I have just been informed that my above comment is false, the CLIP-L is in fact referring to OpenAI's, despite that also being the name of an OpenCLIP model.