AI · Technology · 2026-09-16 20:59
Researchers propose a new method that successfully breaks the 1.58-bit quantization limit for ternary LLMs, offering a new path for efficient inference at extremely low bit widths.
Auto-generated summary for reference only. RSS