# Dense SAE Latents Are Features, Not Bugs

arXiv: https://arxiv.org/abs/2506.15679 · HTML: https://mlsanity.com/paper/2506.15679/

## Key facts

- Cited by 17 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 30 of its 56 references that are in the map.
- First posted on arXiv 18 Jun 2025.

## Placed in

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

## Builds on (most cited first, 30 of 30)

- [PyTorch: An Imperative Style, High-Performance Deep Learning Library](https://mlsanity.com/paper/1912.01703/) (2019)
- [Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer](https://mlsanity.com/paper/1910.10683/) (2019)
- [Distributed Representations of Words and Phrases and their Compositionality](https://mlsanity.com/paper/1310.4546/) (2013)
- [Root Mean Square Layer Normalization](https://mlsanity.com/paper/1910.07467/) (2019)
- [Sparse Autoencoders Find Highly Interpretable Features in Language Models](https://mlsanity.com/paper/2309.08600/) (2023)
- [Man is to Computer Programmer as Woman is to Homemaker? Debiasing Word Embeddings](https://mlsanity.com/paper/1607.06520/) (2016)
- [Toy Models of Superposition](https://mlsanity.com/paper/2209.10652/) (2022)
- [The Linear Representation Hypothesis and the Geometry of Large Language Models](https://mlsanity.com/paper/2311.03658/) (2023)
- [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)
- [Emergent Linear Representations in World Models of Self-Supervised Sequence Models](https://mlsanity.com/paper/2309.00941/) (2023)
- [Jumping Ahead: Improving Reconstruction Fidelity with JumpReLU Sparse Autoencoders](https://mlsanity.com/paper/2407.14435/) (2024)
- [RedPajama: an Open Dataset for Training Large Language Models](https://mlsanity.com/paper/2411.12372/) (2024)
- [Linear Algebraic Structure of Word Senses, with Applications to Polysemy](https://mlsanity.com/paper/1601.03764/) (2016)
- [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)
- [Automatically Interpreting Millions of Features in Large Language Models](https://mlsanity.com/paper/2410.13928/) (2024)
- [Transformer visualization via dictionary learning: contextualized embedding as a linear superposition of transformer factors](https://mlsanity.com/paper/2103.15949/) (2021)
- [How do Language Models Bind Entities in Context?](https://mlsanity.com/paper/2310.17191/) (2023)
- [Universal Neurons in GPT2 Language Models](https://mlsanity.com/paper/2401.12181/) (2024)
- [Sparse Autoencoders Do Not Find Canonical Units of Analysis](https://mlsanity.com/paper/2502.04878/) (2025)
- [Confidence Regulation Neurons in Language Models](https://mlsanity.com/paper/2406.16254/) (2024)
- [Spectral Filters, Dark Signals, and Attention Sinks](https://mlsanity.com/paper/2402.09221/) (2024)
- [Fishing for Magikarp: Automatically Detecting Under-trained Tokens in Large Language Models](https://mlsanity.com/paper/2405.05417/) (2024)
- [Sparse Overcomplete Word Vector Representations](https://mlsanity.com/paper/1506.02004/) (2015)
- [Interpreting Attention Layer Outputs with Sparse Autoencoders](https://mlsanity.com/paper/2406.17759/) (2024)
- [Monitoring Latent World States in Language Models with Propositional Probes](https://mlsanity.com/paper/2406.19501/) (2024)
- [Interpretability in Parameter Space: Minimizing Mechanistic Description Length with Attribution-based Parameter Decomposition](https://mlsanity.com/paper/2501.14926/) (2025)
- [Word Embedding Visualization Via Dictionary Learning](https://mlsanity.com/paper/1910.03833/) (2019)

## Built on by (most cited first, 17 of 17)

- [Temporal Sparse Autoencoders: Leveraging the Sequential Nature of Language for Interpretability](https://mlsanity.com/paper/2511.05541/) (2025)
- [Transformers learn factored representations](https://mlsanity.com/paper/2602.02385/) (2026)
- [Finding Manifolds With Bilinear Autoencoders](https://mlsanity.com/paper/2510.16820/) (2025)
- [Structuring Sparsity: Block-Sparse Featurizers Capture Visual Concept Manifolds](https://mlsanity.com/paper/2606.25234/) (2026)
- [Evaluating Sparse Autoencoders for Monosemantic Representation](https://mlsanity.com/paper/2508.15094/) (2025)
- [The Confidence Manifold: Geometric Structure of Correctness Representations in Language Models](https://mlsanity.com/paper/2602.08159/) (2026)
- [Tree SAE: Learning Hierarchical Feature Structures in Sparse Autoencoders](https://mlsanity.com/paper/2605.07922/) (2026)
- [Legible-by-Construction: Attention and End-to-End Transformers](https://mlsanity.com/paper/2607.04319/) (2026)
- [Semantic Regexes: Auto-Interpreting LLM Features with a Structured Language](https://mlsanity.com/paper/2510.06378/) (2025)
- [On the Interpretability of Whisper Encodings Using Sparse Autoencoders](https://mlsanity.com/paper/2605.12225/) (2026)
- [What Does a Chemical Language Model Know About Molecules?](https://mlsanity.com/paper/2606.23443/) (2026)
- [Binary Autoencoder for Mechanistic Interpretability of Large Language Models](https://mlsanity.com/paper/2509.20997/) (2025)
- [WriteSAE: Sparse Autoencoders for Recurrent State](https://mlsanity.com/paper/2605.12770/) (2026)
- [Capability neq Interpretability: Human Interpretability of Vision Foundation Models](https://mlsanity.com/paper/2605.20337/) (2026)
- [How Optimality Structures Sparse Dictionaries: Theory for Interpreting SAE Representations](https://mlsanity.com/paper/2606.02385/) (2026)
- [Decompose Sparsely Where You Should, Absorb Densely Where You Should No](https://mlsanity.com/paper/2606.14040/) (2026)
- [MAxBench: A Multinomial Concept Recovery Benchmark](https://mlsanity.com/paper/2609.13072/) (2026)

## Abstract (arXiv)

Sparse autoencoders (SAEs) are designed to extract interpretable features from language models by enforcing a sparsity constraint. Ideally, training an SAE would yield latents that are both sparse and semantically meaningful. However, many SAE latents activate frequently (i.e., are dense), raising concerns that they may be undesirable artifacts of the training procedure. In this work, we systematically investigate the geometry, function, and origin of dense latents and show that they are not only persistent but often reflect meaningful model representations. We first demonstrate that dense latents tend to form antipodal pairs that reconstruct specific directions in the residual stream, and that ablating their subspace suppresses the emergence of new dense features in retrained SAEs -- suggesting that high density features are an intrinsic property of the residual space. We then introduce a taxonomy of dense latents, identifying classes tied to position tracking, context binding, entropy regulation, letter-specific output signals, part-of-speech, and principal component reconstruction. Finally, we analyze how these features evolve across layers, revealing a shift from structural features in early layers, to semantic features in mid layers, and finally to output-oriented signals in the last layers of the model. Our findings indicate that dense latents serve functional roles in language model computation and should not be dismissed as training noise.