Full downstream results after standard 4-stage post-training pipeline (SFT, DPO, RLVR1, RLVR2) for 13B TRKD post-training performance.
Source: source_paper:Table 2, Page 9
Reported and observed measurements
downstream_accuracy
t2_13b_trkd_gsm8k
Reported 69.6 percentage_points
Observed — percentage_points
downstream_accuracy
t2_13b_trkd_gsm_s
Reported 48.5 percentage_points
Observed — percentage_points
downstream_accuracy
t2_13b_trkd_gsmplus
Reported 42.6 percentage_points
Observed — percentage_points
downstream_accuracy
t2_13b_trkd_bbh
Reported 31.5 percentage_points
Observed — percentage_points
downstream_accuracy
t2_13b_trkd_drop
Reported 34.6 percentage_points
Observed — percentage_points
downstream_accuracy
t2_13b_trkd_math
Reported 12.2 percentage_points
Observed — percentage_points
downstream_accuracy
t2_13b_trkd_tqa
Reported 53.9 percentage_points
Observed — percentage_points
downstream_accuracy
t2_13b_trkd_nq
Reported 22.2 percentage_points
Observed — percentage_points
downstream_accuracy
t2_13b_trkd_sqa
Reported 8.2 percentage_points
Observed — percentage_points
downstream_accuracy
t2_13b_trkd_mmlu
Reported 37 percentage_points
Observed — percentage_points
downstream_accuracy
t2_13b_trkd_mmlu_p
Reported 17.3 percentage_points
Observed — percentage_points
downstream_accuracy
t2_13b_trkd_arc_c
Reported 55.4 percentage_points
Observed — percentage_points
downstream_accuracy
t2_13b_trkd_obqa
Reported 54 percentage_points
Observed — percentage_points
downstream_accuracy
t2_13b_trkd_wino
Reported 51.3 percentage_points
Observed — percentage_points
downstream_accuracy
t2_13b_trkd_agi
Reported 33.6 percentage_points
Observed — percentage_points
downstream_accuracy
t2_13b_trkd_ife
Reported 64.1 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.