CodeOCR: On the Effectiveness of Vision Language Models in Code Understanding

TL;DR

This paper investigates using multimodal vision-language models with image representations of code to improve efficiency and maintain accuracy in code understanding tasks, showing promising results with significant compression.

Contribution

It is the first systematic study demonstrating that vision-language models can understand compressed code images effectively, reducing token costs substantially.

Findings

MLLMs can understand code with up to 8x token reduction.

Visual cues like syntax highlighting improve performance under compression.

Clone detection tasks are highly resilient to visual compression.

Abstract

Large Language Models (LLMs) have achieved remarkable success in source code understanding, yet as software systems grow in scale, computational efficiency has become a critical bottleneck. Currently, these models rely on a text-based paradigm that treats source code as a linear sequence of tokens, which leads to a linear increase in context length and associated computational costs. The rapid advancement of Multimodal LLMs (MLLMs) introduces an opportunity to optimize efficiency by representing source code as rendered images. Unlike text, which is difficult to compress without losing semantic meaning, the image modality is inherently suitable for compression. By adjusting resolution, images can be scaled to a fraction of their original token cost while remaining recognizable to vision-capable models. To explore the feasibility of this approach, we conduct the first systematic study on the effectiveness of MLLMs for code understanding. Our experiments reveal that: (1) MLLMs can effectively understand code with substantial token reduction, achieving up to 8x compression; (2) MLLMs can effectively leverage visual cues such as syntax highlighting, improving code completion performance under 4x compression; and (3) Code-understanding tasks like clone detection exhibit exceptional resilience to visual compression, with some compression ratios even slightly outperforming raw text inputs. Our findings highlight both the potential and current limitations of MLLMs in code understanding, which points out a shift toward image-modality code representation as a pathway to more efficient inference.