Calibration
ERQA → OmniDocBench 1.5
OmniDocBench 1.5 is estimated from ERQA with a linear curve fitted on 7 models measured on both: y = 0.1604·x + 0.7913, R² = 0.45, cross-validated error 2.5 pp. It is used for 2 estimates.
| Estimated model | ERQA | OmniDocBench 1.5 | Source |
|---|---|---|---|
| Glimpse | 83.8% | 92.6% | estimated ± 2.5 pp, low confidence |
| Claude Opus 4.6 | 51.6% | 87.4% | estimated ± 2.5 pp, low confidence |
Fitted on
| Model | ERQA | OmniDocBench 1.5 |
|---|---|---|
| Claude Opus 4.5 | 46.8% | 87.7% |
| Gemini 3 Pro | 70.5% | 88.5% |
| GPT-5.2 | 59.8% | 85.7% |
| Qwen3.5 397B | 67.5% | 90.8% |
| Qwen3.7 Plus | 69.8% | 91.4% |
| Qwen3.8-27B | 65.5% | 91.1% |
| Qwen3.8 Max | 77.8% | 92.1% |