LLaVA-1.6-7B: benchmark results
Provider: LLaVA Team.
Aggregate Rating 1450 ± 29, rank #3785 of 10916 rated models, from 39 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| AICA-Bench - Emotion Reasoning | 73.81 | Emotion reasoning score on a percent scale: the model explai | 72.7 |
| Med-RewardBench | 57.77 | Expert-Agreement Accuracy (%; pairwise, six dimensions) | 66.7 |
| MLVU-Test - Action Order | 21.4 | Multiple-choice accuracy (%) | 57.4 |
| MLVU-Test - Action Count | 16.7 | Multiple-choice accuracy (%) | 50 |
| MirrorBench - Level 2 (Implicit Mirror Discovery) | -0.08 | Average score (mean of task success rate, step-wise improvem | 47.1 |
| MirrorBench - Level 3 (Self-Referential Mirror Recognition) | -0.25 | Average score (mean of task success rate, step-wise improvem | 47.1 |
| MLVU-Test - Plot QA | 30 | Multiple-choice accuracy (%) | 41.2 |
| MLVU-Test | 25.3 | Multiple-choice accuracy (%) | 32.4 |
| MLVU-Test - Ego Reasoning | 26.4 | Multiple-choice accuracy (%) | 32.4 |
| MLVU-Test - Topic Reasoning | 63.7 | Multiple-choice accuracy (%) | 32.4 |
| MLVU-Dev | 39.3 | Multiple-choice accuracy (%) | 31.6 |
| MirrorBench | -0.28 | Overall average score (mean of task success rate, step-wise | 29.4 |
Interactive version: theaggregate.ai/model?slug=llava-1-6-7b · How It Works · Data refreshed daily, snapshot 2026-11-08.