MAGID: An Automated Pipeline for Generating Synthetic Multi-modal Datasets
This addresses the problem of limited multimodal data for developers of interactive systems, though it is incremental as it builds on existing augmentation and diffusion methods.
The paper tackles the lack of rich multimodal conversational data for LLMs by introducing MAGID, a framework that augments text-only dialogues with diverse, high-quality images using a diffusion model and feedback loops, achieving results comparable to or better than SOTA baselines with significant improvements in human evaluation, especially against retrieval baselines with small image databases.
Development of multimodal interactive systems is hindered by the lack of rich, multimodal (text, images) conversational data, which is needed in large quantities for LLMs. Previous approaches augment textual dialogues with retrieved images, posing privacy, diversity, and quality constraints. In this work, we introduce Multimodal Augmented Generative Images Dialogues (MAGID), a framework to augment text-only dialogues with diverse and high-quality images. Subsequently, a diffusion model is applied to craft corresponding images, ensuring alignment with the identified text. Finally, MAGID incorporates an innovative feedback loop between an image description generation module (textual LLM) and image quality modules (addressing aesthetics, image-text matching, and safety), that work in tandem to generate high-quality and multi-modal dialogues. We compare MAGID to other SOTA baselines on three dialogue datasets, using automated and human evaluation. Our results show that MAGID is comparable to or better than baselines, with significant improvements in human evaluation, especially against retrieval baselines where the image database is small.