Visual-language pre-training (VLP) has achieved remarkable success in multi-modal tasks, largely attributed to the availability of large-scale image-text datasets. In this work, we demonstrate that multi-modal large language models (MLLMs) can enhance visual-language representation learning by improving data quality. Our approach is simple, utilizing MLLMs to extend multiple captions for each image. To prevent the bias introduced by MLLMs' hallucinations and intrinsic caption styles, we propose "text shearing" to maintain the same length for extended captions as that of the original captions. In image-text retrieval, our method consistently obtains 5.6 ~ 35.0% and 16.8 ~ 46.1% improvement on R@1 under the fine-tuning and zero-shot settings, respectively. Notably, we obtain zero-shot results that are comparable to fine-tuning on target datasets, which encourages more exploration of the versatile use of MLLMs.

翻译：视觉-语言预训练（VLP）在多模态任务中取得了显著成功，这主要归功于大规模图像-文本数据集的可用性。本研究表明，多模态大语言模型（MLLMs）可通过提升数据质量来增强视觉-语言表示学习。我们采用简洁方法，利用MLLMs为每张图像扩展多个描述文本。为避免MLLMs的幻觉现象及固有描述风格引入的偏差，我们提出"文本剪枝"策略，使扩展描述文本保持与原始描述文本相同的长度。在图像-文本检索任务中，本方法在微调和零样本设置下分别持续获得R@1指标5.6%~35.0%和16.8%~46.1%的提升。值得注意的是，我们获得的零样本结果可与目标数据集上的微调结果相媲美，这激励了对MLLMs多用途应用的更深入探索。