Video Quality Justification on S2I-Bench
3.02CI ScoreLLaVA-OneVision
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| LLaVA-OneVisionLLM Variant=Vicuna-v1.1-7B, S2I-Tuned=Yes2025.06 | 3.02 | 2.49 | 2.13 | 2.24 | 9.88 | |
| InternVL-ChatLLM Variant=Vicuna-7B, S2I-Tuned=Yes2025.06 | 2.68 | 2.74 | 2.11 | 2.19 | 9.72 | |
| LLaVA-Next-VideoLLM Variant=Mistral-7B, S2I-Tuned=Yes2025.06 | 2.43 | 2.35 | 2.16 | 2.07 | 9.01 | |
| VideoLLaMA3LLM Variant=Qwen2.5-7B, S2I-Tuned=Yes2025.06 | 2.38 | 2.28 | 2.25 | 2.36 | 9.27 | |
| Video-LLaVALLM Variant=Vicuna-v1.5-7B, S2I-Tuned=Yes2025.06 | 2.28 | 2.37 | 1.97 | 1.88 | 8.5 | |
| GPT-4o2025.06 | 2.28 | 2.23 | 2.11 | 2.22 | 8.84 | |
| LLaVA-VideoLLM Variant=Qwen2-7B, S2I-Tuned=Yes2025.06 | 2.22 | 2.3 | 2.25 | 2.31 | 9.08 | |
| Gemini 1.5 Pro2025.06 | 2.21 | 2.22 | 2.15 | 1.95 | 8.53 | |
| InternVL-ChatLLM Variant=Vicuna-7B, S2I-Tuned=No (Baseline)2025.06 | 2.19 | 2.26 | 1.9 | 1.67 | 8.02 | |
| VideoLLaMA3LLM Variant=Qwen2.5-7B, S2I-Tuned=No (Baseline)2025.06 | 2.14 | 2.14 | 1.97 | 2.06 | 8.31 | |
| VideoLLaMA3-7Bbaseline=true2025.06 | 2.14 | 2.14 | 1.97 | 2.06 | 8.31 | |
| LLaVA-OneVisionLLM Variant=Vicuna-v1.1-7B, S2I-Tuned=No (Baseline)2025.06 | 2.11 | 2.3 | 1.83 | 1.72 | 7.96 | |
| GPT-4o mini2025.06 | 2.11 | 2.08 | 2.09 | 1.97 | 8.25 | |
| LLaVA-Next-VideoLLM Variant=Mistral-7B, S2I-Tuned=No (Baseline)2025.06 | 2.08 | 2.19 | 1.77 | 1.64 | 7.68 | |
| LLaVA-VideoLLM Variant=Qwen2-7B, S2I-Tuned=No (Baseline)2025.06 | 2.05 | 2.08 | 2.06 | 1.98 | 8.17 | |
| LLaVA-Video-7Bbaseline=true2025.06 | 2.05 | 2.08 | 2.06 | 1.98 | 8.17 | |
| Video-LLaVALLM Variant=Vicuna-v1.5-7B, S2I-Tuned=No (Baseline)2025.06 | 2.03 | 2.06 | 1.6 | 1.46 | 7.15 |