Structure-Aware Corpus Construction and User-Perception-Aligned Metrics for Large-Language-Model Code Completion

TL;DR

This paper introduces new evaluation metrics aligned with user perception and a structure-preserving data processing method to improve large language model code completion performance.

Contribution

It proposes two user-perception-aligned evaluation metrics and a novel SPSR-Graph data processing technique for better structural semantic modeling in code completion.

Findings

Evaluation metrics show higher correlation with user perception.

SPSR-Graph improves model performance in repository-level code completion.

Theoretical analysis confirms the effectiveness of proposed methods.

Abstract

Code completion technology based on large language model has significantly improved the development efficiency of programmers. However, in practical applications, there remains a gap between current commonly used code completion evaluation metrics and users' actual perception. To address this issue, we propose two evaluation metrics for code completion tasks--LCP and ROUGE-LCP, from the perspective of probabilistic modeling. Furthermore, to tackle the lack of effective structural semantic modeling and cross-module dependency information in LLMs for repository-level code completion scenarios, we propose a data processing method based on a Structure-Preserving and Semantically-Reordered Code Graph (SPSR-Graph). Through theoretical analysis and experimental validation, we demonstrate the superiority of the proposed evaluation metrics in terms of user perception consistency, as well as the effectiveness of the data processing method in enhancing model performance.