arm_gap=0.1763by_cell={"L1/low":0.9833,"L2/low":0.9556,"L3/low":0.9028,"L4/low":0.9417,"L0/none":0.98…n_scored=108primary_p=0.0098curve_shape=lineardose_by_cell={"L1/high":[1,2,3,4],"L2/high":[3,4,5,6,7],"L3/high":[6,7,8,9,10,11],"L4/high":…primary_test={"note":"positive mean_diff means the high-overlap arm scored worse","mean_diff…bulk_probe_L4={"note":"L4 raises bulk in BOTH arms; registered separately so it cannot enter …by_difficulty={"easy":0.9333,"edge":0.7889,"hard":0.9156,"medium":0.8427}by_skill_pair={"calendar_create_event <- calendar_create_hold":3}primary_tiers=["medium","hard"]by_stack_level={"L0":0.9833,"L1":0.9266,"L2":0.8681,"L3":0.8132,"L4":0.8229}completion_mean=0.9136by_overlap_class={"low":0.9458,"high":0.7695,"none":0.9833}correctness_mean=0.8512curve_high_drops=[0.1291,0.0302,0.1314]n_solution_error=0by_failure_reason={"near_miss":13,"incomplete":5,"over_eager":4,"bad_arguments":2,"instruction_bl…curve_low_overlap={"L0":1,"L1":1,"L2":0.963,"L3":0.9259,"L4":0.9778}low_overlap_score=0.9458n_skipped_no_gold=0by_competitor_dose={"1":1,"2":0.8743,"3":0.7611,"4":0.775,"5":0.7222,"6":0.9333,"7":0.76,"8":0.741…curve_high_overlap={"L0":1,"L1":0.8709,"L2":0.8407,"L3":0.7093,"L4":0.6611}high_overlap_score=0.7695secondary_sign_test_L3={"note":"one-sided exact sign test on discordant pairs from the pre-registered …curve_high_by_difficulty={"easy":{"L0":1,"L1":0.8,"L2":0.8,"L3":0.8,"L4":1},"edge":{"L0":0.9,"L1":0.9,"L…n_failures_without_a_pair=38curve_low_overlap_all_tiers={"L0":0.9833,"L1":0.9833,"L2":0.9556,"L3":0.9028,"L4":0.9417}bled_by_instruction_strength={"blunt":19,"medium":13,"subtle":1}curve_high_overlap_all_tiers={"L0":0.9833,"L1":0.8698,"L2":0.7806,"L3":0.7236,"L4":0.7042}solution_errors_by_stack_level={}score_excluding_solution_errors=0.8717
Fields here are the task author's — trap neither interprets nor validates them.