MMEB Video Classification (Kinetics-700, SSv2, HMDB, UCF, Breakfast v2)
63.2Classification AccuracyFreeRet
Evaluation results
| Method | Links | |
|---|---|---|
| FreeRetMLLM=Qwen2-VL-7B, Train data (M)=–, Video data=–2025.09 | 63.2 | |
| FreeRetMLLM=Qwen2-VL-2B, Train data (M)=–, Video data=–2025.09 | 58.3 | |
| FreeRet-embedMLLM=Qwen2-VL-7B, Train data (M)=–, Video data=–2025.09 | 54.3 | |
| FreeRet-embedMLLM=Qwen2-VL-2B, Train data (M)=–, Video data=–2025.09 | 47.7 | |
| LamRAMLLM=Qwen2-VL-7B, Train data (M)=1.4, Video data=✗2025.09 | 39.3 | |
| VLM2Vec-V2MLLM=Qwen2-VL-2B, Train data (M)=1.7, Video data=✓2025.09 | 39.3 | |
| VLM2VecMLLM=Qwen2-VL-7B, Train data (M)=1.0, Video data=✗2025.09 | 39.1 | |
| GMEMLLM=Qwen2-VL-7B, Train data (M)=8.0, Video data=✗2025.09 | 37.4 | |
| GMEMLLM=Qwen2-VL-2B, Train data (M)=8.0, Video data=✗2025.09 | 34.9 | |
| VLM2VecMLLM=Qwen2-VL-2B, Train data (M)=1.0, Video data=✗2025.09 | 33.4 |