Claims
| Claim | Type | Assessment | Score | Review checks | References |
|---|---|---|---|---|---|
| Embedding a well-log-based sweet-spot index as a geological prior in an attention-gated KG-SE-ResNet18 model improved reservoir-type classification accuracy from 86.95% to 89.28% in the Jiyuan Oilfield case study. | QUANTITATIVE | Supported | 0.9734 | 1 of 4 | 121 |
| In a 12-cored-well case study from the Chang 8 Member of China's Jiyuan Oilfield, an SE-ResNet18 well-log model achieved R² 0.86 and RMSE 0.287 mD for tight-sandstone permeability regression. | QUANTITATIVE | Supported | 0.8581 | 1 of 4 | 111 |