Full downstream results after standard 4-stage post-training pipeline (SFT, DPO, RLVR1, RLVR2) for NTP post-training baseline performance.
Source: source_paper:Table 2, Page 9
Reported and observed measurements
downstream_accuracy
t2_ntp_shared_gsm8k
Reported 67 percentage_points
Observed — percentage_points
downstream_accuracy
t2_ntp_shared_gsm_s
Reported 46.8 percentage_points
Observed — percentage_points
downstream_accuracy
t2_ntp_shared_gsmplus
Reported 40.6 percentage_points
Observed — percentage_points
downstream_accuracy
t2_ntp_shared_bbh
Reported 31.1 percentage_points
Observed — percentage_points
downstream_accuracy
t2_ntp_shared_drop
Reported 32.7 percentage_points
Observed — percentage_points
downstream_accuracy
t2_ntp_shared_math
Reported 12.6 percentage_points
Observed — percentage_points
downstream_accuracy
t2_ntp_shared_tqa
Reported 53.3 percentage_points
Observed — percentage_points
downstream_accuracy
t2_ntp_shared_nq
Reported 21.6 percentage_points
Observed — percentage_points
downstream_accuracy
t2_ntp_shared_sqa
Reported 7.5 percentage_points
Observed — percentage_points
downstream_accuracy
t2_ntp_shared_mmlu
Reported 32.3 percentage_points
Observed — percentage_points
downstream_accuracy
t2_ntp_shared_mmlu_p
Reported 16.6 percentage_points
Observed — percentage_points
downstream_accuracy
t2_ntp_shared_arc_c
Reported 52.4 percentage_points
Observed — percentage_points
downstream_accuracy
t2_ntp_shared_obqa
Reported 52.4 percentage_points
Observed — percentage_points
downstream_accuracy
t2_ntp_shared_wino
Reported 51.4 percentage_points
Observed — percentage_points
downstream_accuracy
t2_ntp_shared_agi
Reported 32.3 percentage_points
Observed — percentage_points
downstream_accuracy
t2_ntp_shared_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.