The last couple of years gave us a fairly clean story around LLM pretraining: scale up the data, scale up the compute, use next-token prediction as a universal loss, and watch as a kind of implicit multi-task learning emerges. The evals ecosystem followed this arc - benchmark suites for reasoning,...
* Introduction * Understanding vs. Generation in Vision * Unifying Vision – Are We There Yet? * Vision Without Words: Do Models Need Language? * The Many Paths of Vision Introduction “Think with images” - this is the promise of recent models like o3 from OpenAI, which can integrate images...
This post is a collection of observations around self-supervised learning for vision-based datasets. This is by no means a complete survey of SSL techniques, and assumes familiarity with Contrastive Learning, Masked Image Modeling and Masked AutoEncoder frameworks. For readers intending to get up to speed on these topics, there are...