Text-based Visual Question Answering on TextVQA (Efficiency/Latency)
1,668Token CountInternVL2
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| InternVL2Parameters=2B2024.12 | 1,668 | 5.09 | 8.8 | 0.24 | 3.2 | |
| Qwen2-VLParameters=2B2024.12 | 977 | 4.14 | 40.8 | 0.56 | 1.4 | |
| IMPParameters=3.1B2024.12 | 729 | 0.31 | 8.4 | 0.12 | 7.1 | |
| LLaVAParameters=7B, Version=V1.52024.12 | 576 | 4.76 | 16 | 0.11 | 5.4 | |
| FlashSloth-HDParameters=3.2B2024.12 | 414 | 2.83 | 8.1 | 0.11 | 5.6 | |
| FlashSlothParameters=3.2B2024.12 | 90 | 0.1 | 7.6 | 0.05 | 9.8 |