EyeMulator: Improving Code Language Models by Mimicking Human Visual Attention

TL;DR

EyeMulator enhances code language models by aligning their attention with human visual focus, leading to substantial improvements in code translation and summarization tasks.

Contribution

It introduces a novel method to mimic human visual attention in code models without changing their architecture, improving performance.

Findings

Over 30 CodeBLEU points improvement in translation

Up to 22 BERTScore points in summarization

Attention alignment with human gaze improves model accuracy

Abstract

Code Language Models (CodeLLMs) traditionally learn attention based solely on statistical input-output token correlations ("machine attention"). In contrast, human developers rely on intuition, selectively fixating on semantically salient tokens during program comprehension. We present EyeMulator, a model-agnostic technique to align CodeLLM attention with human visual attention without architectural changes. By extracting scan paths from eye-tracking data, we derive token-level attention weights used to augment the loss function during fine-tuning. This induces the model to mimic human focus. Our evaluation across StarCoder, Llama-3.2, and DeepSeek-Coder shows that EyeMulator significantly outperforms baselines, achieving gains of over 30 CodeBLEU points in translation and up to 22 BERTScore points in summarization. Ablation studies confirm that these gains stem directly from replicating human attention dynamics. Artifacts are available at https://zenodo.org/records/17205682.