PLLaVA-13B: benchmark results
Provider: Other.
Aggregate Rating 1502 ± 41, rank #2329 of 10916 rated models, from 11 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| VidPair-Halluc | 32.06 | Binary-QA weighted accuracy wAcc (%): question-pair and vide | 78.6 |
| CCTVBench - Balanced Accuracy | 59.81 | Balanced accuracy (%) over all binary video-question instanc | 66.7 |
| VidPair-Halluc - Open-Ended Description | 41.31 | Open-ended description score (%) judged by GPT-4o on VidPair | 42.9 |
| LongVideoBench | 45.1 | Test Total Accuracy (%) | 35.3 |
| CCTVBench | 6.19 | Quadruple accuracy (%): share of question pairs in which the | 33.3 |
| INFACT (Factuality) | 52.5 | Average reliability under induced perturbations (0-1, shown | 15.4 |
| TVBench | 36.4 | Average Accuracy | 11.1 |
| INFACT (Factuality) - Clean Accuracy | 41 | Accuracy (0-1, shown times 100) on clean video (Mode I) for | 7.7 |
| INFACT (Faithfulness) - Clean Accuracy | 45.8 | Accuracy (0-1, shown times 100) on clean video (Mode I) for | 7.7 |
| VidPair-Halluc - Multi-Choice F1 | 8.92 | Multiple-choice F1 (%) for assessing and ranking clip-level | 7.1 |
| INFACT (Faithfulness) | 52.3 | Average reliability under induced perturbations (0-1, shown | 0 |
Interactive version: theaggregate.ai/model?slug=pllava-13b · How It Works · Data refreshed daily, snapshot 2026-11-08.