AutoRound 两行参数修正：GGUF 量化精度大幅提升
原标题：Same bytes, closer to the original: two lines of AutoRound we had wrong
AI 摘要
Archsloth 在 Hugging Face 博客中披露，其 GGUF 量化流程中 AutoRound 的两个参数设置错误：误用 W4A16 方案却以 Q4_K_M 格式导出，且未启用 SignRoundV2 符号梯度搜索。修正后，Qwen3-4B 的 Q4_K_M 量化在韩语、代码等十个测试轴上 KL 散度较 unsloth 版本降低 20.9% 至 54.4%。文章还指出校准文本语言对 AutoRound 结果影响显著，韩英混合校准使韩语 KLD 降低 29.6%，远超逐层比特分配带来的收益。
正文节选
Two lines of AutoRound we had wrong, and one lever nobody prices in We publish GGUF quantizations under Archsloth. This post is about the part that is actually useful to you: two flags we had set wrong for months, and one input that turns out to decide the result even though the field treats it as neutral. Every number below is measured on files you can download, against comparison files you can download, with a command you can run. The raw per-run logs ship inside each repository. Take unsloth/