THUNDER: Tile-level Histopathology image UNDERstanding benchmark
Pierre Marza, Leo Fillioux, Sofiène Boutaj, Kunal Mahatha, Christian Desrosiers, Pablo Piantanida, Jose Dolz, Stergios Christodoulidis,, Maria Vakalopoulou · 2025
Read THUNDER: Tile-level Histopathology image UNDERstanding benchmark by Pierre Marza, Leo Fillioux, Sofiène Boutaj, Kunal Mahatha, Christian Desrosiers, Pablo Piantanida, Jose Dolz, Stergios Christodoulidis,, Maria Vakalopoulou (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-2507-07860.txt.
Introduction
Pierre Marza and colleagues build THUNDER, a fast benchmark for the flood of foundation models in digital pathology. New feature extractors arrive monthly, and downstream performance alone cannot distinguish them responsibly. In healthcare, robustness and uncertainty matter alongside accuracy. THUNDER compares models at tile level across diverse data sets and tasks.
It also probes the geometry of their feature spaces and stress-tests prediction reliability. User-defined local models can join the comparison directly. The paper reports a comprehensive run: twenty-three foundation models across sixteen data sets, with feature analysis and robustness checks. The code is public and the benchmark designed to stay dynamic as models multiply. For a field drowning in releases, THUNDER offers a shared yardstick with the safety dimensions left in.
Catalog record
| CREATOR | Pierre Marza, Leo Fillioux, Sofiène Boutaj, Kunal Mahatha, Christian Desrosiers, Pablo Piantanida, Jose Dolz, Stergios Christodoulidis,, Maria Vakalopoulou |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 249 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2507-07860