Feature-Guided SAE Steering for Refusal-Rate Control using Contrasting Prompts

Samaksh Bhargav, Zining Zhu

Keywords: AI Safety, Sparse Autoencoders, Steering

Other Keywords: Contrasting Prompts

TL;DR: Developed a method to identify and steer specific neural features in Large Language Models using Sparse Autoencoders, achieving simultaneous improvements in AI safety (+18.9%) and utility (+11.1%)

Abstract: Large Language Model (LLM) deployment requires guiding the LLM to recognize and not answer unsafe prompts while complying with safe prompts. Previous methods for achieving this require adjusting model weights along with other expensive procedures. While recent advances in Sparse Autoencoders (SAEs) have enabled interpretable feature extraction from LLMs, existing approaches lack systematic feature selection methods and principled evaluation of safety-utility tradeoffs. We explored using different steering features and steering strengths using Sparse Auto Encoders (SAEs) to provide a solution. Using an accurate and innovative contrasting prompt method with the AI-Generated Prompts Dataset from teknium/OpenHermes-2p5-Mistral-7B and Air Bench eu-dataset to efficiently choose the best features in the model to steer, we tested this method on Llama-3 8B. We conclude that using this method, our approach achieves an 18.9\% improvement in safety performance while simultaneously increasing utility by 11.1\%, demonstrating that targeted SAE steering can overcome traditional safety-utility tradeoffs when optimal features are identified through principled selection methods.

Submission Number: 53