Table 4.
Performance over five independent experimental runs on the external Sherif dataset.
| Model | Accuracy (%) | Macro-F1 (%) |
|---|---|---|
| CNN ensemble | 96.00 ± 0.12 | 94.10 ± 0.18 |
| ViT ensemble | 94.39 ± 0.21 | 91.58 ± 0.27 |
| NeuroTrustNet | 94.00 ± 0.19 | 91.00 ± 0.24 |
| RapidNet | 75.00 ± 0.31 | 68.15 ± 0.42 |
| AdaptoVision | 74.00 ± 0.35 | 67.03 ± 0.48 |
Bold values highlight the proposed NeuroTrustNet/AAS configuration.