Explaining latent representations of generative models with large multimodal models
This work addresses the interpretability of generative AI for researchers and practitioners, but it appears incremental as it applies existing multimodal models to a known bottleneck.
The authors tackled the problem of interpreting latent variables in generative models by using large multimodal models to generate explanations, and they quantitatively evaluated the performance and uncertainty of these explanations across multiple models.
Learning interpretable representations of data generative latent factors is an important topic for the development of artificial intelligence. With the rise of the large multimodal model, it can align images with text to generate answers. In this work, we propose a framework to comprehensively explain each latent variable in the generative models using a large multimodal model. We further measure the uncertainty of our generated explanations, quantitatively evaluate the performance of explanation generation among multiple large multimodal models, and qualitatively visualize the variations of each latent variable to learn the disentanglement effects of different generative models on explanations. Finally, we discuss the explanatory capabilities and limitations of state-of-the-art large multimodal models.