# Feature Hedging: Correlated Features Break Narrow Sparse Autoencoders

arXiv: https://arxiv.org/abs/2505.11756 · HTML: https://mlsanity.com/paper/2505.11756/

## Key facts

- Cited by 16 papers in the map (data through 14 Oct 2026).
- Placed in Sparse autoencoder training and evaluation, one of 13 directions in Mechanistic interpretability.
- Builds on 16 of its 29 references that are in the map.
- First posted on arXiv 16 May 2025.

## Placed in

- Mechanistic interpretability / [Sparse autoencoder training and evaluation](https://mlsanity.com/mechinterp/sparse-autoencoder-training-and-evaluation/)

## Builds on (most cited first, 16 of 16)

- [The Pile: An 800GB Dataset of Diverse Text for Language Modeling](https://mlsanity.com/paper/2101.00027/) (2020)
- [Sparse Autoencoders Find Highly Interpretable Features in Language Models](https://mlsanity.com/paper/2309.08600/) (2023)
- [Toy Models of Superposition](https://mlsanity.com/paper/2209.10652/) (2022)
- [Scaling and evaluating sparse autoencoders](https://mlsanity.com/paper/2406.04093/) (2024)
- [Sparse Feature Circuits: Discovering and Editing Interpretable Causal Graphs in Language Models](https://mlsanity.com/paper/2403.19647/) (2024)
- [Gemma Scope: Open Sparse Autoencoders Everywhere All At Once on Gemma 2](https://mlsanity.com/paper/2408.05147/) (2024)
- [Finding Neurons in a Haystack: Case Studies with Sparse Probing](https://mlsanity.com/paper/2305.01610/) (2023)
- [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)
- [Not All Language Model Features Are One-Dimensionally Linear](https://mlsanity.com/paper/2405.14860/) (2024)
- [A is for Absorption: Studying Feature Splitting and Absorption in Sparse Autoencoders](https://mlsanity.com/paper/2409.14507/) (2024)
- [Learning Multi-Level Features with Matryoshka Sparse Autoencoders](https://mlsanity.com/paper/2503.17547/) (2025)
- [BatchTopK Sparse Autoencoders](https://mlsanity.com/paper/2412.06410/) (2024)
- [SAEBench: A Comprehensive Benchmark for Sparse Autoencoders in Language Model Interpretability](https://mlsanity.com/paper/2503.09532/) (2025)
- [Are Sparse Autoencoders Useful? A Case Study in Sparse Probing](https://mlsanity.com/paper/2502.16681/) (2025)
- [Decomposing The Dark Matter of Sparse Autoencoders](https://mlsanity.com/paper/2410.14670/) (2024)

## Built on by (most cited first, 16 of 16)

- [Sparse but Wrong: Incorrect L0 Leads to Incorrect Features in Sparse Autoencoders](https://mlsanity.com/paper/2508.16560/) (2025)
- [SynthSAEBench: Evaluating Sparse Autoencoders on Scalable Realistic Synthetic Data](https://mlsanity.com/paper/2602.14687/) (2026)
- [OrtSAE: Orthogonal Sparse Autoencoders Uncover Atomic Features](https://mlsanity.com/paper/2509.22033/) (2025)
- [Causality is Key for Interpretability Claims to Generalise](https://mlsanity.com/paper/2602.16698/) (2026)
- [From Atoms to Trees: Building a Structured Feature Forest with Hierarchical Sparse Autoencoders](https://mlsanity.com/paper/2602.11881/) (2026)
- [Stop Probing, Start Coding: Why Linear Probes and Sparse Autoencoders Fail at Compositional Generalisation](https://mlsanity.com/paper/2603.28744/) (2026)
- [From Reasoning Traces to Reusable Modules: Understanding Compositional Generalization in Language Model Reasoning](https://mlsanity.com/paper/2606.18089/) (2026)
- [AlignSAE: Concept-Aligned Sparse Autoencoders](https://mlsanity.com/paper/2512.02004/) (2025)
- [The Rate-Distortion-Polysemanticity Tradeoff in SAEs](https://mlsanity.com/paper/2605.14694/) (2026)
- [Do Sparse Autoencoders Learn Meaningful Concept Hierarchies?](https://mlsanity.com/paper/2606.22994/) (2026)
- [Towards Atoms of Large Language Models](https://mlsanity.com/paper/2509.20784/) (2025)
- [Sign-Aware Gated Sparse Autoencoders: Modeling Anticorrelated Features with Bi-Jump-ReLU Activations](https://mlsanity.com/paper/2605.28149/) (2026)
- [Beyond a Bag of Features: Set-Level Instability in Sparse Autoencoders](https://mlsanity.com/paper/2608.11197/) (2026)
- [HiPACE: Hierarchical Phase-Boundary Analysis and Controlled Evaluation of Feature Absorption in Sparse Autoencoders](https://mlsanity.com/paper/2609.29551/) (2026)
- [When Trees Are Not Enough: Learning Mixed-Topology Feature Graphs with Adaptive Graph Sparse Autoencoders](https://mlsanity.com/paper/2609.36294/) (2026)
- [The Independence Prior of SAEs Fragments Visual Concepts](https://mlsanity.com/paper/2610.04112/) (2026)

## Abstract (arXiv)

It is assumed that sparse autoencoders (SAEs) decompose polysemantic activations into interpretable linear directions, as long as the activations are composed of sparse linear combinations of underlying features. However, we find that if an SAE is more narrow than the number of underlying "true features" on which it is trained, and there is correlation between features, the SAE will merge components of correlated features together, thus destroying monosemanticity. In LLM SAEs, these two conditions are almost certainly true. This phenomenon, which we call feature hedging, is caused by SAE reconstruction loss, and is more severe the narrower the SAE. In this work, we introduce the problem of feature hedging and study it both theoretically in toy models and empirically in SAEs trained on LLMs. We suspect that feature hedging may be one of the core reasons that SAEs consistently underperform supervised baselines. Finally, we use our understanding of feature hedging to propose an improved variant of matryoshka SAEs. Importantly, our work shows that SAE width is not a neutral hyperparameter: narrower SAEs suffer more from hedging than wider SAEs.