Object Hallucination Evaluation on HallB (Hallucination-Bench)
64.7ScoreInternVL2-8B
Evaluation Results
| Method | Links | |
|---|---|---|
| InternVL2-8BModel size=8B2025.04 | 64.7 | |
| LyricsLLM Backbone=Vicuna-13B2023.12 | 62.6 | |
| Qwen2VL-2BModel size=2B2025.04 | 61.2 | |
| Qwen2VL-7BModel size=7B2025.04 | 59.9 | |
| InternVL2-2BModel size=2B2025.04 | 59.3 | |
| SpaRE-2BModel size=2B2025.04 | 58.2 | |
| Qwen-VL-ChatLLM Backbone=Qwen-7B2023.12 | 56.4 | |
| SpaRE-7BModel size=7B2025.04 | 56.3 | |
| InstructBLIPLLM Backbone=Vicuna-7B2023.12 | 53.6 | |
| ShareGPT4VLLM Backbone=Vicuna-7B2023.12 | 49.8 | |
| LLaVA-1.5LLM Backbone=Vicuna-13B2023.12 | 46.7 | |
| IDEFICS-80BLLM Backbone=LLaMA-65B2023.12 | 46.1 | |
| LLaVALLM Backbone=Vicuna-7B2023.12 | 44.1 | |
| SpaceLLaVa2025.04 | 43.9 | |
| LLaVA-NeXT-8BModel size=8B2025.04 | 39 |