Cal-DPO: Calibrated Direct Preference Optimization for Language Model Alignment
This work addresses alignment issues for LLM developers and users, offering an incremental improvement over existing preference optimization methods.
The paper tackles the problem of aligning large language models with human preference data by addressing suboptimal alignment due to ignoring actual reward values in contrastive objectives, proposing Cal-DPO which calibrates implicit rewards to match ground-truth scales, resulting in substantial improvements on standard benchmarks.
We study the problem of aligning large language models (LLMs) with human preference data. Contrastive preference optimization has shown promising results in aligning LLMs with available preference data by optimizing the implicit reward associated with the policy. However, the contrastive objective focuses mainly on the relative values of implicit rewards associated with two responses while ignoring their actual values, resulting in suboptimal alignment with human preferences. To address this limitation, we propose calibrated direct preference optimization (Cal-DPO), a simple yet effective algorithm. We show that substantial improvement in alignment with the given preferences can be achieved simply by calibrating the implicit reward to ensure that the learned implicit rewards are comparable in scale to the ground-truth rewards. We demonstrate the theoretical advantages of Cal-DPO over existing approaches. The results of our experiments on a variety of standard benchmarks show that Cal-DPO remarkably improves off-the-shelf methods.