Knowledge Distillation During Mid-Training Favors Reasoning over Factual Recall
尚未评估计划受阻指标结果claim_t3_sd_fkl
Mid-training ablation results for Switch Distillation with FKL relative difference across Reasoning, Factual Recall, and Knowledge task groups.
来源：source_paper:Table 3, Page 9
报告指标与观测值
delta_percentage_points
t3_sd_fkl_reasoning
论文报告 -2.9 percentage_points
实际观测 — percentage_points
delta_percentage_points
t3_sd_fkl_factual_recall
论文报告 -0.2 percentage_points
实际观测 — percentage_points
delta_percentage_points
t3_sd_fkl_knowledge
论文报告 -1.4 percentage_points
实际观测 — percentage_points
Assessment（0）
这条 Claim 暂无已发布的不可变 Assessment。
关联运行（0）
这条 Claim 暂无关联执行。