Translated data: The research teams from the University of Southern California and Harvard University have jointly launched the DreamDistribution generative model, which achieves highly diverse and personalized image generation by learning from very few reference images through prompting. This method is not only applicable to text-to-image generation but also excels in the field of 3D generation. DreamDistribution has achieved outstanding results in evaluations, demonstrating its potential for application in a wider range of generative tasks.