Check-Eval: A Checklist-based Approach for Evaluating Text Quality

TL;DR

Check-Eval introduces a checklist-based framework leveraging large language models to evaluate generated text quality, achieving higher correlation with human judgments than existing metrics across benchmark datasets.

Contribution

It presents a novel, structured evaluation method combining checklist generation and assessment, improving alignment with human evaluations for natural language generation.

Findings

Outperforms existing metrics like G-Eval and GPTScore in correlation with human judgments.

Works effectively as both reference-free and reference-dependent evaluation.

Validated on Portuguese Legal Semantic Textual Similarity and SummEval datasets.

Abstract

Evaluating the quality of text generated by large language models (LLMs) remains a significant challenge. Traditional metrics often fail to align well with human judgments, particularly in tasks requiring creativity and nuance. In this paper, we propose \textsc{Check-Eval}, a novel evaluation framework leveraging LLMs to assess the quality of generated text through a checklist-based approach. \textsc{Check-Eval} can be employed as both a reference-free and reference-dependent evaluation method, providing a structured and interpretable assessment of text quality. The framework consists of two main stages: checklist generation and checklist evaluation. We validate \textsc{Check-Eval} on two benchmark datasets: Portuguese Legal Semantic Textual Similarity and \textsc{SummEval}. Our results demonstrate that \textsc{Check-Eval} achieves higher correlations with human judgments compared to existing metrics, such as \textsc{G-Eval} and \textsc{GPTScore}, underscoring its potential as a more reliable and effective evaluation framework for natural language generation tasks. The code for our experiments is available at \url{https://anonymous.4open.science/r/check-eval-0DB4}