General Multimodal Understanding on Aggregated Benchmarks
71Average ScoreQwen2.5-VL-3B
Evaluation results
| Method | Links | |
|---|---|---|
| Qwen2.5-VL-3BKV Size=2562026.05 | 71 | |
| Cambrian-4B + Gaze attentionKV Size=288+42026.05 | 69.6 | |
| Cambrian-4B + Gaze attentionKV Size=144+42026.05 | 69 | |
| Qwen2.5-VL-3B + InfiniPot-VKV Size=1282026.05 | 68.8 | |
| LLaVA-OV-7BKV Size=1962026.05 | 68.7 | |
| Cambrian-4BKV Size=5762026.05 | 68.6 | |
| Cambrian-4B + Gaze attentionKV Size=72+42026.05 | 68.1 | |
| Cambrian-4B + TokenCompre. + Gaze attentionKV Size=54+42026.05 | 67.4 | |
| Cambrian-4B + TokenCompre.KV Size=1442026.05 | 66.5 | |
| Cambrian-4B + TokenCompre. + Gaze attentionKV Size=36+42026.05 | 66.4 | |
| LLaVA-OV-7B + HERMESKV Size=1002026.05 | 64.7 | |
| Qwen2.5-VL-3B + InfiniPot-VKV Size=502026.05 | 61.4 | |
| LLaVA-OV-7B + HERMESKV Size=402026.05 | 58.7 |