arXivJake Hyun, Ziyu Zhao, Jae-sun Seo, Mohamed S. AbdelfattahThu, Nov 5, 2026, 9:42 AM PST
score 16.9
Compressed AI models borrow extra precision from RAM and disk
Original: StrataFlux: Memory-Hierarchy-Native Low-Bit LLM Inference
Source: arxiv.org ↗
Writing ELI5 summary…