VQA evaluation on VLMs
1
Qwen2.5-VL-72B (SFT)
47.1
2
Qwen2.5-VL-32B (SFT)
46.8
3
Qwen2.5-VL-7B (SFT)
45.3
4
GPT-4o
29.8
5
Gemini-2.5-Pro
28.2
6
Qwen2.5-VL-32B
27.1
7
Qwen2.5-VL-72B
27.4
8
Qwen2.5-VL-7B
25.9
9
mPLUG-Owl3-7B
25.4
-
Random Chance
25.0
10
Gemini-2-Flash
24.9
11
LLaVA-OneVision-7B
24.7
12
LLaVA-Video-7B
24.1
13
InternVL2.5-26B
19.8
14
InternVL3-78B
19.8
15
InternVL2.5-8B
16.7
16
InternVL3-8B
16.7
17
InternLMXComposer2.5-7B
9.3
18
InternVideo2-Chat-8B
5.3
19
Tarsier-Recap-7B
4.8