# Interpreting Brain Responses to Language with Sparse Features from Language Models

arXiv: https://arxiv.org/abs/2606.06857 · HTML: https://mlsanity.com/paper/2606.06857/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Placed in SAE features and their applications, one of 13 directions in Mechanistic interpretability.
- Builds on 17 of its 70 references that are in the map.
- First posted on arXiv 5 Jun 2026.

## Placed in

- Mechanistic interpretability / [SAE features and their applications](https://mlsanity.com/mechinterp/sae-features-and-their-applications/)

## Builds on (most cited first, 17 of 17)

- [Gemma 2: Improving Open Language Models at a Practical Size](https://mlsanity.com/paper/2408.00118/) (2024)
- [Gemma Scope: Open Sparse Autoencoders Everywhere All At Once on Gemma 2](https://mlsanity.com/paper/2408.05147/) (2024)
- [Jumping Ahead: Improving Reconstruction Fidelity with JumpReLU Sparse Autoencoders](https://mlsanity.com/paper/2407.14435/) (2024)
- [AxBench: Steering LLMs? Even Simple Baselines Outperform Sparse Autoencoders](https://mlsanity.com/paper/2501.17148/) (2025)
- [Causal Abstraction: A Theoretical Foundation for Mechanistic Interpretability](https://mlsanity.com/paper/2301.04709/) (2023)
- [CRAFT: Concept Recursive Activation FacTorization for Explainability](https://mlsanity.com/paper/2211.10154/) (2022)
- [Learning Multi-Level Features with Matryoshka Sparse Autoencoders](https://mlsanity.com/paper/2503.17547/) (2025)
- [On the Predictive Power of Neural Language Models for Human Real-Time Comprehension Behavior](https://mlsanity.com/paper/2006.01912/) (2020)
- [Low-Dimensional Structure in the Space of Language Representations is Reflected in Brain Responses](https://mlsanity.com/paper/2106.05426/) (2021)
- [From Language to Cognition: How LLMs Outgrow the Human Language Network](https://mlsanity.com/paper/2503.01830/) (2025)
- [Language models and brains align due to more than next-word prediction and word-level information](https://mlsanity.com/paper/2212.00596/) (2022)
- [Information-Restricted Neural Language Models Reveal Different Brain Regions' Sensitivity to Semantics, Syntax and Context](https://mlsanity.com/paper/2302.14389/) (2023)
- [From Atoms to Trees: Building a Structured Feature Forest with Hierarchical Sparse Autoencoders](https://mlsanity.com/paper/2602.11881/) (2026)
- [Generative causal testing to bridge data-driven models and scientific theories in language neuroscience](https://mlsanity.com/paper/2410.00812/) (2024)
- [Superposition disentanglement of neural representations reveals hidden alignment](https://mlsanity.com/paper/2510.03186/) (2025)
- [Sparse Autoencoders Bridge The Deep Learning Model and The Brain](https://mlsanity.com/paper/2506.11123/) (2025)
- [Sparse Autoencoders Map Brain-LLM Alignment onto Cortical Semantic Topography](https://mlsanity.com/paper/2605.23035/) (2026)

## Built on by (most cited first, 1 of 1)

- [The Platonic brain bridge hypothesis: human brain networks as an architectural prior for multimodal large language models](https://mlsanity.com/paper/2609.10947/) (2026)

## Abstract (arXiv)

A central goal of cognitive neuroscience is to characterize the features that are represented by human language cortex. Artificial language models (LMs) have emerged as a powerful tool to address this challenge, but studies relating biological and artificial representations are often criticized as relating one black box to another. The present work introduces Augmented Sparse Encoding Models, an encoding framework that replaces dense LM hidden states with hierarchically-organized sparse autoencoder (SAE) features, while explicitly including surprisal as a predictor. Using this approach, we (i) produce interpretations of neural responses and (ii) test whether model-brain alignment reflects primary or idiosyncratic variation in LM representations. Using a high-field 7T fMRI dataset of eight participants listening to 200 linguistically diverse sentences, we first validate our modeling framework by recovering previous interpretations of voxel populations tuned to processing difficulty and meaning abstractness. We then interpret a previously-uncharacterized (but reliable) voxel population and find that it is tuned to people-related content. Next, we show that the fronto-temporal human language network is predicted by a common set of features across its constituent regions, but find that frontal regions are relatively well-explained by surprisal alone, even in the absence of LM-based features. Finally, we show that brain responses during language processing are not merely predictable from an arbitrary set of LM features. Rather, brain responses are best explained by the features that tend to capture the most general information encoded in LM representations, suggesting a nontrivial correspondence between brain and LM language representation.