|
|
--- |
|
|
title: pyvene |
|
|
emoji: 👀 |
|
|
colorFrom: pink |
|
|
colorTo: purple |
|
|
sdk: static |
|
|
pinned: false |
|
|
--- |
|
|
|
|
|
# Who are we? |
|
|
|
|
|
We are a group of hackers from Stanford's NLP group, and we are interested in LLM interpretability. |
|
|
|
|
|
`pyvene` is where we started, which stands for **py**torch model inter**vene**tion. |
|
|
|
|
|
# Resources |
|
|
|
|
|
**Supervised dictionary learning models (SDLs) and datasets releases for Gemma 2 2B and 9B: [`AxBench Collection`](https://huggingface.co/collections/pyvene/axbench-release-6787576a14657bb1fc7a5117).** |
|
|
|
|
|
**Benchmark interpretability methods at scale (AxBench) library: [`AxBench`](https://github.com/stanfordnlp/axbench).** |
|
|
|
|
|
**Representation finetuning (ReFT) library: [`pyreft`](https://github.com/stanfordnlp/pyreft).** |
|
|
|
|
|
**PyTorch model intervention library: [`pyvene`](https://github.com/stanfordnlp/pyvene).** |
|
|
|
