# Interpreting and steering diffusion models

Area: [Mechanistic interpretability](https://mlsanity.com/mechinterp/index.md). HTML: https://mlsanity.com/mechinterp/interpreting-and-steering-diffusion-models/

## Key facts

- Interpreting and steering diffusion models: 27 papers on arXiv since 2022 (data through 14 Oct 2026).
- 11 of its 27 mapped papers (41%) cite “SAeUron: Interpretable Concept Unlearning in Diffusion Models with Sparse Autoencoders” (2025, arXiv 2501.18052).
- 7 of its 27 mapped papers (26%) cite “One-Step is Enough: Sparse Autoencoders for Text-to-Image Diffusion Models” (2024, arXiv 2410.22366).
- 4 new papers 24 Sep–14 Oct.

## Most cited here

- [SAeUron: Interpretable Concept Unlearning in Diffusion Models with Sparse Autoencoders](https://mlsanity.com/paper/2501.18052/) (2025), cited by 11 papers here
- [One-Step is Enough: Sparse Autoencoders for Text-to-Image Diffusion Models](https://mlsanity.com/paper/2410.22366/) (2024), cited by 7 papers here
- [textitRevelio: Interpreting and leveraging semantic information in diffusion models](https://mlsanity.com/paper/2411.16725/) (2024), cited by 5 papers here
- [TIDE : Temporal-Aware Sparse Autoencoders for Interpretable Diffusion Transformers in Image Generation](https://mlsanity.com/paper/2503.07050/) (2025), cited by 4 papers here
- [Concept Steerers: Leveraging K-Sparse Autoencoders for Test-Time Controllable Generations](https://mlsanity.com/paper/2501.19066/) (2025), cited by 3 papers here
- [Interpreting Large Text-to-Image Diffusion Models with Dictionary Learning](https://mlsanity.com/paper/2505.24360/) (2025), cited by 2 papers here
- [SAEmnesia: Erasing Concepts in Diffusion Models with Supervised Sparse Autoencoders](https://mlsanity.com/paper/2509.21379/) (2025), cited by 2 papers here
- [A Single Neuron Works: Precise Concept Erasure in Text-to-Image Diffusion Models](https://mlsanity.com/paper/2509.21008/) (2025), cited by 2 papers here
- [Emergence and Evolution of Interpretable Concepts in Diffusion Models](https://mlsanity.com/paper/2504.15473/) (2025), cited by 1 papers here
- [OrthoEraser: Coupled-Neuron Orthogonal Projection for Concept Erasure](https://mlsanity.com/paper/2603.11493/) (2026), cited by 1 papers here

## New (24 Sep–14 Oct)

- [Beyond the Linear Representation Hypothesis: Non-Linear Activation Steering in Text-to-Image Models](https://mlsanity.com/paper/2610.06945/) (2026-10-03)
- [RASteer: Retain-Aware Activation Steering for Concept Erasure in Diffusion Models](https://mlsanity.com/paper/2610.01969/) (2026-10-01)
- [D-Scope: Decomposing and Steering Diffusion Transformers with Sparse Autoencoders](https://mlsanity.com/paper/2609.39625/) (2026-09-30)
- [SharedSAE: a unified sparse dictionary across u-net blocks and denoising time for diffusion models](https://mlsanity.com/paper/2610.13997/) (2026-09-27)