Update README.md
Browse files
README.md
CHANGED
|
@@ -163,7 +163,20 @@ Example Output:
|
|
| 163 |
|
| 164 |
## Performance
|
| 165 |
|
| 166 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 167 |
|
| 168 |
Document VQA performance of different models and prompting strategies on the BoundingDocs v2.0 dataset. <br>
|
| 169 |
Smol stands for SmolVLM-2.2B, Qwen stands for Qwen2-VL-7B, and D.E. stands for DocExplainerV0. <br>
|
|
|
|
| 163 |
|
| 164 |
## Performance
|
| 165 |
|
| 166 |
+
| Architecture | Prompting | ANLS | MeanIoU |
|
| 167 |
+
|--------------------------------|------------|-------|---------|
|
| 168 |
+
| Smol | Zero-shot | 0.527 | 0.011 |
|
| 169 |
+
| | Anchors | 0.543 | 0.026 |
|
| 170 |
+
| | CoT | 0.561 | 0.011 |
|
| 171 |
+
| Qwen | Zero-shot | 0.691 | 0.048 |
|
| 172 |
+
| | Anchors | 0.694 | 0.051 |
|
| 173 |
+
| | CoT | <ins>0.720</ins> | 0.038 |
|
| 174 |
+
| Claude Sonnet 4 | Zero-shot | **0.737** | 0.031 |
|
| 175 |
+
| Smol + D.E. | Zero-shot | 0.572 | 0.175 |
|
| 176 |
+
| Qwen + D.E. | Zero-shot | 0.689 | 0.188 |
|
| 177 |
+
| Smol + Naive OCR | Zero-shot | 0.556 | <ins>0.405</ins> |
|
| 178 |
+
| Qwen + Naive OCR | Zero-shot | 0.690 | **0.494** |
|
| 179 |
+
|
| 180 |
|
| 181 |
Document VQA performance of different models and prompting strategies on the BoundingDocs v2.0 dataset. <br>
|
| 182 |
Smol stands for SmolVLM-2.2B, Qwen stands for Qwen2-VL-7B, and D.E. stands for DocExplainerV0. <br>
|