Per-task downstream accuracy for mid-training ablation oracle_domain using OLMo-2 7B Instruct teacher.
Source: source_paper:Table 9, Page 33
Reported and observed measurements
downstream_accuracy
t9_oracle_domain_gsm8k
Reported 61 percentage_points
Observed — percentage_points
downstream_accuracy
t9_oracle_domain_gsm_s
Reported 45.6 percentage_points
Observed — percentage_points
downstream_accuracy
t9_oracle_domain_gsmplus
Reported 38.5 percentage_points
Observed — percentage_points
downstream_accuracy
t9_oracle_domain_bbh
Reported 31.6 percentage_points
Observed — percentage_points
downstream_accuracy
t9_oracle_domain_drop
Reported 40.1 percentage_points
Observed — percentage_points
downstream_accuracy
t9_oracle_domain_math
Reported 8 percentage_points
Observed — percentage_points
downstream_accuracy
t9_oracle_domain_tqa
Reported 53 percentage_points
Observed — percentage_points
downstream_accuracy
t9_oracle_domain_nq
Reported 22.6 percentage_points
Observed — percentage_points
downstream_accuracy
t9_oracle_domain_sqa
Reported 8.3 percentage_points
Observed — percentage_points
downstream_accuracy
t9_oracle_domain_mmlu
Reported 49.4 percentage_points
Observed — percentage_points
downstream_accuracy
t9_oracle_domain_mmlu_p
Reported 18.8 percentage_points
Observed — percentage_points
downstream_accuracy
t9_oracle_domain_arc_c
Reported 61.1 percentage_points
Observed — percentage_points
downstream_accuracy
t9_oracle_domain_obqa
Reported 61.4 percentage_points
Observed — percentage_points
downstream_accuracy
t9_oracle_domain_wino
Reported 52.6 percentage_points
Observed — percentage_points
downstream_accuracy
t9_oracle_domain_agi
Reported 38.9 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.