On the Generalization of Preference Learning with DPO
Keywords: preference learning, generalization bound
Abstract: Large language models (LLMs) have demonstrated remarkable capabilities but often struggle to align with human preferences, leading to harmful or undesirable outputs. Preference learning, which trains models to distinguish between preferred and non-preferred responses based on human feedback, has become a crucial component for ensuring that LLMs align with human values. Despite the widespread adoption in real-world systems, a thorough theoretical understanding of the generalization guarantees for these models remains lacking. This paper bridges that gap by introducing a new theoretical framework to analyze the generalization guarantees of models trained with direct preference optimization. While existing generalization theory often focuses on overparameterized models achieving near-optimal loss or models independent of the training process, our framework rigorously assesses how well models generalize after a finite number of gradient steps, reflecting real-world LLM training practices. By analyzing the reward margin associated with each sample and its trajectory throughout training, we can effectively bound the generalization error. We derive learning guarantees showing that, under specific conditions, models trained with DPO can correctly discern preferred responses on unseen data with high probability. These insights are empirically validated on contemporary LLMs, underscoring the practical relevance of our theory.
Primary Area: foundation or frontier models, including LLMs
Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.
Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.
Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.
No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.
Submission Number: 10640
Loading