SCAR: Sparse Conditioned Autoencoders for Concept Detection and Steering in LLMs

Ruben Härle, Felix Friedrich, Manuel Brack, Björn Deiseroth, Patrick Schramowski, Kristian Kersting

Track: Technical

Keywords: Sparse Autoencoder, LLM, Toxicity, Safety, Alignment

TL;DR: We introduce Sparse Conditioned Autoencoder (SCAR), a novel module that facilitates concept detection and steering in LLMs, e.g. toxicity, thereby promoting ethical and safe real-world deployment.

Abstract:

Large Language Models (LLMs) have demonstrated remarkable capabilities in generating human-like text, but their output may not be aligned with the user or even produce harmful content.
This paper presents a novel approach to detect and steer concepts such as toxicity before generation.
We introduce the Sparse Conditioned Autoencoder (SCAR), a single trained module that extends the otherwise untouched LLM.
SCAR ensures full steerability, towards and away from concepts (e.g., toxic content), without compromising the quality of the model's text generation on standard evaluation benchmarks.
We demonstrate the effective application of our approach through a variety of concepts, including toxicity, safety, and writing style alignment.
As such, this work establishes a robust framework for controlling LLM generations, ensuring their ethical and safe deployment in real-world applications.

Submission Number: 32