Multimodal Question Answering on CRIT (Video Frame) 1.0 (test)
38.8Exact Match (EM)Qwen2.5-VL_CRIT
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Qwen2.5-VL_CRITModel Parameters=7B, Evaluation Setting=CoT, Fine-tuned on CRIT=Yes2026.04 | 38.8 | 42.2 | |
| LLaVA-OnevisionModel Parameters=7B, Evaluation Setting=CoT2026.04 | 33.3 | 37.4 | |
| GPT-4oEvaluation Setting=CoT2026.04 | 32 | 38.9 | |
| Idefics2_CRITModel Parameters=8B, Evaluation Setting=CoT, Fine-tuned on CRIT=Yes2026.04 | 31.2 | 33.9 | |
| Gemini 2.0 FlashEvaluation Setting=CoT2026.04 | 30.6 | 35.3 | |
| Intern2.5-VLModel Parameters=8B, Evaluation Setting=CoT2026.04 | 30.1 | 34.8 | |
| Qwen2.5-VLModel Parameters=72B, Evaluation Setting=CoT2026.04 | 30.1 | 33.9 | |
| GPT-4o-miniEvaluation Setting=CoT2026.04 | 28.7 | 33.3 | |
| Phi3.5-VisionModel Parameters=4B, Evaluation Setting=CoT2026.04 | 24 | 26.5 | |
| Qwen2.5-VLModel Parameters=7B, Evaluation Setting=CoT2026.04 | 24 | 27.8 | |
| Idefics2Model Parameters=8B, Evaluation Setting=CoT2026.04 | 20.2 | 25.2 |