Table 3.
Balanced accuracy comparison of CarLab 2025 and other frameworks with in-domain testing.
| Framework | Balanced accuracy |
|---|---|
| CaRLab 2025 | 67.20 |
| USVAC 2025 Level 1 | 61.03 |
| Compton 2022 | 56.61 |
| da Silva Moura 2024 | |
| Za’im 2023 | 56.60 |
Bold values mark the best-performing configuration.