Shan Chen, Jack Gallifant, Kuleen Sasse, Danielle Bitterman[1] Please read this as a work in progress where we are colleagues sharing this in a lab (https://www.bittermanlab.org) meeting to help/motivate potential parallel research. TL;DR: * Recent work has evaluated the generalizability of Sparse Autoencoder (SAE) features; this study examines their effectiveness...
A few months ago, we explored whether Sparse Autoencoder (SAE) features from a base model remained meaningful when transferred to a multimodal system—specifically, LLaVA—in our preliminary post Are SAE Features from the Base Model still meaningful to LLaVA?. Today, I’m excited to share how that initial work has evolved. Our...
Shan Chen, Jack Gallifant, Kuleen Sasse, Danielle Bitterman[1] Please read this as a work in progress where we are colleagues sharing this in a lab (https://www.bittermanlab.org) meeting to help/motivate potential parallel research. TL;DR: * Recent work has evaluated the generalizability of Sparse Autoencoder (SAE) features; this study examines their effectiveness...
Shan Chen, Jack Gallifant, Kuleen Sasse, Danielle Bitterman[1]
Please read this as a work in progress where we are colleagues sharing this in a lab (https://www.bittermanlab.org) meeting to help/motivate potential parallel research.
A few months ago, we explored whether Sparse Autoencoder (SAE) features from a base model remained meaningful when transferred to a multimodal system—specifically, LLaVA—in our preliminary post Are SAE Features from the Base Model still meaningful to LLaVA?. Today, I’m excited to share how that initial work has evolved. Our new arXiv paper, Sparse Autoencoder Features for Classifications and Transferability.
Our study makes three key contributions to the field of interpretable AI and feature extraction in Large Language Models (LLMs). First, it establishes classification benchmarks by introducing a robust methodology for evaluating and selecting Sparse Autoencoder (SAE)-based features in safety-critical classification tasks, demonstrating their superior performance over traditional baselines. Second, it provides a... (read more)
thank you! full paper and code will be ready soon!
Thank you! And thanks for making llava-gemma!
Shan Chen, Jack Gallifant, Kuleen Sasse, Danielle Bitterman[1]
Please read this as a work in progress where we are colleagues sharing this in a lab (https://www.bittermanlab.org) meeting to help/motivate potential parallel research.