Risk-aware Direct Preference Optimization under Nested Risk Measure

Jun Wang (iWudao Tech) · Yaodong Yang (AIG) · Lijun Zhang (Nanjing University (NJU)) · Wei Wei (shanxi university) · Lin Li (Shanxi University) · Yajie Qi (Shanxi University) · Huizhong Song (Shanxi University)

advantage function alignment performance bradley-terry model constrained optimization fine-tuning kl divergence likelihood maximizationmodel driftnested risk measuresreward maximization risk controlrisk-aware direct preference optimizationsequential risk ratiotoken-level representation

When fine-tuning pre-trained Large Language Models (LLMs) to align with human values and intentions, maximizing the estimated reward can lead to superior performance, but it also introduces potential risks due to deviations from the reference model's intended behavior. Most existing methods typically introduce KL divergence to constrain deviations between the trained model and the reference model; however, this may not be sufficient in certain applications that require tight risk control. In this paper, we introduce Risk-aware Direct Preference Optimization (Ra-DPO), a novel approach that incorporates risk-awareness by employing a class of nested risk measures. This approach formulates a constrained risk-aware advantage function maximization problem and then converts the Bradley-Terry model into a token-level representation. The objective function maximizes the likelihood of the policy while suppressing the deviation between a trained model and the reference model using a sequential risk ratio, thereby enhancing the model's risk-awareness. Experimental results across three open-source datasets: IMDb Dataset, Anthropic HH Dataset, and AlpacaEval, demonstrate the proposed method's superior performance in balancing alignment performance and model drift.