Science Question Answering on SQA IMG
97.67ScoreTopV
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| TopVModel=InternVL2-26B, FLOPs Ratio=47%2025.03 | 97.67 | 56.73 | 2,426 | 2.75 | |
| BaselineModel=InternVL2-26B, FLOPs Ratio=02025.03 | 97.47 | 57.61 | 2,843 | 2.34 | |
| FastVModel=InternVL2-26B, FLOPs Ratio=46%2025.03 | 96.82 | 58.12 | 3,312 | 2.02 | |
| TopVModel=InternVL2-2B, FLOPs Ratio=48%2025.03 | 94.22 | 5.99 | 450 | 6.39 | |
| BaselineModel=InternVL2-2B, FLOPs Ratio=02025.03 | 94.2 | 6.63 | 607 | 5.5 | |
| FastVModel=InternVL2-2B, FLOPs Ratio=47%2025.03 | 93.75 | 7.69 | 745 | 4.34 | |
| S-MFTLanguage Model Backbone=Qwen2.5-7B, Vision Tower=SigLIP-ViT-Large2025.12 | 78.1 | — | — | — | |
| FFTLanguage Model Backbone=Qwen2.5-7B, Vision Tower=SigLIP-ViT-Large2025.12 | 76.4 | — | — | — | |
| LoRALanguage Model Backbone=Qwen2.5-7B, Vision Tower=SigLIP-ViT-Large2025.12 | 74.5 | — | — | — | |
| TopVModel=LLaVA-v1.5-13B, FLOPs Ratio=50%2025.03 | 73.27 | 28.66 | 522 | 6.57 | |
| FastVModel=LLaVA-v1.5-13B, FLOPs Ratio=48%2025.03 | 73.17 | 29.51 | 702 | 4.78 | |
| TopVModel=LLaVA-v1.5-13B, FLOPs Ratio=35%2025.03 | 73.12 | 28.32 | 507 | 6.26 | |
| BaselineModel=LLaVA-v1.5-13B, FLOPs Ratio=02025.03 | 72.93 | 28.92 | 615 | 5.3 | |
| S-MFTLanguage Model Backbone=Vicuna-7B, Vision Tower=CLIP-ViT-Large2025.12 | 70.3 | — | — | — | |
| LoRALanguage Model Backbone=Vicuna-7B, Vision Tower=CLIP-ViT-Large2025.12 | 69.7 | — | — | — | |
| TopVModel=LLaVA-v1.5-7B, FLOPs Ratio=51%2025.03 | 69.61 | 14.17 | 532 | 6.27 | |
| FastVModel=LLaVA-v1.5-7B, FLOPs Ratio=47%2025.03 | 69.56 | 14.17 | 532 | 6.27 | |
| BaselineModel=LLaVA-v1.5-7B, FLOPs Ratio=02025.03 | 69.51 | 14.37 | 633 | 5.18 | |
| FFTLanguage Model Backbone=Vicuna-7B, Vision Tower=CLIP-ViT-Large2025.12 | 69.5 | — | — | — | |
| S-MFTLanguage Model Backbone=LLaMA2-7B, Vision Tower=CLIP-ViT-Large2025.12 | 69 | — | — | — | |
| TopVModel=LLaVA-v1.5-7B, FLOPs Ratio=35%2025.03 | 68.91 | 14.3 | 602 | 6.18 | |
| FFTLanguage Model Backbone=LLaMA2-7B, Vision Tower=CLIP-ViT-Large2025.12 | 68.8 | — | — | — | |
| LoRALanguage Model Backbone=LLaMA2-7B, Vision Tower=CLIP-ViT-Large2025.12 | 68.3 | — | — | — |