CodeV: Issue Resolving with Visual Data

TL;DR

CodeV introduces a novel method that incorporates visual data into large language models to improve GitHub issue resolution, addressing a gap in current approaches that focus solely on textual information.

Contribution

This paper presents the first approach leveraging visual data for issue resolving with LLMs and introduces a new benchmark, Visual SWE-bench, for evaluating such methods.

Findings

CodeV outperforms existing text-only models on the new benchmark.

Visual data significantly enhances issue resolution accuracy.

Extensive experiments validate the effectiveness of incorporating visual information.

Abstract

Large Language Models (LLMs) have advanced rapidly in recent years, with their applications in software engineering expanding to more complex repository-level tasks. GitHub issue resolving is a key challenge among these tasks. While recent approaches have made progress on this task, they focus on textual data within issues, neglecting visual data. However, this visual data is crucial for resolving issues as it conveys additional knowledge that text alone cannot. We propose CodeV, the first approach to leveraging visual data to enhance the issue-resolving capabilities of LLMs. CodeV resolves each issue by following a two-phase process: data processing and patch generation. To evaluate CodeV, we construct a benchmark for visual issue resolving, namely Visual SWE-bench. Through extensive experiments, we demonstrate the effectiveness of CodeV, as well as provide valuable insights into leveraging visual data to resolve GitHub issues.