Fetching the paper…

Optimize Weight Rounding via Signed Gradient Descent for the Quantization of LLMs · Around