Downstream task evaluation after post-training stage SFT for 7b_sd.
Source: source_paper:Table 8, Page 33, After SFT
Reported and observed measurements
downstream_accuracy
t8_sft_7b_sd_gsm8k
Reported 63.7 percentage_points
Observed — percentage_points
downstream_accuracy
t8_sft_7b_sd_gsm_s
Reported 42.7 percentage_points
Observed — percentage_points
downstream_accuracy
t8_sft_7b_sd_gsmplus
Reported 36.7 percentage_points
Observed — percentage_points
downstream_accuracy
t8_sft_7b_sd_bbh
Reported 33.5 percentage_points
Observed — percentage_points
downstream_accuracy
t8_sft_7b_sd_drop
Reported 48.9 percentage_points
Observed — percentage_points
downstream_accuracy
t8_sft_7b_sd_math
Reported 12.8 percentage_points
Observed — percentage_points
downstream_accuracy
t8_sft_7b_sd_tqa
Reported 55.1 percentage_points
Observed — percentage_points
downstream_accuracy
t8_sft_7b_sd_nq
Reported 24.5 percentage_points
Observed — percentage_points
downstream_accuracy
t8_sft_7b_sd_sqa
Reported 8.5 percentage_points
Observed — percentage_points
downstream_accuracy
t8_sft_7b_sd_mmlu
Reported 50.3 percentage_points
Observed — percentage_points
downstream_accuracy
t8_sft_7b_sd_mmlu_p
Reported 19.4 percentage_points
Observed — percentage_points
downstream_accuracy
t8_sft_7b_sd_arc_c
Reported 61.3 percentage_points
Observed — percentage_points
downstream_accuracy
t8_sft_7b_sd_obqa
Reported 61.4 percentage_points
Observed — percentage_points
downstream_accuracy
t8_sft_7b_sd_wino
Reported 56.1 percentage_points
Observed — percentage_points
downstream_accuracy
t8_sft_7b_sd_agi
Reported 40.6 percentage_points
Observed — percentage_points
downstream_accuracy
t8_sft_7b_sd_ife
Reported 45.5 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.