Efficiency Metrics for Visual Question Answering on GQA
1,699Token CountInternVL2
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| InternVL2Parameters=2B2024.12 | 1,699 | 5.1 | 9.3 | 0.24 | 3.7 | |
| IMPParameters=3.1B2024.12 | 729 | 0.3 | 8.4 | 0.12 | 9.2 | |
| LLaVAParameters=7B, Version=V1.52024.12 | 576 | 4.41 | 15.3 | 0.11 | 7.4 | |
| FlashSloth-HDParameters=3.2B2024.12 | 414 | 2.71 | 7.8 | 0.11 | 7.1 | |
| Qwen2-VLParameters=2B2024.12 | 352 | 1.5 | 9.3 | 0.13 | 5.5 | |
| FlashSlothParameters=3.2B2024.12 | 90 | 0.08 | 7.6 | 0.05 | 12.6 |