Random Network Distillation as a Diversity Metric for Both Image and Text Generation
This provides a tool for researchers and practitioners to better evaluate diversity in generative AI, addressing a known bottleneck in model assessment.
The authors tackled the problem of quantifying data diversity in generative models by developing a new metric based on random network distillation, which they validated on both image and text data, including few-shot image generation.
Generative models are increasingly able to produce remarkably high quality images and text. The community has developed numerous evaluation metrics for comparing generative models. However, these metrics do not effectively quantify data diversity. We develop a new diversity metric that can readily be applied to data, both synthetic and natural, of any type. Our method employs random network distillation, a technique introduced in reinforcement learning. We validate and deploy this metric on both images and text. We further explore diversity in few-shot image generation, a setting which was previously difficult to evaluate.