Linear representations in language models can change dramatically over a conversation Paper • 2601.20834 • Published 10 days ago • 21
daVinci-Agency: Unlocking Long-Horizon Agency Data-Efficiently Paper • 2602.02619 • Published 5 days ago • 47
OmniSIFT: Modality-Asymmetric Token Compression for Efficient Omni-modal Large Language Models Paper • 2602.04804 • Published 3 days ago • 44
