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

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

28 Apr 2025arXiv:2504.19475archive 2025-07-28

Sonia Joseph, Praneet Suresh, Lorenz Hufe, Edward Stevinson, Robert Graham, Yash Vadi, Danilo Bzdok, Sebastian Lapuschkin, Lee Sharkey, Blake Aaron Richards

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.

PaperPDFCodeCode Syntology ran

In Syntology Open this paper in Syntology's Atlas, the map of the papers in Syntology's graph and their citations.

For agents, Syntology's MCP tool lists every function and class Syntology harvested from this paper and whether it ran (how to connect): get_harvested_code_for_paper(arxiv_id="2504.19475")

Code

Syntology Ran 2 of 16 code samples harvested from 2 repositories linked to this paper; 14 have no recorded run. Of those that ran: 2 ran · our draft was wrong.

By repository: official repository: 2 samples from 1 repository, 2 ran; community (archive-listed): 14 samples from 1 repository, 0 ran. The run record, sample by sample. “Ran” means executed on a synthesized input, not that the code is correct or reproduces the paper.

prisma-multimodal/vit-prisma officialmentioned in papermentioned on GitHubjax report
soniajoseph/vit-prisma mentioned on GitHubjaxMIT report

Repository list and official/mentioned flags are the archive's, frozen 2025-07-28. Reachability, where shown, is from one Syntology probe window (2026-09-16 to 2026-09-18); repositories not probed show nothing. GitHub stars are not tracked.

Code Syntology ran Syntology

16 samples harvested; 2 ran; 0 honoured the contract we drafted; 14 have no recorded run. Read from Syntology's graph 2026-09-24; that is when this build read the record, not when the samples ran.

2ran · our draft was wrong
14unverified

Licence: 0 of the 16 samples are pointer only, meaning Syntology does not serve that copy's text. This page shows no code text for any sample; each one links to its file in the repository.

Harvested from 2 repositories linked to this paper, official or community; each sample names its own and says which. “Ran” means the sample executed on a synthesized input. It does not mean the output is correct, and nothing here reproduces the paper's results. “Honoured” and “violated” refer to a contract Syntology drafted from the code itself; “our draft was wrong” and “fixture could not drive it” are failures of Syntology's instrument, not of the code.

Each sample ends with its code_sha256, Syntology's identity for that exact code. An agent fetches the stored sample with Syntology's MCP tool get_code(code_sha256="…") (how to connect); click an identity to copy that call.

Repository labels, per sample. official repository: The archive marks this repository official for the paper. named in the paper: The archive records that the paper mentions this repository; it is not marked official. community (archive-listed): In the archive's code links for this paper, not marked official and not recorded as mentioned in the paper. found in paper text by Syntology: Syntology found this repository in the paper's own text; whether it is the authors' implementation is not asserted. community: Not in the archive's code links for this paper; a community repository Syntology harvested. Samples from a repository marked official are listed first. Licence labels name the repository's licence as recorded at harvest. “Pointer only” means Syntology does not serve that copy's text, for one of four reasons: no licence file was found; the licence was not identified; the licence is recorded as permissive but that copy's record is not marked cleared; or the licence is outside the permissive list Syntology serves text under (MIT, Apache-2.0, BSD and similar). Some licences outside that list permit redistribution, such as WTFPL, and GPL-3.0 under its conditions; they are simply not on the list. Hover a licence label for the reason. File links open the file on GitHub at the default branch, which may have changed since the harvest.

check_model_name prisma-multimodal/vit-prisma/src/vit_prisma/models/model_loader.py official repository ran · our draft was wrong fingerprinted MIT (permissive) · a1fc8248bd1ad1ba · report
get_activation_fn prisma-multimodal/vit-prisma/src/vit_prisma/sae/sae.py official repository ran · our draft was wrong MIT (permissive) · bd6e1088e609f3eb · report
accuracy soniajoseph/vit-prisma/src/vit_prisma/model_eval/evaluate_imagenet.py community (archive-listed) unverified MIT (permissive) · 6ab336cad19fdbae · report
create_train_info soniajoseph/vit-prisma/src/vit_prisma/model_eval/evaluate_clip_benchmark.py community (archive-listed) unverified MIT (permissive) · 44243798f6173648 · report
draw_circle soniajoseph/vit-prisma/src/vit_prisma/dataloaders/induction.py community (archive-listed) unverified MIT (permissive) · d41780f7d4f76a23 · report
draw_circle_with_points soniajoseph/vit-prisma/src/vit_prisma/dataloaders/circle.py community (archive-listed) unverified MIT (permissive) · cd3823f34b47e5a7 · report
draw_line soniajoseph/vit-prisma/src/vit_prisma/dataloaders/induction.py community (archive-listed) unverified MIT (permissive) · 92ed88c91dec647a · report
draw_x soniajoseph/vit-prisma/src/vit_prisma/dataloaders/induction.py community (archive-listed) unverified MIT (permissive) · 187a24f916bfbe42 · report
extract_and_parse_text_labels soniajoseph/vit-prisma/src/vit_prisma/dataloaders/imagenet_dataset.py community (archive-listed) unverified MIT (permissive) · ee86a4895f763cc5 · report
get_cc_dataloaders soniajoseph/vit-prisma/src/vit_prisma/dataloaders/conceptual_captions.py community (archive-listed) unverified MIT (permissive) · 0d68e4cb26762d11 · report
get_imagenet_index_to_name soniajoseph/vit-prisma/src/vit_prisma/dataloaders/imagenet_dataset.py community (archive-listed) unverified MIT (permissive) · b62018180f976b10 · report
get_imagenet_transforms_clip soniajoseph/vit-prisma/src/vit_prisma/dataloaders/imagenet_dataset.py community (archive-listed) unverified MIT (permissive) · d1b187c5b98793b5 · report
get_train_test_data soniajoseph/vit-prisma/src/vit_prisma/dataloaders/circle.py community (archive-listed) unverified MIT (permissive) · 0f946c7b88043fde · report
plot_four_objects soniajoseph/vit-prisma/src/vit_prisma/dataloaders/polygenic_induction.py community (archive-listed) unverified MIT (permissive) · 6237c44a5a1559b3 · report
run soniajoseph/vit-prisma/src/vit_prisma/model_eval/evaluate_imagenet.py community (archive-listed) unverified MIT (permissive) · f265fd87ea3c65a9 · report
train_test_dataset soniajoseph/vit-prisma/src/vit_prisma/dataloaders/dsprites.py community (archive-listed) unverified MIT (permissive) · 627da91f7fee55bf · report

Results from the paper archive 2025-07-28

No leaderboard rows for this paper in the archive.

Methods

Sparse Autoencoder

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections