Downstream task evaluation after post-training stage RLVR1 for 13b_rkd.
来源：source_paper:Table 8, Page 33, After RLVR1
报告指标与观测值
downstream_accuracy
t8_rlvr1_13b_rkd_gsm8k
论文报告 75.6 percentage_points
实际观测 — percentage_points
downstream_accuracy
t8_rlvr1_13b_rkd_gsm_s
论文报告 60.1 percentage_points
实际观测 — percentage_points
downstream_accuracy
t8_rlvr1_13b_rkd_gsmplus
论文报告 50.3 percentage_points
实际观测 — percentage_points
downstream_accuracy
t8_rlvr1_13b_rkd_bbh
论文报告 33.3 percentage_points
实际观测 — percentage_points
downstream_accuracy
t8_rlvr1_13b_rkd_drop
论文报告 38.1 percentage_points
实际观测 — percentage_points
downstream_accuracy
t8_rlvr1_13b_rkd_math
论文报告 15.8 percentage_points
实际观测 — percentage_points
downstream_accuracy
t8_rlvr1_13b_rkd_tqa
论文报告 53.3 percentage_points
实际观测 — percentage_points
downstream_accuracy
t8_rlvr1_13b_rkd_nq
论文报告 22.3 percentage_points
实际观测 — percentage_points
downstream_accuracy
t8_rlvr1_13b_rkd_sqa
论文报告 8 percentage_points
实际观测 — percentage_points
downstream_accuracy
t8_rlvr1_13b_rkd_mmlu
论文报告 36.2 percentage_points
实际观测 — percentage_points
downstream_accuracy
t8_rlvr1_13b_rkd_mmlu_p
论文报告 16.9 percentage_points
实际观测 — percentage_points
downstream_accuracy
t8_rlvr1_13b_rkd_arc_c
论文报告 55.9 percentage_points
实际观测 — percentage_points
downstream_accuracy
t8_rlvr1_13b_rkd_obqa
论文报告 55.2 percentage_points
实际观测 — percentage_points
downstream_accuracy
t8_rlvr1_13b_rkd_wino
论文报告 51.3 percentage_points
实际观测 — percentage_points
downstream_accuracy
t8_rlvr1_13b_rkd_agi
论文报告 36.4 percentage_points
实际观测 — percentage_points
downstream_accuracy
t8_rlvr1_13b_rkd_ife
论文报告 63.6 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。