Multimodal Hallucination - VH POPE COCO: leaderboard
The "VH POPE COCO" subtask of Multimodal Hallucination. A task measuring hallucination in multimodal models — how often a vision-language model asserts objects, attributes or facts not supported by the image.
Metric: Score (%). Source: huggingface.co. 9 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Llama-VID | 90.1 | |
| 2 | InstructBLIP | 84.1 | |
| 3 | LLaVA-v1.5 | 82.9 | |
| 4 | FAVOR | 79.8 | |
| 5 | MPLUG-Owl2 | 71.9 | |
| 6 | Video-LLaVA | 64.2 | |
| 7 | Chat-Univi | 61 | |
| 8 | Video-Llama | 38.7 | |
| 9 | Valley | 38.5 |
Interactive version: theaggregate.ai/benchmark?slug=multimodal-hallucination-vh-pope-coco · How It Works · Data refreshed daily, snapshot 2026-10-20.