Benchmarking Foundation Models for Mitotic Figure Classification DOI: 10.59275/j.melba.2024-AAAA Volume: 2
\firstname Jonas \surname Ammeling \aff 1 \orcid 0000-0002-0335-1194, \name Jonathan Ganz \aff 2, 1 \orcid 0009-0008-1299-8716, \name Emely Rosbach \aff 1 \orcid 0009-0001-7526-9923, \name Ludwig Lausser \aff 1, \name Christof A. Bertram \aff 3 \orcid 0000-0002-2402-9997, \name Katharina Breininger · 2025
Read Benchmarking Foundation Models for Mitotic Figure Classification DOI: 10.59275/j.melba.2024-AAAA Volume: 2 by \firstname Jonas \surname Ammeling \aff 1 \orcid 0000-0002-0335-1194, \name Jonathan Ganz \aff 2, 1 \orcid 0009-0008-1299-8716, \name Emely Rosbach \aff 1 \orcid 0009-0001-7526-9923, \name Ludwig Lausser \aff 1, \name Christof A. Bertram \aff 3 \orcid 0000-0002-2402-9997, \name Katharina Breininger (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2508-04441.txt.
Description
Jonas Ammeling and co-authors benchmark foundation models at spotting dividing cells in tumor slides. Mitotic counts help grade tumors, but labeled pathology images are scarce. Foundation models trained on unlabeled data supply rich features for small tasks. The team tested how those features scale and travel across tumor types.
Simple linear probes on frozen features set the baseline. Low-rank tuning of attention, called LoRA, did clearly better. With only a tenth of the training data, LoRA models near full-data skill. On unseen tumor domains, the newest models nearly closed the gap. Fully tuned classic networks stayed competitive, though, keeping the race honest.
Catalog record
| CREATOR | \firstname Jonas \surname Ammeling \aff 1 \orcid 0000-0002-0335-1194, \name Jonathan Ganz \aff 2, 1 \orcid 0009-0008-1299-8716, \name Emely Rosbach \aff 1 \orcid 0009-0001-7526-9923, \name Ludwig Lausser \aff 1, \name Christof A. Bertram \aff 3 \orcid 0000-0002-2402-9997, \name Katharina Breininger |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 49 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2508-04441