PCaBiop (slide-level)

1,000 prostate biopsy slides sourced from PANDA, labelled benign or cancer, contributed by two data providers (Karolinska Institutet and Radboud UMC). Unlike the three tile cohorts, the evaluation unit is a whole slide: each of the 5 encoders is a whole-slide foundation model producing one embedding per slide, and every number below is computed over those slide embeddings.

This is a different panel. The slide roster shares no encoder ranking with the 26-model tile panel, so this cohort takes no part in the cross-cohort aggregate and its ranks — the table below orders these five encoders against each other, nothing more. There is also no natural-image control at slide level, so no row carries the † mark.

Don't see a slide encoder? Request an evaluation

PCaBiop, sorted by median CRoMa. Columns are explained under Reading the columns.

Model	bio bacc	conf bacc	RI	MaRI	CRoMa	F(0)	LTM₁₀	support
PRISM2	0.975	0.995	0.434	0.367	0.27	0.187	-0.24	20.9%
PRISM	0.968	0.992	0.281	0.195	0.26	0.288	-0.39	10.1%
MOOZY	0.971	1.000	0.236	0.181	-0.02	0.535	-0.41	24.3%
TITAN	0.915	1.000	0.015	0.001	-0.30	0.895	-0.60	49.1%
Prov-GigaPath	0.720	0.995	0.020	0.006	-0.41	0.990	-0.59	97.7%

The confounder is unusually dominant here: every encoder’s confounder k-NN accuracy is at or near 1.0 — the data provider is essentially perfectly decodable from every slide embedding — and 3 of the five encoders hold a negative median margin. Support spans 10% to 98%, the widest spread of any published cohort, so the count-based indices rest on very different amounts of evidence per row: Prov-GigaPath’s low RI is measured on nearly every slide, while PRISM’s rests on a tenth of them.

The two rankings, on this cohort alone:

Median CRoMa against tail severity LTM₁₀. Better is up and to the right; ringed points are undominated on both axes. Hover or tab to any point to name it with its two values. With five encoders the panel is sparse by construction — read it as a picture of the table, not as a frontier claim of tile-panel strength.

The distribution explorer

The same explorer as the tile panel’s, restricted to the slide roster. Click a row to move the detail, drag across the detail curve to count the slides in any range, and pick a second encoder under Compare with to overlay its shape.

The shape is again where the medians stop telling the story. PRISM2 and PRISM sit 0.014 apart on median CRoMa — indistinguishable on that column — while PRISM carries 1.5× the confounder-dominant mass and 1.7× the tail severity. Overlay the two above to see it.

The operating point

This cohort is reported at k*, not the tile panel’s shared median-k: each encoder is evaluated at its own kNN-optimal k (here spanning k = 3–13; the per-encoder values are recorded in results/PROVENANCE.json). With only five encoders, a shared median is dominated by panel composition — adding a single encoder to the panel moved the would-be shared k from 3 to 9 — so pinning one k would make every number hostage to who else happens to be on the roster. The trade is stated rather than hidden: RI and MaRI are protocol-dependent, so those two columns are not measured at one shared operating point here, while CRoMa, F(0) and LTM₁₀ are k-free and unaffected.

Shortcut susceptibility

Shortcut susceptibility for the five whole-slide encoders, in domain (ID) and out of domain (OOD). Rows are ranked by Change at V = 1, the normalized change at maximum confounding, because nIPD averages over the whole range: an early gain there can pay for a late collapse, so a curve ending at chance can outrank one that never moved. Rows ending at or below -0.900 are marked ≈ chance, where none of the above-chance margin survives.

Bold marks the leading ranked encoder in each column where higher is better, so a column that disagrees with the ranking shows it at a glance. Each caption reports Spearman ρ, the rank correlation between the CRoMa and nIPD columns: how closely the two order the encoders the same way. Shortcut susceptibility defines the measure and holds the interactive explorer.

PCaBiop — ID; Spearman ρ = 0.90; n=5 ranked pathology encoders; descriptive

Model	Median CRoMa (m=5)	Change at V = 1	nIPD	Baseline balanced accuracy
PRISM2	0.270	-0.216	-0.032	0.957
PRISM	0.257	-0.293	-0.078	0.953
MOOZY	-0.016	-0.313	-0.065	0.967
TITAN	-0.295	-0.803	-0.189	0.912
Prov-GigaPath	-0.413	-0.970 ≈ chance	-0.245	0.803

PCaBiop — OOD; Spearman ρ = 0.60; n=5 ranked pathology encoders; descriptive

Model	Median CRoMa (m=5)	Change at V = 1	nIPD	Baseline balanced accuracy
PRISM2	0.270	-0.040	-0.011	0.945
PRISM	0.257	-0.304	-0.083	0.917
MOOZY	-0.016	-0.759	0.048	0.790
TITAN	-0.295	-0.931 ≈ chance	-0.126	0.855
Prov-GigaPath	-0.413	-1.161 ≈ chance	-0.107	0.548

Read MOOZY’s positive OOD value with care. Its pretraining corpus includes the PANDA slides every probe trains on, so the apparent improvement under confounding plausibly reflects familiarity with the training data rather than robustness on the held-out PAR cohort, which no encoder has seen.