# Interpreting CLIP with Hierarchical Sparse Autoencoders

arXiv: https://arxiv.org/abs/2502.20578 · HTML: https://mlsanity.com/paper/2502.20578/

## Key facts

- Cited by 60 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 57 of its 74 references that are in the map.
- First posted on arXiv 27 Feb 2025.

## Placed in

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

## Builds on (most cited first, 30 of 57)

- [Learning Transferable Visual Models From Natural Language Supervision](https://mlsanity.com/paper/2103.00020/) (2021)
- [ImageNet Large Scale Visual Recognition Challenge](https://mlsanity.com/paper/1409.0575/) (2014)
- [Visual Instruction Tuning](https://mlsanity.com/paper/2304.08485/) (2023)
- [Grad-CAM: Visual Explanations from Deep Networks via Gradient-based Localization](https://mlsanity.com/paper/1610.02391/) (2016)
- [A Unified Approach to Interpreting Model Predictions](https://mlsanity.com/paper/1705.07874/) (2017)
- ["Why Should I Trust You?": Explaining the Predictions of Any Classifier](https://mlsanity.com/paper/1602.04938/) (2016)
- [SDXL: Improving Latent Diffusion Models for High-Resolution Image Synthesis](https://mlsanity.com/paper/2307.01952/) (2023)
- [Deep Learning Face Attributes in the Wild](https://mlsanity.com/paper/1411.7766/) (2014)
- [Axiomatic Attribution for Deep Networks](https://mlsanity.com/paper/1703.01365/) (2017)
- [Sigmoid Loss for Language Image Pre-Training](https://mlsanity.com/paper/2303.15343/) (2023)
- [Deep Inside Convolutional Networks: Visualising Image Classification Models and Saliency Maps](https://mlsanity.com/paper/1312.6034/) (2013)
- [Visualizing and Understanding Convolutional Networks](https://mlsanity.com/paper/1311.2901/) (2013)
- [Gemma 2: Improving Open Language Models at a Practical Size](https://mlsanity.com/paper/2408.00118/) (2024)
- [Similarity of Neural Network Representations Revisited](https://mlsanity.com/paper/1905.00414/) (2019)
- [LAION-400M: Open Dataset of CLIP-Filtered 400 Million Image-Text Pairs](https://mlsanity.com/paper/2111.02114/) (2021)
- [Sparse Autoencoders Find Highly Interpretable Features in Language Models](https://mlsanity.com/paper/2309.08600/) (2023)
- [Interpretability Beyond Feature Attribution: Quantitative Testing with Concept Activation Vectors (TCAV)](https://mlsanity.com/paper/1711.11279/) (2017)
- [Reproducible scaling laws for contrastive language-image learning](https://mlsanity.com/paper/2212.07143/) (2022)
- [Sanity Checks for Saliency Maps](https://mlsanity.com/paper/1810.03292/) (2018)
- [Concept Bottleneck Models](https://mlsanity.com/paper/2007.04612/) (2020)
- [Quantifying Attention Flow in Transformers](https://mlsanity.com/paper/2005.00928/) (2020)
- [Towards Automated Circuit Discovery for Mechanistic Interpretability](https://mlsanity.com/paper/2304.14997/) (2023)
- [CogVLM: Visual Expert for Pretrained Language Models](https://mlsanity.com/paper/2311.03079/) (2023)
- [Representation Alignment for Generation: Training Diffusion Transformers Is Easier Than You Think](https://mlsanity.com/paper/2410.06940/) (2024)
- [Scaling and evaluating sparse autoencoders](https://mlsanity.com/paper/2406.04093/) (2024)
- [Mechanistic Interpretability for AI Safety -- A Review](https://mlsanity.com/paper/2404.14082/) (2024)
- [Gemma Scope: Open Sparse Autoencoders Everywhere All At Once on Gemma 2](https://mlsanity.com/paper/2408.05147/) (2024)
- [The Platonic Representation Hypothesis](https://mlsanity.com/paper/2405.07987/) (2024)
- [How Much Can CLIP Benefit Vision-and-Language Tasks?](https://mlsanity.com/paper/2107.06383/) (2021)
- [Jumping Ahead: Improving Reconstruction Fidelity with JumpReLU Sparse Autoencoders](https://mlsanity.com/paper/2407.14435/) (2024)

## Built on by (most cited first, 30 of 60)

- [From Flat to Hierarchical: Extracting Sparse Representations with Matching Pursuit](https://mlsanity.com/paper/2506.03093/) (2025)
- [Not All Features Are Created Equal: A Mechanistic Study of Vision-Language-Action Models](https://mlsanity.com/paper/2603.19233/) (2026)
- [DNA: Uncovering Universal Latent Forgery Knowledge](https://mlsanity.com/paper/2601.22515/) (2026)
- [Measuring and Guiding Monosemanticity](https://mlsanity.com/paper/2506.19382/) (2025)
- [Explaining Similarity in Vision-Language Encoders with Weighted Banzhaf Interactions](https://mlsanity.com/paper/2508.05430/) (2025)
- [Emergent World Representations in OpenVLA](https://mlsanity.com/paper/2509.24559/) (2025)
- [Interpretable and Steerable Concept Bottleneck Sparse Autoencoders](https://mlsanity.com/paper/2512.10805/) (2025)
- [Causal Interpretation of Sparse Autoencoder Features in Vision](https://mlsanity.com/paper/2509.00749/) (2025)
- [MoME: Mixture of Matryoshka Experts for Audio-Visual Speech Recognition](https://mlsanity.com/paper/2510.04136/) (2025)
- [Model Science: getting serious about verification, explanation and control of AI systems](https://mlsanity.com/paper/2508.20040/) (2025)
- [Representations of Text and Images Align From Layer One](https://mlsanity.com/paper/2601.08017/) (2026)
- [Latent Anomaly Knowledge Excavation: Unveiling Sparse Sensitive Neurons in Vision-Language Models](https://mlsanity.com/paper/2604.07802/) (2026)
- [Inside the Visual Mind: Neuroscience-Motivated Concept Circuits for Interpreting and Steering Vision Transformers](https://mlsanity.com/paper/2606.06664/) (2026)
- [Semantic Regexes: Auto-Interpreting LLM Features with a Structured Language](https://mlsanity.com/paper/2510.06378/) (2025)
- [MedSAE: Dissecting MedCLIP Representations with Sparse Autoencoders](https://mlsanity.com/paper/2510.26411/) (2025)
- [Decomposing multimodal embedding spaces with group-sparse autoencoders](https://mlsanity.com/paper/2601.20028/) (2026)
- [Sparse CLIP: Co-Optimizing Interpretability and Performance in Contrastive Learning](https://mlsanity.com/paper/2601.20075/) (2026)
- [Evaluating the Interpretability of Sparse Autoencoders with Concept Annotations](https://mlsanity.com/paper/2606.24716/) (2026)
- [FaCT: Faithful Concept Traces for Explaining Neural Network Decisions](https://mlsanity.com/paper/2510.25512/) (2025)
- [Extracting Interaction-Aware Monosemantic Concepts in Recommender Systems](https://mlsanity.com/paper/2511.18024/) (2025)
- [Inspiration Seeds: Learning Non-Literal Visual Combinations for Generative Exploration](https://mlsanity.com/paper/2602.08615/) (2026)
- [Hierarchical Concept Embedding & Pursuit for Interpretable Image Classification](https://mlsanity.com/paper/2602.11448/) (2026)
- [SEM: Sparse Embedding Modulation for Post-Hoc Debiasing of Vision-Language Models](https://mlsanity.com/paper/2603.19028/) (2026)
- [Beyond Semantics: Disentangling Information Scope in Sparse Autoencoders for CLIP](https://mlsanity.com/paper/2604.05724/) (2026)
- [From Local to Global to Mechanistic: An iERF-Centered Unified Framework for Interpreting Vision Models](https://mlsanity.com/paper/2605.00474/) (2026)
- [SoftSAE: Dynamic Top-K Selection for Adaptive Sparse Autoencoders](https://mlsanity.com/paper/2605.06610/) (2026)
- [TEVI: Text-Conditioned Editing of Visual Representations via Sparse Autoencoders for Improved Vision-Language Alignment](https://mlsanity.com/paper/2606.07451/) (2026)
- [Cascaded Sparse Autoencoders Learn Multi-Level Visual Concepts in Multimodal LLMs](https://mlsanity.com/paper/2606.16193/) (2026)
- [Do Sparse Autoencoders Learn Meaningful Concept Hierarchies?](https://mlsanity.com/paper/2606.22994/) (2026)
- [When Structured Sparse Autoencoders Learn Consistent Concepts Across Modalities](https://mlsanity.com/paper/2607.08605/) (2026)

## Abstract (arXiv)

Sparse autoencoders (SAEs) are useful for detecting and steering interpretable features in neural networks, with particular potential for understanding complex multimodal representations. Given their ability to uncover interpretable features, SAEs are particularly valuable for analyzing large-scale vision-language models (e.g., CLIP and SigLIP), which are fundamental building blocks in modern systems yet remain challenging to interpret and control. However, current SAE methods are limited by optimizing both reconstruction quality and sparsity simultaneously, as they rely on either activation suppression or rigid sparsity constraints. To this end, we introduce Matryoshka SAE (MSAE), a new architecture that learns hierarchical representations at multiple granularities simultaneously, enabling a direct optimization of both metrics without compromise. MSAE establishes a new state-of-the-art Pareto frontier between reconstruction quality and sparsity for CLIP, achieving 0.99 cosine similarity and less than 0.1 fraction of variance unexplained while maintaining ~80% sparsity. Finally, we demonstrate the utility of MSAE as a tool for interpreting and controlling CLIP by extracting over 120 semantic concepts from its representation to perform concept-based similarity search and bias analysis in downstream tasks like CelebA. We make the codebase available at https://github.com/WolodjaZ/MSAE.