HuggingFace Top Papers June 1-7: PEFT Scaling, Speculative Decoding, Nvidia Cosmos 3

Notable papers this week include scaling PEFT toward million personal models on trillion-parameter bases, Domino's approach to decoupling causal modeling from autoregressive drafting in speculative decoding, and Nvidia's Cosmos 3 omnimodal world models for physical AI. Code2LoRA uses hypernetworks to generate LoRA adapters dynamically under software evolution.