Learning from Reference Answers: Versatile Language Model Alignment without Binary Human Preference Data
This addresses the challenge of reducing data collection costs for LLM alignment in domains like safety and confidence, though it is incremental as it builds on existing alignment frameworks.
The paper tackles the problem of aligning large language models (LLMs) to be helpful, harmless, and honest without requiring resource-intensive binary human preference data and reward modeling, by using similarity between generated outputs and reference answers as a reward function, achieving performance comparable to prior methods across multiple scenarios.
Large language models~(LLMs) are expected to be helpful, harmless, and honest. In different alignment scenarios, such as safety, confidence, and general preference alignment, binary preference data collection and reward modeling are resource-intensive but play a central role in transferring human preferences. In this work, we explore using the similarity between sampled generations and reference answers as a supplementary reward function for alignment. When unary reference answers are available, such similarity-based rewards can circumvent the need for binary preference data and explicit reward modeling. We introduce \textit{RefAlign}, a versatile REINFORCE-style alignment algorithm that does not rely on reward or reference models. RefAlign utilizes language generation evaluation metrics, such as BERTScore, between sampled generations and reference answers as surrogate rewards. Beyond general preference optimization, RefAlign can be naturally extended to diverse scenarios, including safety and confidence alignment, by combining similarity-based rewards with task-specific objectives. Across multiple scenarios, RefAlign achieves performance comparable to prior alignment methods while operating without binary preference data or reward models. The code is available at https://github.com/mzhaoshuai/RefAlign.