Search papers, labs, and topics across Lattice.
HiPO extends Direct Preference Optimization (DPO) by decomposing LLM responses into reasoning segments (query clarification, reasoning steps, answer) and applying a weighted DPO loss to each. This allows for segment-specific training, addressing DPO's limitations in providing granular feedback for complex reasoning tasks. Experiments on 7B LLMs fine-tuned with HiPO on the Math Stack Exchange dataset show improved performance on math benchmarks and enhanced reasoning structure compared to DPO.
LLMs can learn to reason more effectively by breaking down the reasoning process and optimizing each step individually.
Direct Preference Optimization (DPO) is an effective framework for aligning large language models with human preferences, but it struggles with complex reasoning tasks. DPO optimizes for the likelihood of generating preferred over dispreferred responses in their entirety and lacks the granularity to provide feedback on subsections of many-step solutions typical of reasoning tasks. Existing methods excel at either stable preference learning (e.g., DPO variants like KTO and RSO) or structured reasoning (e.g., ReMA's multi-agent RL framework, Tree of Thoughts), but fail to merge these complementary strengths. We propose HiPO (Hierarchical Preference Optimization), an extension of DPO that separates responses into reasoning segments (query clarification and context, reasoning steps, and answer) and computes loss as a weighted sum of the DPO loss for each segment. Our approach enables segment-specific training while maintaining DPO's computational efficiency and training stability. We demonstrate that for multiple 7B LLMs fine-tuned using HiPO and DPO on the Math Stack Exchange preference dataset, the models trained with HiPO outperform the others on a variety of common math benchmarks and achieve greater organization, logical flow, and consistency as measured by GPT-4.1.
