Visual Question Answering on ScienceQA (SQA^I) (test)
97AccuracyInternVL2-8B
Evaluation Results
| Method | Links | |
|---|---|---|
| InternVL2-8BModel=InternVL2-8B, Pruning Method=None, Pruning Ratio (r)=1.02025.09 | 97 | |
| PTPModel=InternVL2-8B, Pruning Method=PTP, Pruning Ratio (r)=0.5, Balancing Weight (alpha)=0.52025.09 | 96.7 | |
| GSearchModel=InternVL2-8B, Pruning Method=GSearch2025.09 | 94.3 | |
| InternVL2-2BModel=InternVL2-2B, Pruning Method=None, Pruning Ratio (r)=1.02025.09 | 94.1 | |
| PTPModel=InternVL2-2B, Pruning Method=PTP, Pruning Ratio (r)=0.5, Balancing Weight (alpha)=0.52025.09 | 93.9 | |
| FastVModel=InternVL2-8B, Pruning Method=FastV, Pruning Ratio (r)=0.52025.09 | 93.2 | |
| GSearchModel=InternVL2-2B, Pruning Method=GSearch2025.09 | 92.5 | |
| PDropModel=InternVL2-8B, Pruning Method=PDrop2025.09 | 92.1 | |
| FastVModel=InternVL2-2B, Pruning Method=FastV, Pruning Ratio (r)=0.52025.09 | 91 | |
| Spatially-RModel=InternVL2-2B, Pruning Method=Spatially-R, Pruning Ratio (r)=0.52025.09 | 90.4 | |
| RandomModel=InternVL2-2B, Pruning Method=Random, Pruning Ratio (r)=0.52025.09 | 90.1 | |
| PDropModel=InternVL2-2B, Pruning Method=PDrop2025.09 | 89.9 | |
| VTWModel=InternVL2-2B, Pruning Method=VTW2025.09 | 75.2 | |
| Dense ConnectorResolution=336, Tokens=144, PT+IT=0.5M+0.6M, LLM=Vicuna-7B2024.05 | 68.8 | |
| LLaVAResolution=336, Tokens=576, PT+IT=0.5M+0.6M, LLM=Vicuna-7B2024.05 | 66.8 | |
| VTWModel=InternVL2-8B, Pruning Method=VTW2025.09 | 62.3 | |
| Qwen-VL-ChatResolution=448, Tokens=256, PT+IT=1.4B+50M, LLM=Qwen-7B2024.05 | 61.5 |