眼见为实:提示GPT-4V以实现更好的视觉指令微调
To See is to Believe: Prompting GPT-4V for Better Visual Instruction Tuning
AI总结:
本文提出细粒度视觉指令数据集LVIS-Instruct4V,通过GPT-4V基于LVIS图像生成220K条上下文感知指令,替换LLaVA-Instruct后显著提升LLaVA-1.5在多个多模态基准上的性能。
摘要:
现有的视觉指令微调方法通常通过文本描述来提示大语言模型,以生成指令跟随数据。尽管取得了令人鼓舞的性能,但这些描述来源于图像标注,往往粒度较粗。此外,在未观察完整视觉上下文的情况下,指令甚至可能与视觉内容相矛盾。为了解决这一挑战,我们引入了一个细粒度的视觉指令数据集LVIS-Instruct4V,其中包含220K条视觉对齐且具备上下文感知的指令,这些指令是通过使用来自LVIS的图像提示强大的GPT-4V生成的。通过实验验证和案例研究,我们证明高质量的视觉指令数据能够以明显优势提升当前最先进的大型多模态模型LLaVA-1.5在广泛基准上的性能。值得注意的是,仅通过用我们的LVIS-Instruct4V替换LLaVA-Instruct,我们在大多数具有挑战性的LMM基准上取得了优于LLaVA的结果,例如LLaVA$^w$(76.7对70.7)和MM-Vet(40.2对35.4)。我们在https://github.com/X2FD/LVIS-INSTRUCT4V发布了数据和模型。
英文摘要:
Existing visual instruction tuning methods typically prompt large language models with textual descriptions to generate instruction-following data. Despite the promising performance achieved, these descriptions are derived from image annotations, which are oftentimes coarse-grained. Furthermore, the instructions might even contradict the visual content without observing the entire visual context. To address this challenge, we introduce a fine-grained visual instruction dataset, LVIS-Instruct4V, which contains 220K visually aligned and context-aware instructions produced by prompting the powerful GPT-4V with images from LVIS. Through experimental validation and case studies, we demonstrate that high-quality visual instructional data could improve the performance of LLaVA-1.5, a state-of-the-art large multimodal model, across a wide spectrum of benchmarks by clear margins. Notably, by simply replacing the LLaVA-Instruct with our LVIS-Instruct4V, we achieve better results than LLaVA on most challenging LMM benchmarks, e.g., LLaVA$^w$ (76.7 vs. 70.7) and MM-Vet (40.2 vs. 35.4). We release our data and model at https://github.com/X2FD/LVIS-INSTRUCT4V.
机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。