AI RESEARCH
CaptionQA: Is Your Caption as Useful as the Image Itself?
arXiv CS.CV
•
ArXi:2511.21025v2 Announce Type: replace Image captions serve as efficient surrogates for visual content in multimodal systems such as retrieval, recommendation, and multi-step agentic inference pipelines. Yet current evaluation practices miss a fundamental question: Can captions stand-in for images in real downstream tasks? We propose a utility-based benchmark, CaptionQA, to evaluate model-generated captions, where caption quality is measured by how well it s downstream tasks.