Large Language Models Are State-of-the-Art Evaluators of Code Generation

Recent advancements in the field of natural language generation have facilitated the use of large language models to assess the quality of generated text. Although these models have shown promising results in tasks such as machine translation and summarization, their applicability in code generation tasks remains limited without human involvement. The complexity of programming concepts required for such tasks makes it difficult to develop evaluation metrics that align with human judgment. Token-matching-based metrics, such as BLEU, have demonstrated weak correlations with human practitioners in code generation tasks. Moreover, the utilization of human-written test suites to evaluate functional correctness can be challenging in domains with low resources. To overcome these obstacles, we propose a new evaluation framework based on the GPT-3.5 (GPT-3.5-turbo), Our framework addresses the limitations of existing approaches by achieving superior correlations with functional correctness and human preferences, without the need for test oracles or references. We evaluate the efficacy of our framework on two different aspects (human preference and execution success) and four programming languages, comparing its performance with the state-of-the-art CodeBERTScore metric, which relies on a pre-trained model. Our results demonstrate that our framework surpasses CodeBERTScore, delivering high levels of accuracy and consistency across various programming languages and tasks. Natural language generation (NLG) systems have seen significant progress with the development of large language models (LLMs). These models have shown great promise in generating high-quality and diverse texts that can be difficult to distinguish from human-written texts (Ouyang et al., 2022). However, evaluating the quality of NLG systems remains a challenging task, primarily due to the limitations of traditional evaluation metrics. Token-matching-based metrics, such as BLEU (Papineni et al., 2002) and ROUGE (Lin, 2004), have been widely used to evaluate NLG systems but have demonstrated poor correlation with human judgment and a lack of ability to capture semantic meanings (Kocmi et al., 2021). Furthermore, these metrics require reference output, which can be challenging to obtain for new tasks and low-resource domains (Liu et al., 2023).