Nepotistically Trained Generative-AI Models Collapse
This identifies a critical failure mode for generative-AI models, which could undermine their reliability in applications like content creation.
The paper tackles the problem of generative-AI models collapsing when retrained on their own outputs, showing that even small amounts of such retraining produce highly distorted images that persist even after retraining on real images.
Trained on massive amounts of human-generated content, AI-generated image synthesis is capable of reproducing semantically coherent images that match the visual appearance of its training data. We show that when retrained on even small amounts of their own creation, these generative-AI models produce highly distorted images. We also show that this distortion extends beyond the text prompts used in retraining, and that once affected, the models struggle to fully heal even after retraining on only real images.