CRoMapathologyrobustnessH&E
Cross-confounder robustness margin measuring how much of a model's embedding neighbourhood is driven by biology rather than by the acquisition center. Builds on PathoROB's Robustness Index and scores 25 encoders on its three tile cohorts, reporting a median margin alongside a lower-tail severity.
Detailed Results
Showing 3 of 3 tile cohorts. CRoMa is the median signed margin at the headline radius m=5, in (-1, 1) and neutral at 0. Each cell shows it above its lower-tail mean LTM₁₀. Models are ordered by the mean of the CRoMa rank and the tail rank (lower is better) — a reading order, not a score.
| Model | Mean rank | CRoMa rank | Tail rank | Average CRoMa | Camelyon CRoMa | TCGA-4x4 CRoMa | Tolkach ESCA CRoMa |
|---|---|---|---|---|---|---|---|
| 1.7 | 2.0 | 1.3 | 0.354 | 0.285tail -0.016 | 0.270tail -0.109 | 0.507tail 0.009 | |
| 4.3 | 2.3 | 6.3 | 0.334 | 0.324tail -0.024 | 0.192tail -0.163 | 0.485tail -0.003 | |
| 4.5 | 4.3 | 4.7 | 0.275 | 0.243tail -0.040 | 0.168tail -0.124 | 0.412tail -0.011 | |
| 5.5 | 4.7 | 6.3 | 0.270 | 0.235tail -0.047 | 0.169tail -0.137 | 0.407tail -0.024 | |
| 7.5 | 7.3 | 7.7 | 0.244 | 0.187tail -0.145 | 0.153tail -0.130 | 0.392tail -0.029 | |
| 8.5 | 7.0 | 10.0 | 0.246 | 0.191tail -0.069 | 0.155tail -0.193 | 0.390tail -0.022 | |
| 9.2 | 8.3 | 10.0 | 0.226 | 0.199tail -0.106 | 0.128tail -0.170 | 0.351tail -0.037 | |
| 9.2 | 6.0 | 12.3 | 0.261 | 0.196tail -0.202 | 0.146tail -0.152 | 0.439tail -0.041 | |
| 9.3 | 12.7 | 6.0 | 0.143 | 0.082tail -0.138 | 0.088tail -0.103 | 0.260tail -0.042 | |
| 10.0 | 12.7 | 7.3 | 0.147 | 0.111tail -0.177 | 0.081tail -0.110 | 0.249tail -0.030 | |
| 12.0 | 9.3 | 14.7 | 0.222 | 0.167tail -0.160 | 0.121tail -0.188 | 0.380tail -0.074 | |
12Virchow | 12.2 | 10.3 | 14.0 | 0.206 | 0.155tail -0.181 | 0.092tail -0.180 | 0.371tail -0.055 |
| 12.2 | 4.7 | 19.7 | 0.363 | 0.108tail -0.349 | 0.396tail -0.210 | 0.584tail -0.083 | |
| 12.3 | 15.3 | 9.3 | 0.110 | 0.045tail -0.154 | 0.054tail -0.111 | 0.232tail -0.080 | |
15UNI2 | 13.3 | 15.3 | 11.3 | 0.114 | 0.045tail -0.209 | 0.075tail -0.123 | 0.221tail -0.063 |
| 14.3 | 16.7 | 12.0 | 0.090 | 0.022tail -0.177 | 0.059tail -0.124 | 0.187tail -0.089 | |
| 14.5 | 15.0 | 14.0 | 0.129 | 0.042tail -0.220 | 0.051tail -0.140 | 0.293tail -0.066 | |
18UNI | 16.0 | 19.3 | 12.7 | 0.063 | -0.034tail -0.217 | 0.047tail -0.120 | 0.175tail -0.080 |
| 17.3 | 19.3 | 15.3 | 0.064 | 0.009tail -0.190 | 0.051tail -0.151 | 0.132tail -0.160 | |
| 18.3 | 19.0 | 17.7 | 0.058 | -0.104tail -0.361 | 0.037tail -0.153 | 0.241tail -0.100 | |
21Hibou-B | 20.5 | 21.3 | 19.7 | 0.027 | -0.089tail -0.363 | 0.036tail -0.177 | 0.134tail -0.167 |
| 21.7 | 21.0 | 22.3 | 0.005 | -0.196tail -0.476 | 0.036tail -0.200 | 0.175tail -0.188 | |
| 22.5 | 23.7 | 21.3 | -0.019 | -0.208tail -0.496 | 0.029tail -0.187 | 0.122tail -0.175 | |
| 24.0 | 24.0 | 24.0 | -0.067 | -0.318tail -0.640 | -0.010tail -0.250 | 0.127tail -0.236 | |
25Hibou-L | 24.2 | 23.3 | 25.0 | -0.097 | -0.443tail -0.659 | 0.039tail -0.252 | 0.111tail -0.291 |
★ marks the Pareto frontier — the encoders no other pathology encoder beats on both the CRoMa rank and the tail rank at once. It is a set, not an order.
TCGA marks encoders whose disclosed pretraining overlaps TCGA, one of the three cohorts behind these ranks.
CRoMa also scores DINOv2-B as a natural-image calibration floor and PRISM/PRISM2/TITAN/MOOZY on a separate slide-level cohort. Neither takes part in the ranks above, and generic vision baselines are out of scope for Histoboard — see the full panel.