Downstream task evaluation after post-training stage DPO for 13b_sd.
Source: source_paper:Table 8, Page 33, After DPO
Reported and observed measurements
downstream_accuracy
t8_dpo_13b_sd_gsm8k
Reported 69.9 percentage_points
Observed — percentage_points
downstream_accuracy
t8_dpo_13b_sd_gsm_s
Reported 45.7 percentage_points
Observed — percentage_points
downstream_accuracy
t8_dpo_13b_sd_gsmplus
Reported 39.5 percentage_points
Observed — percentage_points
downstream_accuracy
t8_dpo_13b_sd_bbh
Reported 32.2 percentage_points
Observed — percentage_points
downstream_accuracy
t8_dpo_13b_sd_drop
Reported 44.2 percentage_points
Observed — percentage_points
downstream_accuracy
t8_dpo_13b_sd_math
Reported 10 percentage_points
Observed — percentage_points
downstream_accuracy
t8_dpo_13b_sd_tqa
Reported 54.5 percentage_points
Observed — percentage_points
downstream_accuracy
t8_dpo_13b_sd_nq
Reported 23.9 percentage_points
Observed — percentage_points
downstream_accuracy
t8_dpo_13b_sd_sqa
Reported 8.2 percentage_points
Observed — percentage_points
downstream_accuracy
t8_dpo_13b_sd_mmlu
Reported 47.3 percentage_points
Observed — percentage_points
downstream_accuracy
t8_dpo_13b_sd_mmlu_p
Reported 17.9 percentage_points
Observed — percentage_points
downstream_accuracy
t8_dpo_13b_sd_arc_c
Reported 56.1 percentage_points
Observed — percentage_points
downstream_accuracy
t8_dpo_13b_sd_obqa
Reported 59.2 percentage_points
Observed — percentage_points
downstream_accuracy
t8_dpo_13b_sd_wino
Reported 52.5 percentage_points
Observed — percentage_points
downstream_accuracy
t8_dpo_13b_sd_agi
Reported 38.7 percentage_points
Observed — percentage_points
downstream_accuracy
t8_dpo_13b_sd_ife
Reported 61 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.