Document Visual Question Answering on PFL-DocVQA (test)
80AccuracyPix2Struct
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| Pix2StructExplainable=false, Patches=4,0962026.05 | 80 | 92 | — | — | — | |
| CoExVQA (Crop Variant)Explainable=true, Patches=5122026.05 | 61 | 78 | 47 | 76 | 4.92 | |
| DocVXQAExplainable=true, Patches=1,0242026.05 | 43 | 66 | — | — | — | |
| CoExVQA (Mask Variant)Explainable=true, Patches=5122026.05 | 35 | 63 | 43 | 69 | 2.85 |