mlsanityData to 7 Oct 2026

Prisma: An Open Source Toolkit for Mechanistic Interpretability in Vision and Video

arXiv 2504.19475 ↗First posted 28 Apr 2025cs.CV cs.AI cs.LGcited by 20

Not on the map. Why?

Abstract (arXiv)

Robust tooling and publicly available pre-trained models have helped drive recent advances in mechanistic interpretability for language models. However, similar progress in vision mechanistic interpretability has been hindered by the lack of accessible frameworks and pre-trained weights. We present Prisma (Access the codebase here: https://github.com/Prisma-Multimodal/ViT-Prisma), an open-source framework designed to accelerate vision mechanistic interpretability research, providing a unified toolkit for accessing 75+ vision and video transformers; support for sparse autoencoder (SAE), transcoder, and crosscoder training; a suite of 80+ pre-trained SAE weights; activation caching, circuit analysis tools, and visualization tools; and educational resources. Our analysis reveals surprising findings, including that effective vision SAEs can exhibit substantially lower sparsity patterns than language SAEs, and that in some instances, SAE reconstructions can decrease model loss. Prisma enables new research directions for understanding vision model internals while lowering barriers to entry in this emerging field.

Key facts. Cited by 20 papers in the map (data through 7 Oct 2026). Builds on 31 of its 57 references that are in the map. First posted on arXiv 28 Apr 2025.

Builds on

31 of 57 references.

References · 31

  1. 2021Learning Transferable Visual Models From Natural Language Supervisioncited by 33,456
  2. 2020An Image is Worth 16x16 Words: Transformers for Image Recognition at Scalecited by 29,915
  3. 2023DINOv2: Learning Robust Visual Features without Supervisioncited by 7,764
  4. 2021Emerging Properties in Self-Supervised Vision Transformerscited by 6,893
  5. 2021ViViT: A Video Vision Transformercited by 1,610
  6. 2022Interpretability in the Wild: a Circuit for Indirect Object Identification in GPT-2 smallcited by 1,130
  7. 2022Toy Models of Superpositioncited by 928
  8. 2023Progress measures for grokking via mechanistic interpretabilitycited by 822
  9. 2023Towards Automated Circuit Discovery for Mechanistic Interpretabilitycited by 700
  10. 2024Scaling and evaluating sparse autoencoderscited by 530
  11. 2024Revisiting Feature Prediction for Learning Visual Representations from Videocited by 490
  12. 2021Intriguing Properties of Vision Transformerscited by 446
19 more
  1. 2024Sparse Feature Circuits: Discovering and Editing Interpretable Causal Graphs in Language Modelscited by 412
  2. 2024Jumping Ahead: Improving Reconstruction Fidelity with JumpReLU Sparse Autoencoderscited by 283
  3. 2023Interpreting CLIP's Image Representation via Text-Based Decompositioncited by 189
  4. 2024Transcoders Find Interpretable LLM Feature Circuitscited by 186
  5. 2024Improving Dictionary Learning with Gated Sparse Autoencoderscited by 171
  6. 2019REPAIR: Removing Representation Bias by Dataset Resamplingcited by 157
  7. 2022From Attribution Maps to Human-Understandable Explanations through Concept Relevance Propagationcited by 141
  8. 2019Why Can't I Dance in the Mall? Learning to Mitigate Scene Bias in Action Recognitioncited by 117
  9. 2024AtP*: An efficient and scalable method for localizing LLM behaviour to componentscited by 100
  10. 2024Sparse autoencoders reveal selective remapping of visual concepts during adaptationcited by 66
  11. 2024Interpreting the Second-Order Effects of Neurons in CLIPcited by 36
  12. 2025Steering CLIP's vision transformer with sparse autoencoderscited by 33
  13. 2023Analyzing Vision Transformers for Image Classification in Class Embedding Spacecited by 26
  14. 2023Towards Evaluating Explanations of Vision Transformers for Medical Imagingcited by 22
  15. 2024Understanding Video Transformers via Universal Concept Discoverycited by 22
  16. 2018Video Time: Properties, Encoders and Evaluationcited by 19
  17. 2022Is Appearance Free Action Recognition Possible?cited by 16
  18. 2022Quantifying and Learning Static vs. Dynamic Information in Deep Spatiotemporal Networkscited by 7
  19. 2024Interpretability in Action: Exploratory Analysis of VPT, a Minecraft Agentcited by 4

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2025: 10; 2026: 10.

20 papers.

Sort:

By area:

  1. 2026Sanity Checks for Sparse Autoencoders: Do SAEs Beat Random Baselines?cited by 10
  2. 2025From Behavioral Performance to Internal Competence: Interpreting Vision-Language Models with VLM-Lenscited by 9
  3. 2025Adversarial Attacks Leverage Interference Between Features in Superpositioncited by 7
  4. 2025Too Late to Recall: Explaining the Two-Hop Problem in Multimodal Knowledge Retrievalcited by 6
  5. 2025From Noise to Narrative: Tracing the Origins of Hallucinations in Transformerscited by 4
  6. 2025A Geometric Unification of Concept Learning with Concept Conescited by 3
  7. 2025DB-KSVD: Scalable Alternating Optimization for Disentangling High-Dimensional Embedding Spacescited by 2
  8. 2025Visual Exploration of Feature Relationships in Sparse Autoencoders with Curated Conceptscited by 2
  9. 2026Sparse CLIP: Co-Optimizing Interpretability and Performance in Contrastive Learningcited by 2
  10. 2025Interpreto: An Explainability Library for Transformerscited by 1
  11. 2026Seeing Through Circuits: Faithful Mechanistic Interpretability for Vision Transformerscited by 1
  12. 2026VFUSE: Virulent Feature Understanding with Sparse autoEncoderscited by 1
8 more
  1. 2025CircuitProbe: Tracing Visual Temporal Evidence Flow in Video Language Modelscited by 0
  2. 2025The Deleuzian Representation Hypothesiscited by 0
  3. 2026DreamReader: An Interpretability Toolkit for Text-to-Image Modelscited by 0
  4. 2026Contrastive Semantic Projection: Faithful Neuron Labeling with Contrastive Examplescited by 0
  5. 2026When Confidence Lacks Concepts: Interpretable OOD Detection via Representation Perturbationscited by 0
  6. 2026IRIS: A Visual Cortex-Inspired Framework for Analyzing Orientation Selectivity in Vision Transformerscited by 0
  7. 2026Multimodal Model Diffing for Feature Discovery and Controlcited by 0
  8. 2026SCOUT: Semantic Concept Discovery for Open-Vocabulary Editing of face Recognition Templatescited by 0