LittleBit Achieves Sub-1-Bit LLM Compression

LittleBit (NeurIPS 2025) factorizes weight matrices into low-rank latent factors, binarizes them, and restores magnitude via lightweight learned scales, enabling 1.0 to 0.1 bits per weight while keeping the original inference architecture. The follow-up LittleBit-2 (ICML 2026) adds Joint-ITQ initialization, improving initialization without adding inference overhead.

• Community Discusses How to Write Top-Conference Papers2026-10-14
• NeurIPS 2026 Sydney waitlist draws concern2026-10-15
• ML PhD Student Asks What Industry Internships Really Want2026-10-14
• Call for Papers: Constrained Learning for Scientific Modelling at NeurIPS Paris Pre-event2026-10-12
• Researcher says Google Scholar citations lag for months2026-10-12