PMoL: Parameter Efficient MoE for Preference Mixing of LLM Alignment

TL;DR

PMoL introduces a parameter-efficient architecture combining MoE and LoRA to improve preference mixing and alignment in LLMs, addressing limitations of RLHF with multiple preferences.

Contribution

The paper proposes PMoL, a novel model architecture that enables efficient and flexible preference mixing for LLM alignment, outperforming baseline methods.

Findings

PMoL achieves superior preference mixing capabilities.

PMoL improves preference alignment with lower training costs.

Experimental results validate PMoL's effectiveness with reward models and GPT-4o.

Abstract

Reinforcement Learning from Human Feedback (RLHF) has been proven to be an effective method for preference alignment of large language models (LLMs) and is widely used in the post-training process of LLMs. However, RLHF struggles with handling multiple competing preferences. This leads to a decrease in the alignment of LLMs with human preferences. To address this issue, we propose Preference Mixture of LoRAs (PMoL) from the perspective of model architecture, which can adapt to any number of preferences to mix. PMoL combines Mixture of Experts (MoE) and Low Rank Adaptor (LoRA). This architecture is innovatively applied to the research of preference alignment and has achieved significant performance improvement. The expert group soft loss is used to enable MoE with the ability to mix preferences. Through comprehensive evaluation by the reward model and GPT-4o, the experiment results show that PMoL has superior preference mixing capabilities compared to baseline methods. PMoL achieves better preference alignment with lower training costs.