A Survey on Bridging VLMs and Synthetic Data
Vision-language models (VLMs) have significantly advanced multimodal AI by learning joint representations of visual and textual data. However, their progress is hindered by challenges in acquiring high-quality, aligned datasets, including issues of cost, privacy, and scarcity. On the other hand, synthetic data, created through the use of generative AI—which can even include VLMs—offers a scalable and cost-effective solution to these challenges. This paper presents the first comprehensive survey on bridging VLMs and synthetic data, exploring both the role of synthetic data in VLMs and the role of VLMs in synthetic data. First, we provide a preliminary overview by briefly explaining the architecture of two basic VLMs and, after studying a large number of previous works, offer an extensive survey of the previously proposed methodologies and potential future directions in this area.
Code (1)
Tasks
SurveySimilar Papers 제목 키워드 기반
Do Vision-Language Models Measure Up? Benchmarking Visual Measurement Reading with MeasureBench
Reading measurement instruments is effortless for humans and requires relatively little domain expertise, yet it remains surprisingly challenging for current vision-language models (VLMs) as we find in preliminary evalua…
Vision-Language Models for Deployable Social Robot Navigation: Bridging Semantic Reasoning and Low-Level Control
Social robot navigation (SRN) requires more than geometric path planning; it demands understanding human intentions, social norms, and contextual cues to generate socially compliant behaviors. Although classical navigati…
Collision AvoidanceRobot NavigationContinual Learning for VLMs: A Survey and Taxonomy Beyond Forgetting
Vision-language models (VLMs), spanning predictive architectures to generative Multimodal Large Language Models (MLLMs), have revolutionized artificial intelligence through powerful cross-modal alignment and zero-shot ge…
Compositional Zero-Shot LearningZero-shot GeneralizationContinual LearningSynthetic Tabular Data Generation: A Comparative Survey for Modern Techniques
As privacy regulations become more stringent and access to real-world data becomes increasingly constrained, synthetic data generation has emerged as a vital solution, especially for tabular datasets, which are central t…
Synthetic Data GenerationTabular Data GenerationBridging Resolution: A Survey of the State of the Art
Bridging reference resolution is an anaphora resolution task that is arguably more challenging and less studied than entity coreference resolution. Given that significant progress has been made on coreference resolution …
coreference-resolutionCoreference ResolutionSurvey