Downstream task evaluation after post-training stage DPO for 7b_rkd.
Source: source_paper:Table 8, Page 33, After DPO
Reported and observed measurements
downstream_accuracy
t8_dpo_7b_rkd_gsm8k
Reported 64.5 percentage_points
Observed — percentage_points
downstream_accuracy
t8_dpo_7b_rkd_gsm_s
Reported 40 percentage_points
Observed — percentage_points
downstream_accuracy
t8_dpo_7b_rkd_gsmplus
Reported 36.4 percentage_points
Observed — percentage_points
downstream_accuracy
t8_dpo_7b_rkd_bbh
Reported 34.2 percentage_points
Observed — percentage_points
downstream_accuracy
t8_dpo_7b_rkd_drop
Reported 40.5 percentage_points
Observed — percentage_points
downstream_accuracy
t8_dpo_7b_rkd_math
Reported 10 percentage_points
Observed — percentage_points
downstream_accuracy
t8_dpo_7b_rkd_tqa
Reported 52.8 percentage_points
Observed — percentage_points
downstream_accuracy
t8_dpo_7b_rkd_nq
Reported 22.3 percentage_points
Observed — percentage_points
downstream_accuracy
t8_dpo_7b_rkd_sqa
Reported 7.8 percentage_points
Observed — percentage_points
downstream_accuracy
t8_dpo_7b_rkd_mmlu
Reported 48.8 percentage_points
Observed — percentage_points
downstream_accuracy
t8_dpo_7b_rkd_mmlu_p
Reported 17.3 percentage_points
Observed — percentage_points
downstream_accuracy
t8_dpo_7b_rkd_arc_c
Reported 57.4 percentage_points
Observed — percentage_points
downstream_accuracy
t8_dpo_7b_rkd_obqa
Reported 55 percentage_points
Observed — percentage_points
downstream_accuracy
t8_dpo_7b_rkd_wino
Reported 52.5 percentage_points
Observed — percentage_points
downstream_accuracy
t8_dpo_7b_rkd_agi
Reported 38.2 percentage_points
Observed — percentage_points
downstream_accuracy
t8_dpo_7b_rkd_ife
Reported 64.1 percentage_points
Observed — percentage_points
Assessments (0)
No immutable Assessment has been published for this Claim yet.
Runs (0)
No execution has been linked to this Claim yet.