Downstream task evaluation after post-training stage DPO for 13b_trkd.
Source: source_paper:Table 8, Page 33, After DPO
Reported and observed measurements
downstream_accuracy
t8_dpo_13b_trkd_gsm8k
Reported 53.1 percentage_points
Observed — percentage_points
downstream_accuracy
t8_dpo_13b_trkd_gsm_s
Reported 34.9 percentage_points
Observed — percentage_points
downstream_accuracy
t8_dpo_13b_trkd_gsmplus
Reported 30.7 percentage_points
Observed — percentage_points
downstream_accuracy
t8_dpo_13b_trkd_bbh
Reported 32.9 percentage_points
Observed — percentage_points
downstream_accuracy
t8_dpo_13b_trkd_drop
Reported 35.3 percentage_points
Observed — percentage_points
downstream_accuracy
t8_dpo_13b_trkd_math
Reported 6.8 percentage_points
Observed — percentage_points
downstream_accuracy
t8_dpo_13b_trkd_tqa
Reported 54.9 percentage_points
Observed — percentage_points
downstream_accuracy
t8_dpo_13b_trkd_nq
Reported 22.4 percentage_points
Observed — percentage_points
downstream_accuracy
t8_dpo_13b_trkd_sqa
Reported 8.5 percentage_points
Observed — percentage_points
downstream_accuracy
t8_dpo_13b_trkd_mmlu
Reported 43.4 percentage_points
Observed — percentage_points
downstream_accuracy
t8_dpo_13b_trkd_mmlu_p
Reported 16.5 percentage_points
Observed — percentage_points
downstream_accuracy
t8_dpo_13b_trkd_arc_c
Reported 53.2 percentage_points
Observed — percentage_points
downstream_accuracy
t8_dpo_13b_trkd_obqa
Reported 56 percentage_points
Observed — percentage_points
downstream_accuracy
t8_dpo_13b_trkd_wino
Reported 52 percentage_points
Observed — percentage_points
downstream_accuracy
t8_dpo_13b_trkd_agi
Reported 35.3 percentage_points
Observed — percentage_points
downstream_accuracy
t8_dpo_13b_trkd_ife
Reported 61.7 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.