With rapid progress made in language qualities such as fluency and consistency via large language models (LLMs), there has been increasing interest in assessing alignment with diverse human preferences. Traditional metrics heavily rely on lexical similarity with human-written references and have been observed to suffer from a poor correlation with human evaluation. Furthermore, they ignore the diverse preferences of humans, a key aspect in evaluating open-ended tasks like story generation. Inspired by these challenges, we introduce an interpretable open-ended evaluation framework PerSE to assess the alignment with a specific human preference. It is tuned to deduce the specific preference from a given personal profile and evaluate the alignment between the generation and the personal preference. PerSE also explains its assessment by a detailed comment or several fine-grained scores. This enhances its interpretability, making it more suitable to tailor a personalized generation. Our 13B LLaMA-2-based PerSE shows a 15.8% increase in Kendall correlation and a 13.7% rise in accuracy on zero-shot reviewers compared to GPT-4. It also outperforms GPT-4 by 46.01% in the Kendall correlation on new domains, indicating its transferability.
View on arXiv