AI Paper Atlas

Research map / Label-efficient and robust learning / Knowledge distillation

Pathryoshka: Compressing Pathology Foundation Models via Multi-Teacher Knowledge Distillation with Nested Embeddings

Christian Grashei, Christian Brechenmacher, Rao Umer, Jingsong Liu, Carsten Marr, Peter Schüffler, Ewa Szczurek

ECCV 2026 · Label-efficient and robust learning › Knowledge distillation · approach: Few-shot and metric distillation · 6 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add random non-aligned cropping augmentation (ablation)Pathryoshka-B · 10 pathology benchmarksaverage accuracy: 81.6 -> 82.4Improved
add random cropping augmentation (ablation)Pathryoshka-B · CAMELYON16accuracy: 91.6 -> 95.0Improved
add random cropping augmentation (ablation)Pathryoshka-B · MHISTaccuracy: 85.1 -> 84.0Hurt
add random cropping augmentation (ablation)Pathryoshka-B · PANDA / UniToPathoaccuracy: 77.0 -> 77.0; 54.9 -> 54.9No clear effect
distill from multiple teachers instead of one (ablation)Pathryoshka · 10 pathology benchmarksmedian balanced accuracyImproved
multi-teacher nested distillation vs single-teacher H0-miniPathryoshka-B · 10 pathology benchmarksmedian accuracy: 78.6 -> 85.6Improved
distill ViT-B student from three large teachers vs UNI2-hPathryoshka-B · 10 pathology benchmarksmedian accuracy: 84.3 -> 85.6Improved
reduce model from ViT-B to ViT-SPathryoshka · 10 pathology benchmarksmedian accuracy: 85.6 -> 80.5Hurt
add nested embeddings and truncate to dim/16Pathryoshka-B · ConSep segmentationDice: H0-mini 54.8 vs Pathryoshka-B 62.6Improved
truncate embeddings to dim/64 (nested vs non-nested)Pathryoshka-B · CCRCC patch retrievalRecall@5: UNI2-h 92.4 vs 95.2Improved
truncate embeddings to dim/64 (k-NN)Pathryoshka-B · 6 patch benchmarks k-NNrelative accuracy decrease: 9.4% vs 17.7% (UNI2-h)Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy