Vision SAE

Creator

Creator

Created

Created

2025 Jan 30 0:4

Editor

Editor

Edited

Edited

2025 Jan 30 0:6

Refs

Refs

clip

A suite of Vision Sparse Autoencoders — LessWrong

CLIP-Scope? Inspired by Gemma-Scope We trained 8 Sparse Autoencoders each on 1.2 billion tokens on different layers of a Vision Transformer. These (a…

A suite of Vision Sparse Autoencoders — LessWrong

https://www.lesswrong.com/posts/wrznNDMRmbQABAEMH/a-suite-of-vision-sparse-autoencoders

A suite of Vision Sparse Autoencoders — LessWrong

mats_sae_training_for_ViTs

HugoFry • Updated 2025 Jan 14 20:16

This is a streamlit app to allow users to explore features learnt by a Sparse AutoEncoder (SAE) t...

https://sae-explorer.streamlit.app/

web app

NSFW Features

Towards Multimodal Interpretability: Learning Sparse Interpretable Features in Vision Transformers — LessWrong

Executive Summary In this post I present my results from training a Sparse Autoencoder (SAE) on a CLIP Vision Transformer (ViT) using the ImageNet-1k…

Towards Multimodal Interpretability: Learning Sparse Interpretable Features in Vision Transformers — LessWrong

https://www.lesswrong.com/posts/bCtbuWraqYTDtuARg/towards-multimodal-interpretability-learning-sparse-2

Towards Multimodal Interpretability: Learning Sparse Interpretable Features in Vision Transformers — LessWrong

Recommendations

///////////