llava-onevision-qwen2-7B-ov: benchmark results
Provider: LLaVA Team. Access: Open.
Aggregate Rating 1471 ± 28, rank #2832 of 10916 rated models, from 47 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| SenseBench - Single-Image Perception | 64.15 | Exact-Match Accuracy (%; mean of the whether, what and how q | 92.9 |
| MIOH - Existence | 60.9 | Accuracy (%; existence questions about whether objects are p | 81.5 |
| KOFFVQA - Hallucination and Robustness | 80 | Score (%) | 80.9 |
| SenseBench - Single-Image Description | 3.87 | Judge Score Sum (0-6; Prometheus completeness, correctness a | 65.4 |
| MIOH | 36.4 | Accuracy (%; mean of the existence, counting, attribute and | 59.3 |
| SAW-Bench - Reverse Route Plan | 25.33 | Accuracy (%) on the 229 reverse route plan questions (the mo | 52.3 |
| MIOH - Position | 34.4 | Accuracy (%; position questions about object positions acros | 50 |
| MIOH - Counting | 23.8 | Accuracy (%; counting questions about how many objects there | 48.1 |
| SenseBench - Paired-Image Description | 2.42 | Judge Score Sum (0-6; Prometheus completeness, correctness a | 47.6 |
| MIBench (Multimodal Interaction) - Vision-Centric | 55.5 | Accuracy (%) on the vision-centric samples (the answer must | 42.9 |
| MIBench (Multimodal Interaction) - Synergy | 42.81 | Accuracy (%) on the synergy samples (both modalities are nee | 38.1 |
| SenseBench - Paired-Image Perception | 52.9 | Exact-Match Accuracy (%; mean of the whether, what and how q | 38.1 |
Interactive version: theaggregate.ai/model?slug=llava-onevision-qwen2-7b-ov · How It Works · Data refreshed daily, snapshot 2026-11-08.