Video Action Retrieval on SOVABench Inter-pair 1.0
38.3mAPMiniCPM-V 4.5
Evaluation results
| Method | Links | |
|---|---|---|
| MiniCPM-V 4.5Category=General MLLM, Prompting Strategy=TASK-AWARE, Framework=MLLM-to-Embedding framework2026.01 | 38.3 | |
| CLIP4ClipCategory=Contrastive Video-VLM2026.01 | 36.6 | |
| InternVL3.5 8BCategory=General MLLM, Prompting Strategy=TASK-AWARE, Framework=MLLM-to-Embedding framework2026.01 | 35.4 | |
| VideoCLIPCategory=Contrastive Video-VLM2026.01 | 34.5 | |
| MiniCPM-V 4.5Category=General MLLM, Prompting Strategy=GENERAL, Framework=MLLM-to-Embedding framework2026.01 | 34.4 | |
| Gemini 2.5 FlashCategory=API MLLM, Prompting Strategy=TASK-AWARE, Framework=MLLM-to-Embedding framework2026.01 | 33.2 | |
| ActionCLIPCategory=Contrastive Video-VLM2026.01 | 32.8 | |
| VideoLlama3 7BCategory=Video-MLLM, Prompting Strategy=GENERAL, Framework=MLLM-to-Embedding framework2026.01 | 32.4 | |
| VideoChat-R1 7BCategory=Video-MLLM, Prompting Strategy=TASK-AWARE, Framework=MLLM-to-Embedding framework2026.01 | 31.6 | |
| SigLIP2-GiantCategory=Contrastive Image-VLM2026.01 | 30.6 | |
| VideoLlama3 7BCategory=Video-MLLM, Prompting Strategy=TASK-AWARE, Framework=MLLM-to-Embedding framework2026.01 | 29.2 | |
| CLIP-ViT-L-14Category=Contrastive Image-VLM, Backbone=ViT-L-142026.01 | 29.1 | |
| Qwen3-VL 235B A22BCategory=API MLLM, Prompting Strategy=TASK-AWARE, Framework=MLLM-to-Embedding framework2026.01 | 29.1 | |
| MERUCategory=Contrastive Image-VLM2026.01 | 28.6 | |
| Gemini 2.5 FlashCategory=API MLLM, Prompting Strategy=GENERAL, Framework=MLLM-to-Embedding framework2026.01 | 27.9 | |
| InternVL3.5 8BCategory=General MLLM, Prompting Strategy=GENERAL, Framework=MLLM-to-Embedding framework2026.01 | 27.7 | |
| VideoLLaVA 7BCategory=Video-MLLM, Prompting Strategy=TASK-AWARE, Framework=MLLM-to-Embedding framework2026.01 | 27.3 | |
| VideoChat-R1 7BCategory=Video-MLLM, Prompting Strategy=GENERAL, Framework=MLLM-to-Embedding framework2026.01 | 25.1 | |
| VideoLLaVA 7BCategory=Video-MLLM, Prompting Strategy=GENERAL, Framework=MLLM-to-Embedding framework2026.01 | 24.9 | |
| Qwen3-VL 235B A22BCategory=API MLLM, Prompting Strategy=GENERAL, Framework=MLLM-to-Embedding framework2026.01 | 14.7 | |
| RandomCategory=Baseline2026.01 | 3.4 |