Direct Large Language Model Alignment Through Self-rewarding Contrastive Prompt Distillation

Liu Aiwei, Bai Haoping, Lu Zhiyun, Kong Xiang, Wang Simon, Shan Jiulong, Cao Meng, Wen Lijie. Arxiv 2024

[Paper]    
Distillation Efficiency And Optimization Prompting Reinforcement Learning

Aligning large language models (LLMs) with human expectations without human-annotated preference data is an important problem. In this paper, we propose a method to evaluate the response preference by using the output probabilities of response pairs under contrastive prompt pairs, which could achieve better performance on LLaMA2-7B and LLaMA2-13B compared to RLAIF. Based on this, we propose an automatic alignment method, Direct Large Model Alignment (DLMA). First, we use contrastive prompt pairs to automatically generate preference data. Then, we continue to evaluate the generated preference data using contrastive prompt pairs and calculate a self-rewarding score. Finally, we use the DPO algorithm to effectively align LLMs by combining this self-rewarding score. In the experimental stage, our DLMA method could surpass the \texttt{RLHF} method without relying on human-annotated preference data.

Similar Work