# Interpreting vision and multimodal models

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

## Key facts

- Interpreting vision and multimodal models: 105 papers on arXiv since 2021 (data through 14 Oct 2026).
- 30 of its 105 mapped papers (29%) cite “Sparse autoencoders reveal selective remapping of visual concepts during adaptation” (2024, arXiv 2412.05276).
- 20 of its 105 mapped papers (19%) cite “Sparse Autoencoders Learn Monosemantic Features in Vision-Language Models” (2025, arXiv 2504.02821).
- 4 new papers 24 Sep–14 Oct.

## Most cited here

- [Sparse autoencoders reveal selective remapping of visual concepts during adaptation](https://mlsanity.com/paper/2412.05276/) (2024), cited by 30 papers here
- [Sparse Autoencoders Learn Monosemantic Features in Vision-Language Models](https://mlsanity.com/paper/2504.02821/) (2025), cited by 20 papers here
- [Interpreting CLIP with Hierarchical Sparse Autoencoders](https://mlsanity.com/paper/2502.20578/) (2025), cited by 16 papers here
- [Universal Sparse Autoencoders: Interpretable Cross-Model Concept Alignment](https://mlsanity.com/paper/2502.03714/) (2025), cited by 14 papers here
- [Steering CLIP's vision transformer with sparse autoencoders](https://mlsanity.com/paper/2504.08729/) (2025), cited by 13 papers here
- [Towards Vision-Language Mechanistic Interpretability: A Causal Tracing Tool for BLIP](https://mlsanity.com/paper/2308.14179/) (2023), cited by 8 papers here
- [Large Multi-modal Models Can Interpret Features in Large Multi-modal Models](https://mlsanity.com/paper/2411.14982/) (2024), cited by 7 papers here
- [Interpretable and Testable Vision Features via Sparse Autoencoders](https://mlsanity.com/paper/2502.06755/) (2025), cited by 6 papers here
- [SAE-V: Interpreting Multimodal Models for Enhanced Alignment](https://mlsanity.com/paper/2502.17514/) (2025), cited by 5 papers here
- [Interpreting the linear structure of vision-language model embedding spaces](https://mlsanity.com/paper/2504.11695/) (2025), cited by 5 papers here

## New (24 Sep–14 Oct)

- [Decoding the Functional Roles of Register and High-Norm Patch Tokens in Vision Transformers](https://mlsanity.com/paper/2610.03698/) (2026-10-02)
- [Corrupted but Correct: Why Vision-Language Models Lie to Themselves Internally](https://mlsanity.com/paper/2610.03445/) (2026-10-02)
- [From Patching to Pruning Visual Computation in Vision Language Models](https://mlsanity.com/paper/2610.03389/) (2026-10-02)
- [Augmenting Visual Anomaly Detection with Automated Interpretability](https://mlsanity.com/paper/2609.33818/) (2026-09-27)