LLaVA-OneVision-1.5-8B: benchmark results
Provider: LLaVA Team.
Aggregate Rating 1540 ± 18, rank #1834 of 10916 rated models, from 155 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| InsightVQA - Insight Ranking Top-1 | 67.5 | Top-1 accuracy (%): share of situational-judgment items whos | 100 |
| Sens-VisualNews | 87.6 | Top-1 accuracy (%) on the test split of the full dataset (ma | 100 |
| Sens-VisualNews - Strict | 93.9 | Top-1 accuracy (%) on the test split of the strict subset (s | 100 |
| InsightVQA - Emotion Perception | 56.74 | Accuracy (%) on single-choice emotion-category and valence r | 84.6 |
| EgoCoT-Bench - Next Action Anticipation | 69.19 | Answer accuracy (%, four-way multiple choice, strict exact m | 83.3 |
| LinkS2Bench - Region Reachability | 61.5 | Region Reachability: yes/no accuracy (%) on whether a target | 76.5 |
| OpenRef - Negative Rejection | 75.1 | N3R (0-100), negative relative rejection reliability: mean o | 75 |
| GST-Bench - Trajectory Selection | 26.54 | Accuracy (%; multiple choice of the camera trajectory drawn | 71.4 |
| PinpointQA - Target Presence Verification | 76 | Exact-match accuracy (times 100, micro over QA pairs) of yes | 71.4 |
| InsightVQA - Intent Reasoning | 45.97 | Accuracy (%) on single-choice questions asking for the most | 69.2 |
| CourtSI-Bench - Camera-Object Distance | 3.7 | Threshold mean relative accuracy (%), averaged over confiden | 68.2 |
| CourtSI-Bench - Object-Object Distance | 26.9 | Threshold mean relative accuracy (%), averaged over confiden | 68.2 |
Interactive version: theaggregate.ai/model?slug=llava-onevision-1-5-8b · How It Works · Data refreshed daily, snapshot 2026-11-08.