Sub-1-Bit LLM Compression via Latent Factorization

Sub‑1‑bit LLM compression via latent factorization dramatically reduces model size while preserving performance, making it ideal for prototypes and internal wor