README.md

---
license: mit
---

# SAE for OpenAI's CLIP ViT-B/16 trained on ImageNet-1K Activations

![Overview of a CLIP-trained SAE](https://osu-nlp-group.github.io/SAE-V/assets/overview2.webp)

* **Homepage:** https://osu-nlp-group.github.io/SAE-V
* **Code:** https://github.com/OSU-NLP-Group/SAE-V
* **Preprint:** https://arxiv.org/abs/2502.06755
* **Demos:** https://osu-nlp-group.github.io/SAE-V#demos
* **Point of Contact:** [Sam Stevens](mailto:stevens.994@buckeyemail.osu.edu)

## Inference Instructions

Follow the instructions [here](https://osu-nlp-group.github.io/SAE-V/saev/#inference-instructions).