LLaVA-NeXT-7B: benchmark results
Provider: LLaVA Team.
Aggregate Rating 1435 ± 22, rank #4425 of 10916 rated models, from 126 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| WildRoadBench - Drainage Water | 13 | F1 (%) of predicted boxes at IoU 0.5 for the drainage water | 95.8 |
| EIBench (complex) | 38.71 | Recall | 83.3 |
| CogRail RailThreat (Cog-MRSI, Type-II Visual Prompt) | 39.93 | Overall F1 (%; harmonic mean of macro precision and macro re | 75 |
| GeoMMBench - Thermal Imagery | 53.6 | Micro-averaged accuracy (%) on the thermal imagery questions | 70.6 |
| CogRail RailThreat (Cog-MRSI, Type-I Visual Prompt) | 52.31 | Overall F1 (%; harmonic mean of macro precision and macro re | 66.7 |
| EIBench | 53.82 | Recall | 66.7 |
| WildRoadBench - Road Water | 8 | F1 (%) of predicted boxes at IoU 0.5 for the road water scen | 62.5 |
| GeoMMBench - Spatial | 51.4 | Micro-averaged accuracy (%) on the spatial relation question | 61.8 |
| EgoPoint-Bench - Real-World Adversarial Resilience | 56.19 | Accuracy (%) on the Adversarial Resilience questions among t | 54.5 |
| EgoPoint-Bench - Simulation Adversarial Resilience | 46.59 | Accuracy (%) on the Adversarial Resilience questions among t | 54.5 |
| FAQ (Deepfake Forensics) - Facial Perception | 40 | Accuracy (%) on FAQ's Level 1 facial-perception questions: f | 54.2 |
| PlantMicro - Localization mAP@0.5 | 28.1 | Mean average precision at IoU 0.5 (%) of directly predicted | 50 |
Interactive version: theaggregate.ai/model?slug=llava-next-7b · How It Works · Data refreshed daily, snapshot 2026-11-08.