Feb 2023
In this post, I list all papers on diffusion models that I could find on arxiv. I hope this list will be helpful in tracking papers in this area. It clearly shows that the interest in this area is exploding in the academic community.
April 2022
In this post, I will argue that modern generative models (especially diffusion models) have crossed an inflection point, where synthetic data from them can yield benefits in representation learning. The post mainly covers three topics 1) Motivation and understanding of the role of generative models in representation learning 2) Experimental evidence validating their success 3) Understanding why synthetic data yield benefit in representation learning.