Downstream task evaluation after post-training stage DPO for 7b_sd.
Source: source_paper:Table 8, Page 33, After DPO
Reported and observed measurements
downstream_accuracy
t8_dpo_7b_sd_gsm8k
Reported 70.9 percentage_points
Observed — percentage_points
downstream_accuracy
t8_dpo_7b_sd_gsm_s
Reported 50.8 percentage_points
Observed — percentage_points
downstream_accuracy
t8_dpo_7b_sd_gsmplus
Reported 44.2 percentage_points
Observed — percentage_points
downstream_accuracy
t8_dpo_7b_sd_bbh
Reported 35.5 percentage_points
Observed — percentage_points
downstream_accuracy
t8_dpo_7b_sd_drop
Reported 49 percentage_points
Observed — percentage_points
downstream_accuracy
t8_dpo_7b_sd_math
Reported 15 percentage_points
Observed — percentage_points
downstream_accuracy
t8_dpo_7b_sd_tqa
Reported 54.7 percentage_points
Observed — percentage_points
downstream_accuracy
t8_dpo_7b_sd_nq
Reported 23.6 percentage_points
Observed — percentage_points
downstream_accuracy
t8_dpo_7b_sd_sqa
Reported 8.1 percentage_points
Observed — percentage_points
downstream_accuracy
t8_dpo_7b_sd_mmlu
Reported 50 percentage_points
Observed — percentage_points
downstream_accuracy
t8_dpo_7b_sd_mmlu_p
Reported 20.1 percentage_points
Observed — percentage_points
downstream_accuracy
t8_dpo_7b_sd_arc_c
Reported 61.2 percentage_points
Observed — percentage_points
downstream_accuracy
t8_dpo_7b_sd_obqa
Reported 61 percentage_points
Observed — percentage_points
downstream_accuracy
t8_dpo_7b_sd_wino
Reported 55.5 percentage_points
Observed — percentage_points
downstream_accuracy
t8_dpo_7b_sd_agi
Reported 40.9 percentage_points
Observed — percentage_points
downstream_accuracy
t8_dpo_7b_sd_ife
Reported 62.1 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.