Can Coding Agents Solve Repository-Level Issues with Rendered Code? An Exploratory Study of Visual Representations

2026-08-10

AI Summary

This work investigates whether rendering source code as images to provide visual context can effectively support agent navigation, editing, and validation in repository-scale program repair tasks while reducing prompt token consumption. The authors propose a structured workflow that decouples unguided exploration from targeted repair phases and evaluate it end-to-end on the SWE-bench Verified benchmark. For the first time, they systematically quantify the compression benefits and performance limits of visual code representations within realistic agent workflows: this approach significantly reduces prompt length while preserving repair accuracy, yet exhibits instability under aggressive compression and cannot surpass the inherent performance ceilings imposed by the underlying foundation model or agent architecture.

Abstract

Visual modality has recently been explored as a way to compress textual tokens, including rendering code as images for static code understanding. We study whether this representation can serve as operational context for agentic coding, where an agent must navigate repositories, edit source files, and verify executable patches. Using SWE-bench Verified, we evaluate rendered code in repository-level repair workflows and introduce controlled agent settings to separate unguided repository exploration from more structured repair stages. Our results show a mixed picture. Rendered code consistently reduces prompt-token cost, but the savings do not increase linearly with the nominal visual compression ratio. It largely preserves end-to-end repair accuracy, but does not overcome the performance limits of the underlying model or agent architecture, and can become unstable under aggressive compression. Further analysis suggests that visual code is most useful when raw source reading is a major bottleneck; once repository localization is structured, much of the remaining cost comes from patch--test trial-and-error, where visual compression has limited leverage. Overall, our study positions rendered code as a viable but conditional compression mechanism for realistic coding agents.

Problem

• rendered code
• coding agents
• repository-level repair
• visual representations
• code understanding

Innovation

• rendered code
• visual representation
• coding agents
• repository-level repair
• token compression