Worth noting that this paper is about CLIP only, which is way simpler than llm architectures. (if I’m not mistaken) Still, interesting approach and kind of con

Worth noting that this paper is about CLIP only, which is way simpler than llm architectures. (if I’m not mistaken) Still, interesting approach and kind of confirms the experience of most people where clip models can recognize known concepts but struggle with novel ones.