Stealing LLMs (MIT, Microsoft, Harvard) #ai
14:35
GPT-4o goes CRAZY & Creates SCIENCE ABSTRACTIONS (that work)
29:27
Dirichlet Energy Minimization Explains In-Context Learning (Harvard)
24:09
The Dark Matter of AI [Mechanistic Interpretability]
24:07
AI can't cross this line and we don't know why.
29:11
Decoding on Graphs: Empower LLMs with KGs (MIT)
10:23
Grandes Modelos Conceituais (LCMs) da Meta: A Era da IA depois dos LLMs?
26:43
NEW: Better In-Context Learning ICL, Improved RAG (Harvard)
27:55