Multi-Dimensional Spectral Geometry of Biological Knowledge in Single-Cell Transformer Representations
Ihor Kendiukhov · 2026
Read Multi-Dimensional Spectral Geometry of Biological Knowledge in Single-Cell Transformer Representations by Ihor Kendiukhov (2026) 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-2602-22247.txt.
Description
Single-cell foundation models like scGPT learn rich gene vectors, but what biology those vectors encode has been murky. Ihor Kendiukhov decodes scGPT's inner geometry through dozens of rounds of automated hypothesis screening, testing nearly two hundred ideas. The model, he finds, sorts genes into a structured biological map rather than an opaque feature space.
The main axis separates genes by cell location, from secreted proteins at one pole to cytosolic proteins at the other. Middle layers briefly encode mitochondrial and endoplasmic-reticulum compartments in an order mirroring the secretory pathway. Other axes encode protein-interaction networks with graded fidelity. The study thus reads the model's geometry as biology, not black-box magic.
Catalog record
| CREATOR | Ihor Kendiukhov |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 54 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602-22247