Document Policy Preservation (Implicit Query) on Doc-PP Image 1.0 (test)
75.7Faithfulness (No Policy)Qwen3-VL-235B-A22B-Thinking
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Qwen3-VL-235B-A22B-ThinkingInput Modality=Image, Query Type=Implicit2026.01 | 75.7 | 63.4 | 40.4 | |
| GPT-5.2Input Modality=Image, Query Type=Implicit2026.01 | 74.2 | 68.7 | 25.7 | |
| Gemini-3-Flash-PreviewInput Modality=Image, Query Type=Implicit2026.01 | 70.2 | 69 | 64.6 | |
| Qwen3-VL-235B-A22B-InstructInput Modality=Image, Query Type=Implicit2026.01 | 67.5 | 61.6 | 93.5 | |
| Gemini-3-Pro-PreviewInput Modality=Image, Query Type=Implicit2026.01 | 64.6 | 60.5 | 46.8 | |
| Mistral-Large-2512Input Modality=Image, Query Type=Implicit2026.01 | 64.6 | 60 | 76.8 |