3 ms·
Data Compression Using Contrastive Pre-Training and Large Language ModelsLanguag
- barbazoo 3y agoI just thought about how you could probably "store" a pretty good approximation of a picture in a LLM and some input parameters that are orders of magnitude smaller in size than the original. I think that's what this paper is about right?