# Causal Interpretation of Sparse Autoencoder Features in Vision

arXiv: https://arxiv.org/abs/2509.00749 · HTML: https://mlsanity.com/paper/2509.00749/

## Key facts

- Cited by 5 papers in the map (data through 14 Oct 2026).
- Placed in Interpreting vision and multimodal models, one of 13 directions in Mechanistic interpretability.
- Builds on 11 of its 11 references that are in the map.
- First posted on arXiv 31 Aug 2025.

## Placed in

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

## Builds on (most cited first, 11 of 11)

- [A Unified Approach to Interpreting Model Predictions](https://mlsanity.com/paper/1705.07874/) (2017)
- [Axiomatic Attribution for Deep Networks](https://mlsanity.com/paper/1703.01365/) (2017)
- [Sparse Autoencoders Find Highly Interpretable Features in Language Models](https://mlsanity.com/paper/2309.08600/) (2023)
- [Reproducible scaling laws for contrastive language-image learning](https://mlsanity.com/paper/2212.07143/) (2022)
- [Evaluating the visualization of what a Deep Neural Network has learned](https://mlsanity.com/paper/1509.06321/) (2015)
- [AttnLRP: Attention-Aware Layer-Wise Relevance Propagation for Transformers](https://mlsanity.com/paper/2402.05602/) (2024)
- [Sparse Autoencoders Learn Monosemantic Features in Vision-Language Models](https://mlsanity.com/paper/2504.02821/) (2025)
- [Sparse autoencoders reveal selective remapping of visual concepts during adaptation](https://mlsanity.com/paper/2412.05276/) (2024)
- [Interpreting CLIP with Hierarchical Sparse Autoencoders](https://mlsanity.com/paper/2502.20578/) (2025)
- [Interpretable and Testable Vision Features via Sparse Autoencoders](https://mlsanity.com/paper/2502.06755/) (2025)
- [Respect the model: Fine-grained and Robust Explanation with Sharing Ratio Decomposition](https://mlsanity.com/paper/2402.03348/) (2024)

## Built on by (most cited first, 5 of 5)

- [Sparse Autoencoders enable Robust and Interpretable Fine-tuning of CLIP models](https://mlsanity.com/paper/2605.15961/) (2026)
- [Beyond Semantics: Disentangling Information Scope in Sparse Autoencoders for CLIP](https://mlsanity.com/paper/2604.05724/) (2026)
- [When Structured Sparse Autoencoders Learn Consistent Concepts Across Modalities](https://mlsanity.com/paper/2607.08605/) (2026)
- [Verifying the Linear Representation Hypothesis: How Interpretable Are Vision SAEs?](https://mlsanity.com/paper/2609.35020/) (2026)
- [Metonymic Circuits for Abstract Concept Grounding in Vision Transformers](https://mlsanity.com/paper/2610.06928/) (2026)

## Abstract (arXiv)

Understanding what sparse auto-encoder (SAE) features in vision transformers truly represent is usually done by inspecting the patches where a feature's activation is highest. However, self-attention mixes information across the entire image, so an activated patch often co-occurs with-but does not cause-the feature's firing. We propose Causal Feature Explanation (CaFE), which leverages Effective Receptive Field (ERF). We consider each activation of an SAE feature to be a target and apply input-attribution methods to identify the image patches that causally drive that activation. Across CLIP-ViT features, ERF maps frequently diverge from naive activation maps, revealing hidden context dependencies (e.g., a "roaring face" feature that requires the co-occurrence of eyes and nose, rather than merely an open mouth). Patch insertion tests confirm that CaFE more effectively recovers or suppresses feature activations than activation-ranked patches. Our results show that CaFE yields more faithful and semantically precise explanations of vision-SAE features, highlighting the risk of misinterpretation when relying solely on activation location.