Knowledge Distillation During Mid-Training Favors Reasoning over Factual Recall
尚未评估计划受阻研究发现claim_fig11_factual_recall_kd_analysis_1b_13b
Gold-token probability, gradient attenuation relative to NTP, and factual recall deficits under KD hold consistently for OLMo-2 1B and 13B Instruct teachers.
来源：source_paper:Figure 11, Page 30
报告指标与观测值
这条 Claim 暂无结构化指标。
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。