Intention-to-Document on DocHTML (test)
8.64Layout ScoreAnyDoc
Evaluation Results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| AnyDocbackbone=Qwen2.5-VL-7B-Instruct2026.03 | 8.64 | 8.92 | 8.36 | 8.44 | 7.21 | 2.38 | 0.53 | 0.5 | 8.95 | |
| AnyDocbackbone=Qwen2.5-VL-7B-Instruct2026.03 | 8.62 | 8.73 | 8.39 | 8.34 | 7.36 | 4.94 | 0.21 | 8.5 | — | |
| AnyDocMethod Category=Our Method, base_model=Qwen2.5-VL-7B-Instruct2026.03 | 8.62 | 8.73 | 8.39 | 8.34 | 7.36 | 4.94 | 0.21 | 8.5 | — | |
| GPT-4omode=zero-shot2026.03 | 8.59 | 8.75 | 8.32 | 8.41 | 7.27 | 25.4 | 0.2 | 4.7 | 8.96 | |
| GPT-4omode=zero-shot2026.03 | 8.09 | 8.38 | 7.9 | 7.79 | 6.87 | 25.2 | 0.17 | 21.2 | — | |
| GPT-4oMethod Category=General MLLMs, setting=zero-shot2026.03 | 8.09 | 8.38 | 7.9 | 7.79 | 6.87 | 25.2 | 0.17 | 21.2 | — | |
| OpenCOLE2026.03 | 7.91 | 8.03 | 7.79 | 7.52 | 6.91 | — | — | — | 8.13 | |
| InternVL3-78Bmode=zero-shot2026.03 | 7.62 | 8.16 | 7.26 | 7.37 | 6.31 | 2.86 | 0.14 | 32.3 | 8.02 | |
| FLUX.1-dev2026.03 | 7.58 | 7.78 | 6.54 | 5.16 | 6.25 | — | — | — | 6.91 | |
| LaDeCo2026.03 | 7.58 | 8.25 | 7.11 | 7.32 | 6.71 | 71.5 | 0.22 | 0 | — | |
| LaDeCoMethod Category=Specialized Models2026.03 | 7.58 | 8.25 | 7.11 | 7.32 | 6.71 | 71.5 | 0.22 | 0 | — | |
| InternVL3-78Bmode=zero-shot2026.03 | 7.44 | 8.05 | 7.04 | 7.16 | 6.42 | 1.24 | 0.08 | 93.4 | — | |
| InternVL3-78BMethod Category=General MLLMs, setting=zero-shot2026.03 | 7.44 | 8.05 | 7.04 | 7.16 | 6.42 | 1.24 | 0.08 | 93.4 | — |