# Intermediate Layers Encode Optimal Biological Representations in Single-Cell Foundation Models

arXiv: https://arxiv.org/abs/2604.14838 · HTML: https://mlsanity.com/paper/2604.14838/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 2 of its 15 references that are in the map.
- First posted on arXiv 16 Apr 2026.

## Builds on (most cited first, 2 of 2)

- [Layer by Layer: Uncovering Hidden Representations in Language Models](https://mlsanity.com/paper/2502.02013/) (2025)
- [scDrugMap: Benchmarking Large Foundation Models for Drug Response Prediction](https://mlsanity.com/paper/2505.05612/) (2025)

## Abstract (arXiv)

Current single-cell foundation model benchmarks universally extract final layer embeddings, assuming these represent optimal feature spaces. We systematically evaluate layer-wise representations from scFoundation (100M parameters) and Tahoe-X1 (1.3B parameters) across trajectory inference and perturbation response prediction. Our analysis reveals that optimal layers are task-dependent (trajectory peaks at 60% depth, 31% above final layers) and context-dependent (perturbation optima shift 0-96% across T cell activation states). Notably, first-layer embeddings outperform all deeper layers in quiescent cells, challenging assumptions about hierarchical feature abstraction. These findings demonstrate that "where" to extract features matters as much as "what" the model learns, necessitating systematic layer evaluation tailored to biological task and cellular context rather than defaulting to final-layer embeddings.