Table Question Answering on TableVQA DenTab (test)
68.5Sum AccuracyQwen3-VL-32B
Evaluation results
| Method | Links | |||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Qwen3-VL-32BModel Size=32B2026.04 | 68.5 | 67.8 | 93.3 | 61.9 | 53.3 | 48.5 | 93.6 | 82.9 | 64.9 | 89.3 | 73.2 | |
| Qwen3-VL-30B-A3BModel Size=30B-A3B2026.04 | 65 | 67.2 | 92.7 | 62.7 | 44 | 25 | 90.8 | 71.4 | 50.5 | 90.9 | 66.2 | |
| Qwen3-VL-8BModel Size=8B2026.04 | 62.1 | 45.9 | 89.7 | 55.5 | 29.3 | 16.3 | 88.4 | 77.4 | 60.1 | 86.9 | 61.7 | |
| Gemma-3-27BModel Size=27B2026.04 | 58.1 | 48.1 | 82.4 | 59.3 | 48 | 13.3 | 92 | 66.7 | 37.2 | 76.2 | 56.6 | |
| Qwen3-VL-4BModel Size=4B2026.04 | 57.6 | 40.4 | 90.9 | 45.3 | 18.7 | 11.7 | 88.4 | 76.2 | 34.6 | 81.7 | 55.8 | |
| olmOCR-2-7BModel Size=7B, Type=OCR-centric2026.04 | 56.7 | 32.2 | 89.7 | 41.5 | 21.3 | 9.2 | 85.3 | 70.2 | 47.3 | 78.2 | 54.5 | |
| Gemma-3-12BModel Size=12B2026.04 | 53.2 | 38.3 | 78.8 | 52.1 | 26.7 | 9.2 | 89.6 | 65.9 | 33 | 73.4 | 52.2 | |
| Qwen2.5-VL-7BModel Size=7B2026.04 | 51.7 | 42.6 | 81.2 | 48.7 | 21.3 | 11.7 | 85.7 | 70.6 | 45.2 | 78.2 | 56 | |
| Ministral-3-8BModel Size=8B2026.04 | 36.5 | 38.8 | 50.9 | 48.7 | 33.3 | 11.7 | 66.9 | 54.8 | 35.1 | 59.5 | 44.9 | |
| Gemma-3-4BModel Size=4B2026.04 | 34.5 | 16.9 | 59.4 | 34.3 | 22.7 | 6.1 | 55.4 | 52 | 5.9 | 38.9 | 32.3 | |
| Nanonets-OCR2-3BModel Size=3B, Type=OCR-centric2026.04 | 27.6 | 16.4 | 27.9 | 19.1 | 4 | 7.7 | 27.1 | 52.4 | 11.7 | 46.8 | 27.5 | |
| Idefics3-8BModel Size=8B2026.04 | 25.1 | 8.2 | 38.2 | 25.8 | 6.7 | 7.1 | 50.6 | 45.2 | 6.9 | 53.6 | 30.1 | |
| Nanonets-OCR-sType=OCR-centric2026.04 | 14.3 | 7.7 | 37 | 30.5 | 10.7 | 5.6 | 37.1 | 53.6 | 14.9 | 58.7 | 29.9 |