MinosEval: Distinguishing Factoid and Non-Factoid for Tailored Open-Ended QA Evaluation with LLMs

TL;DR

MinosEval introduces a question-type-aware evaluation framework for open-ended QA that improves alignment with human judgment and interpretability by distinguishing factoid from non-factoid questions and applying tailored scoring strategies.

Contribution

It is the first to explicitly differentiate question types in open-ended QA evaluation, employing adaptive scoring methods for factoid and non-factoid questions to enhance interpretability and accuracy.

Findings

MinosEval outperforms traditional metrics in correlating with human judgments.

It provides more interpretable evaluation results through question-type differentiation.

Experiments on multiple datasets demonstrate its robustness and effectiveness.

Abstract

Open-ended question answering (QA) is a key task for evaluating the capabilities of large language models (LLMs). Compared to closed-ended QA, it demands longer answer statements, more nuanced reasoning processes, and diverse expressions, making refined and interpretable automatic evaluation both crucial and challenging. Traditional metrics like ROUGE and BERTScore struggle to capture semantic similarities due to different patterns between model responses and reference answers. Current LLM-based evaluation approaches, such as pairwise or listwise comparisons of candidate answers, lack intuitive interpretability. While pointwise scoring of each response provides some descriptions, it fails to adapt across different question contents. Most notably, existing methods overlook the distinction between factoid and non-factoid questions. To address these challenges, we propose \textbf{MinosEval}, a novel evaluation method that first distinguishes open-ended questions and then ranks candidate answers using different evaluation strategies. For factoid questions, it applies an adaptive key-point scoring strategy, while for non-factoid questions, it uses an instance-aware listwise ranking strategy. Experiments on multiple open-ended QA datasets, including self-built ones with more candidate responses to complement community resources, show that MinosEval better aligns with human annotations and offers more interpretable results.