Constrain Alignment with Sparse Autoencoders

TL;DR

This paper introduces Feature-level constrained Preference Optimization (FPO), a new method using sparse autoencoders to improve LLM alignment with human preferences efficiently and stably, outperforming existing techniques.

Contribution

FPO leverages pre-trained sparse autoencoders and feature-level constraints to simplify and stabilize the alignment process of large language models.

Findings

Achieves 5.08% higher win rate over baselines

Reduces computational cost significantly

Demonstrates effective alignment on benchmark datasets

Abstract

The alignment of large language models (LLMs) with human preferences remains a key challenge. While post-training techniques like Reinforcement Learning from Human Feedback (RLHF) and Direct Preference Optimization (DPO) have achieved notable success, they often introduce computational inefficiencies and training instability. In this paper, we propose Feature-level constrained Preference Optimization (FPO), a novel method designed to simplify the alignment process while ensuring stability. FPO leverages pre-trained Sparse Autoencoders (SAEs) and introduces feature-level constraints, allowing for efficient, sparsity-enforced alignment. Our approach enjoys efficiency by using sparse features activated in a well-trained sparse autoencoder and the quality of sequential KL divergence by using the feature-level offline reference. Experimental results on benchmark datasets demonstrate that FPO achieves a 5.08% absolute improvement in win rate with much lower computational cost compared to state-of-the-art baselines, making it a promising solution for efficient and controllable LLM alignments.