | 50 |
| TOC-Bench - Duration Category | 23.7 | Duration Category accuracy (%) on TOC-Bench (2,323 human-ver | 50 |
| MLVU-Test - Anomaly Recognition | 38.5 | Multiple-choice accuracy (%) | 48.5 |
| MLVU-Dev | 47.3 | Multiple-choice accuracy (%) | 47.4 |
| Vinoground (Video Score) | 25.8 | Video Score | 43.5 |
Interactive version: theaggregate.ai/model?slug=video-llava-7b · How It Works · Data refreshed daily, snapshot 2026-11-08.