The recipe is evaluated across Qwen2.5, Zephyr, and Gemma backbones only at the 3B-9B scale; transfer to substantially larger models remains untested.

The recipe is evaluated across Qwen2.5, Zephyr, and Gemma backbones only at the 3B-9B scale; transfer to substantially larger models remains untested. · CiteArk

The recipe is evaluated across Qwen2.5, Zephyr, and Gemma backbones only at the 3B-9B scale; transfer to substantially larger models remains untested.

报告指标与观测值

Assessment（0）

关联运行（0）