croma¶

croma measures how much a pathology foundation model’s representation is driven by biology rather than by non-biological technical variation – staining, scanning, tissue preparation – across centers.

It implements three complementary neighbourhood metrics:

RI, the Robustness Index, which counts favourable versus unfavourable neighbours
MaRI, the Margin-aware Robustness Index, which weights that same evidence by distance
CRoMa, the Cross-confounder Robustness Margin, a signed margin that also supports tail-aware reporting

RI was introduced in the PathoROB study. croma provides a clean re-implementation of it, adds MaRI as its margin-aware extension, and introduces CRoMa, which overcomes limitations of both. MaRI and CRoMa are described in Beyond counts: A distributional robustness margin for pathology foundation models.

pip install croma

The panel at a glance¶

Top 8 of 25 ranked pathology encoders¶
Model	mean rank	`CRoMa` rank	tail rank	Camelyon	TCGA-4×4	Tolkach-ESCA
Mascaret (TCGA-exposed pretraining)	1.7	2.0	1.3	0.29/-0.02	0.27/-0.11	0.51/0.01
RudolfV-2-S	4.3	2.3	6.3	0.32/-0.02	0.19/-0.16	0.49/-0.00
RudolfV-2	4.5	4.3	4.7	0.24/-0.04	0.17/-0.12	0.41/-0.01
RudolfV-2-B	5.5	4.7	6.3	0.24/-0.05	0.17/-0.14	0.41/-0.02
CONCHv1.5	7.5	7.3	7.7	0.19/-0.14	0.15/-0.13	0.39/-0.03
GenBio-PathFM (TCGA-exposed pretraining)	8.5	7.0	10.0	0.19/-0.07	0.16/-0.19	0.39/-0.02
CONCH	9.2	6.0	12.3	0.20/-0.20	0.15/-0.15	0.44/-0.04
Virchow2	9.2	8.3	10.0	0.20/-0.11	0.13/-0.17	0.35/-0.04

Bold marks the Pareto frontier — the encoder(s) no other pathology encoder beats on both rankings at once. The two ranks are kept apart on purpose: a strong median margin can hide a brittle tail, so each cohort cell shows both, median CRoMa/LTM₁₀. Orange rows mark encoders whose disclosed pretraining overlaps TCGA (legend). Full panel, cohort detail and the distribution explorer →

Median margin against tail severity, aggregated by rank across the three cohorts. Better is up and to the right; the ringed, named points are undominated. Hover or tab to any point to name it.

Citing¶

@article{grisi2026beyond,
  title   = {Beyond counts: A distributional robustness margin for pathology foundation models},
  author  = {Grisi, Cl{\'e}ment and van der Laak, Jeroen and Litjens, Geert},
  journal = {arXiv preprint arXiv:2607.25497},
  year    = {2026},
  doi     = {10.48550/arXiv.2607.25497},
  url     = {https://arxiv.org/abs/2607.25497}
}

Please cite the PathoROB study alongside it: RI is theirs, and the three tile cohorts scored under Results are theirs too.

croma¶

The panel at a glance¶

Getting started

Metrics

Manifest

Results

Citing¶