# Sparse Autoencoders Bridge The Deep Learning Model and The Brain

arXiv: https://arxiv.org/abs/2506.11123 · HTML: https://mlsanity.com/paper/2506.11123/

## Key facts

- Cited by 3 papers in the map (data through 14 Oct 2026).
- Placed in Interpreting vision and multimodal models, one of 13 directions in Mechanistic interpretability.
- First posted on arXiv 10 Jun 2025.

## Placed in

- Mechanistic interpretability / [Interpreting vision and multimodal models](https://mlsanity.com/mechinterp/interpreting-vision-and-multimodal-models/)

## Built on by (most cited first, 3 of 3)

- [A Unified Theory of Sparse Dictionary Learning in Mechanistic Interpretability: Piecewise Biconvexity and Spurious Minima](https://mlsanity.com/paper/2512.05534/) (2025)
- [CXR-LanIC: Language-Grounded Interpretable Classifier for Chest X-Ray Diagnosis](https://mlsanity.com/paper/2510.21464/) (2025)
- [Interpreting Brain Responses to Language with Sparse Features from Language Models](https://mlsanity.com/paper/2606.06857/) (2026)

## Abstract (arXiv)

We present SAE-BrainMap, a novel framework that directly aligns deep learning visual model representations with voxel-level fMRI responses using sparse autoencoders (SAEs). First, we train layer-wise SAEs on model activations and compute the correlations between SAE unit activations and cortical fMRI signals elicited by the same natural image stimuli with cosine similarity, revealing strong activation correspondence (maximum similarity up to 0.76). Depending on this alignment, we construct a voxel dictionary by optimally assigning the most similar SAE feature to each voxel, demonstrating that SAE units preserve the functional structure of predefined regions of interest (ROIs) and exhibit ROI-consistent selectivity. Finally, we establish fine-grained hierarchical mapping between model layers and the human ventral visual pathway, also by projecting voxel dictionary activations onto individual cortical surfaces, we visualize the dynamic transformation of the visual information in deep learning models. It is found that ViT-B/16CLIP tends to utilize low-level information to generate high-level semantic information in the early layers and reconstructs the low-dimension information later. Our results establish a direct, downstream-task-free bridge between deep neural networks and human visual cortex, offering new insights into model interpretability.