Towards Analyzing and Understanding the Limitations of DPO: A Theoretical Perspective

arXiv:2404.04626v1 Announce Type: new
Abstract: Direct Preference Optimization (DPO), which derives reward signals directly from pairwise preference data, has shown its effectiveness on aligning Large Language Models (LLMs) with human preferences. Despite its widespread use across various tasks, DPO has been criticized for its sensitivity to the SFT's effectiveness and its hindrance to the learning capacity towards human-preferred responses, leading to less satisfactory performance. To overcome those limitations, the theoretical understanding of DPO are indispensable but still lacking. To this …

abstract arxiv cs.ai cs.cl data direct preference optimization human language language models large language large language models limitations llms optimization perspective sensitivity sft tasks type understanding