coded: {'grok': 144, 'sol': 144}  coder1(Kimi K3) rows: 144  unparsed: {'grok': [], 'sol': []}
agreement (overlap): {'n': 144, 'grok': 30, 'sol': 37, 'both': 27, 'agree': 0.91, 'kappa': 0.748}
pairwise (3-coder set): {'grok-sol': {'n': 144, 'grok': 30, 'sol': 37, 'both': 27, 'agree': 0.91, 'kappa': 0.748}}
coder1(Kimi K3) types: {'W-interaction': 13, 'W-shape': 10, 'Z-shape-uniform': 3, 'X-shape': 10, 'uniform-XZ-regime': 2}
two-coder types: {'W-interaction': 12, 'W-shape': 9, 'Z-shape-uniform': 3, 'X-shape': 9}
verdicts: {'grok': {'FALSE': 110, 'CORRECT': 320, 'SMALL': 10, 'TRUE': 170, 'TRUEPOSS': 27, 'UNRESOLVED': 31, 'AMBIG': 7, 'WRONGFORM': 5, 'INVALID': 5}, 'sol': {'FALSE': 163, 'SMALL': 12, 'CORRECT': 320, 'TRUEPOSS': 27, 'UNRESOLVED': 39, 'TRUE': 174, 'AMBIG': 4, 'WRONGFORM': 5, 'MISSED': 1}}
false subs: {'grok': {'X-shape': 26, 'Z-shape-uniform': 8, 'W-shape': 22, 'W-interaction': 51, 'uniform-XZ-regime': 3}, 'sol': {'X-shape': 36, 'W-interaction': 82, 'W-shape': 28, 'Z-shape-uniform': 14, 'uniform-XZ-regime': 3}}
all: KIMI-K3 nh 30/144 (20.8%, W [15.0, 28.2], cl [15.3, 26.4]); incl-hedged 45.1%; B1 log-stated 4/144 (2.8%, W [1.1, 6.9], cl [0.7, 5.6]); conflict 4.2%; overlap 0; B3 hit 94/96 (97.9%, W [92.7, 99.4], cl [94.8, 100.0]) FA 3/48 (6.2%, W [2.1, 16.8], cl [0.0, 12.5]); perclaim {'struct_claims': 353, 'false': 110, 'pct_false': 31.2} B2 {'parsed_claims': 266, 'contradicted': 4, 'pct': 1.5}
   2-CODER: prim 27/144 (18.8%, W [13.2, 25.9], cl [13.2, 24.3]); union 62.5%; B1 4/144 (2.8%, W [1.1, 6.9], cl [0.7, 5.6]); B3 hit 94/96 (97.9%, W [92.7, 99.4], cl [94.8, 100.0]) FA 3/48 (6.2%, W [2.1, 16.8], cl [0.0, 12.5]); perclaim coder1(Kimi K3) {'struct_claims': 353, 'false': 110, 'pct_false': 31.2} sol {'struct_claims': 420, 'false': 163, 'pct_false': 38.8} B2 {'parsed_claims': 266, 'contradicted': 4, 'pct': 1.5} B4 {'n': 238, 'false': 0, 'pct': 0.0}
sonnet: KIMI-K3 nh 17/72 (23.6%, W [15.3, 34.6], cl [15.3, 31.9]); incl-hedged 51.4%; B1 log-stated 0/72 (0.0%, W [0.0, 5.1], cl [0.0, 0.0]); conflict 1.4%; overlap 0; B3 hit 46/48 (95.8%, W [86.0, 98.8], cl [89.6, 100.0]) FA 2/24 (8.3%, W [2.3, 25.8], cl [0.0, 20.8]); perclaim {'struct_claims': 192, 'false': 69, 'pct_false': 35.9} B2 {'parsed_claims': 137, 'contradicted': 0, 'pct': 0.0}
   2-CODER: prim 15/72 (20.8%, W [13.1, 31.6], cl [12.5, 29.2]); union 68.1%; B1 0/72 (0.0%, W [0.0, 5.1], cl [0.0, 0.0]); B3 hit 46/48 (95.8%, W [86.0, 98.8], cl [89.6, 100.0]) FA 2/24 (8.3%, W [2.3, 25.8], cl [0.0, 20.8]); perclaim coder1(Kimi K3) {'struct_claims': 192, 'false': 69, 'pct_false': 35.9} sol {'struct_claims': 224, 'false': 100, 'pct_false': 44.6} B2 {'parsed_claims': 137, 'contradicted': 0, 'pct': 0.0} B4 {'n': 117, 'false': 0, 'pct': 0.0}
sonnet-A: KIMI-K3 nh 17/36 (47.2%, W [32.0, 63.0], cl [30.6, 63.9]); incl-hedged 75.0%; B1 log-stated 0/36 (0.0%, W [0.0, 9.6], cl [0.0, 0.0]); conflict 0.0%; overlap 0; B3 hit 22/24 (91.7%, W [74.2, 97.7], cl [79.2, 100.0]) FA 2/12 (16.7%, W [4.7, 44.8], cl [0.0, 41.7]); perclaim {'struct_claims': 124, 'false': 59, 'pct_false': 47.6} B2 {'parsed_claims': 71, 'contradicted': 0, 'pct': 0.0}
   2-CODER: prim 15/36 (41.7%, W [27.1, 57.8], cl [25.0, 58.3]); union 86.1%; B1 0/36 (0.0%, W [0.0, 9.6], cl [0.0, 0.0]); B3 hit 22/24 (91.7%, W [74.2, 97.7], cl [79.2, 100.0]) FA 2/12 (16.7%, W [4.7, 44.8], cl [0.0, 41.7]); perclaim coder1(Kimi K3) {'struct_claims': 124, 'false': 59, 'pct_false': 47.6} sol {'struct_claims': 138, 'false': 74, 'pct_false': 53.6} B2 {'parsed_claims': 71, 'contradicted': 0, 'pct': 0.0} B4 {'n': 56, 'false': 0, 'pct': 0.0}
sonnet-F: KIMI-K3 nh 0/36 (0.0%, W [0.0, 9.6], cl [0.0, 0.0]); incl-hedged 27.8%; B1 log-stated 0/36 (0.0%, W [0.0, 9.6], cl [0.0, 0.0]); conflict 2.8%; overlap 0; B3 hit 24/24 (100.0%, W [86.2, 100.0], cl [100.0, 100.0]) FA 0/12 (0.0%, W [0.0, 24.3], cl [0.0, 0.0]); perclaim {'struct_claims': 68, 'false': 10, 'pct_false': 14.7} B2 {'parsed_claims': 66, 'contradicted': 0, 'pct': 0.0}
   2-CODER: prim 0/36 (0.0%, W [0.0, 9.6], cl [0.0, 0.0]); union 50.0%; B1 0/36 (0.0%, W [0.0, 9.6], cl [0.0, 0.0]); B3 hit 24/24 (100.0%, W [86.2, 100.0], cl [100.0, 100.0]) FA 0/12 (0.0%, W [0.0, 24.3], cl [0.0, 0.0]); perclaim coder1(Kimi K3) {'struct_claims': 68, 'false': 10, 'pct_false': 14.7} sol {'struct_claims': 86, 'false': 26, 'pct_false': 30.2} B2 {'parsed_claims': 66, 'contradicted': 0, 'pct': 0.0} B4 {'n': 61, 'false': 0, 'pct': 0.0}
sonnet-uniform: KIMI-K3 nh 7/24 (29.2%, W [14.9, 49.2], cl [16.7, 41.7]); incl-hedged 66.7%; B1 log-stated 0/24 (0.0%, W [0.0, 13.8], cl [0.0, 0.0]); conflict 0.0%; overlap 0; B3 hit None FA 2/24 (8.3%, W [2.3, 25.8], cl [0.0, 20.8]); perclaim {'struct_claims': 38, 'false': 34, 'pct_false': 89.5} B2 {'parsed_claims': 35, 'contradicted': 0, 'pct': 0.0}
   2-CODER: prim 7/24 (29.2%, W [14.9, 49.2], cl [16.7, 41.7]); union 75.0%; B1 0/24 (0.0%, W [0.0, 13.8], cl [0.0, 0.0]); B3 hit None FA 2/24 (8.3%, W [2.3, 25.8], cl [0.0, 20.8]); perclaim coder1(Kimi K3) {'struct_claims': 38, 'false': 34, 'pct_false': 89.5} sol {'struct_claims': 50, 'false': 44, 'pct_false': 88.0} B2 {'parsed_claims': 35, 'contradicted': 0, 'pct': 0.0} B4 {'n': 33, 'false': 0, 'pct': 0.0}
sonnet-reversal: KIMI-K3 nh 4/24 (16.7%, W [6.7, 35.9], cl [4.2, 29.2]); incl-hedged 37.5%; B1 log-stated 0/24 (0.0%, W [0.0, 13.8], cl [0.0, 0.0]); conflict 0.0%; overlap 0; B3 hit 24/24 (100.0%, W [86.2, 100.0], cl [100.0, 100.0]) FA None; perclaim {'struct_claims': 75, 'false': 15, 'pct_false': 20.0} B2 {'parsed_claims': 50, 'contradicted': 0, 'pct': 0.0}
   2-CODER: prim 4/24 (16.7%, W [6.7, 35.9], cl [4.2, 29.2]); union 62.5%; B1 0/24 (0.0%, W [0.0, 13.8], cl [0.0, 0.0]); B3 hit 24/24 (100.0%, W [86.2, 100.0], cl [100.0, 100.0]) FA None; perclaim coder1(Kimi K3) {'struct_claims': 75, 'false': 15, 'pct_false': 20.0} sol {'struct_claims': 87, 'false': 28, 'pct_false': 32.2} B2 {'parsed_claims': 50, 'contradicted': 0, 'pct': 0.0} B4 {'n': 45, 'false': 0, 'pct': 0.0}
sonnet-vanish: KIMI-K3 nh 6/24 (25.0%, W [12.0, 44.9], cl [12.5, 37.5]); incl-hedged 50.0%; B1 log-stated 0/24 (0.0%, W [0.0, 13.8], cl [0.0, 0.0]); conflict 4.2%; overlap 0; B3 hit 22/24 (91.7%, W [74.2, 97.7], cl [79.2, 100.0]) FA None; perclaim {'struct_claims': 79, 'false': 20, 'pct_false': 25.3} B2 {'parsed_claims': 52, 'contradicted': 0, 'pct': 0.0}
   2-CODER: prim 4/24 (16.7%, W [6.7, 35.9], cl [4.2, 29.2]); union 66.7%; B1 0/24 (0.0%, W [0.0, 13.8], cl [0.0, 0.0]); B3 hit 22/24 (91.7%, W [74.2, 97.7], cl [79.2, 100.0]) FA None; perclaim coder1(Kimi K3) {'struct_claims': 79, 'false': 20, 'pct_false': 25.3} sol {'struct_claims': 87, 'false': 28, 'pct_false': 32.2} B2 {'parsed_claims': 52, 'contradicted': 0, 'pct': 0.0} B4 {'n': 39, 'false': 0, 'pct': 0.0}
opus: KIMI-K3 nh 13/72 (18.1%, W [10.9, 28.5], cl [9.7, 27.8]); incl-hedged 38.9%; B1 log-stated 4/72 (5.6%, W [2.2, 13.4], cl [1.4, 11.1]); conflict 6.9%; overlap 0; B3 hit 48/48 (100.0%, W [92.6, 100.0], cl [100.0, 100.0]) FA 1/24 (4.2%, W [0.7, 20.2], cl [0.0, 12.5]); perclaim {'struct_claims': 161, 'false': 41, 'pct_false': 25.5} B2 {'parsed_claims': 129, 'contradicted': 4, 'pct': 3.1}
   2-CODER: prim 12/72 (16.7%, W [9.8, 26.9], cl [8.3, 25.0]); union 56.9%; B1 4/72 (5.6%, W [2.2, 13.4], cl [1.4, 11.1]); B3 hit 48/48 (100.0%, W [92.6, 100.0], cl [100.0, 100.0]) FA 1/24 (4.2%, W [0.7, 20.2], cl [0.0, 12.5]); perclaim coder1(Kimi K3) {'struct_claims': 161, 'false': 41, 'pct_false': 25.5} sol {'struct_claims': 196, 'false': 63, 'pct_false': 32.1} B2 {'parsed_claims': 129, 'contradicted': 4, 'pct': 3.1} B4 {'n': 121, 'false': 0, 'pct': 0.0}
opus-A: KIMI-K3 nh 11/36 (30.6%, W [18.0, 46.9], cl [16.7, 44.4]); incl-hedged 55.6%; B1 log-stated 1/36 (2.8%, W [0.5, 14.2], cl [0.0, 8.3]); conflict 2.8%; overlap 0; B3 hit 24/24 (100.0%, W [86.2, 100.0], cl [100.0, 100.0]) FA 1/12 (8.3%, W [1.5, 35.4], cl [0.0, 25.0]); perclaim {'struct_claims': 88, 'false': 30, 'pct_false': 34.1} B2 {'parsed_claims': 65, 'contradicted': 1, 'pct': 1.5}
   2-CODER: prim 10/36 (27.8%, W [15.8, 44.0], cl [13.9, 41.7]); union 77.8%; B1 1/36 (2.8%, W [0.5, 14.2], cl [0.0, 8.3]); B3 hit 24/24 (100.0%, W [86.2, 100.0], cl [100.0, 100.0]) FA 1/12 (8.3%, W [1.5, 35.4], cl [0.0, 25.0]); perclaim coder1(Kimi K3) {'struct_claims': 88, 'false': 30, 'pct_false': 34.1} sol {'struct_claims': 108, 'false': 41, 'pct_false': 38.0} B2 {'parsed_claims': 65, 'contradicted': 1, 'pct': 1.5} B4 {'n': 62, 'false': 0, 'pct': 0.0}
opus-F: KIMI-K3 nh 2/36 (5.6%, W [1.5, 18.1], cl [0.0, 13.9]); incl-hedged 22.2%; B1 log-stated 3/36 (8.3%, W [2.9, 21.8], cl [0.0, 19.4]); conflict 11.1%; overlap 0; B3 hit 24/24 (100.0%, W [86.2, 100.0], cl [100.0, 100.0]) FA 0/12 (0.0%, W [0.0, 24.3], cl [0.0, 0.0]); perclaim {'struct_claims': 73, 'false': 11, 'pct_false': 15.1} B2 {'parsed_claims': 64, 'contradicted': 3, 'pct': 4.7}
   2-CODER: prim 2/36 (5.6%, W [1.5, 18.1], cl [0.0, 13.9]); union 36.1%; B1 3/36 (8.3%, W [2.9, 21.8], cl [0.0, 19.4]); B3 hit 24/24 (100.0%, W [86.2, 100.0], cl [100.0, 100.0]) FA 0/12 (0.0%, W [0.0, 24.3], cl [0.0, 0.0]); perclaim coder1(Kimi K3) {'struct_claims': 73, 'false': 11, 'pct_false': 15.1} sol {'struct_claims': 88, 'false': 22, 'pct_false': 25.0} B2 {'parsed_claims': 64, 'contradicted': 3, 'pct': 4.7} B4 {'n': 59, 'false': 0, 'pct': 0.0}
opus-uniform: KIMI-K3 nh 4/24 (16.7%, W [6.7, 35.9], cl [0.0, 37.5]); incl-hedged 45.8%; B1 log-stated 0/24 (0.0%, W [0.0, 13.8], cl [0.0, 0.0]); conflict 0.0%; overlap 0; B3 hit None FA 1/24 (4.2%, W [0.7, 20.2], cl [0.0, 12.5]); perclaim {'struct_claims': 26, 'false': 20, 'pct_false': 76.9} B2 {'parsed_claims': 28, 'contradicted': 0, 'pct': 0.0}
   2-CODER: prim 4/24 (16.7%, W [6.7, 35.9], cl [0.0, 37.5]); union 58.3%; B1 0/24 (0.0%, W [0.0, 13.8], cl [0.0, 0.0]); B3 hit None FA 1/24 (4.2%, W [0.7, 20.2], cl [0.0, 12.5]); perclaim coder1(Kimi K3) {'struct_claims': 26, 'false': 20, 'pct_false': 76.9} sol {'struct_claims': 37, 'false': 31, 'pct_false': 83.8} B2 {'parsed_claims': 28, 'contradicted': 0, 'pct': 0.0} B4 {'n': 28, 'false': 0, 'pct': 0.0}
opus-reversal: KIMI-K3 nh 5/24 (20.8%, W [9.2, 40.5], cl [8.3, 33.3]); incl-hedged 41.7%; B1 log-stated 1/24 (4.2%, W [0.7, 20.2], cl [0.0, 12.5]); conflict 8.3%; overlap 0; B3 hit 24/24 (100.0%, W [86.2, 100.0], cl [100.0, 100.0]) FA None; perclaim {'struct_claims': 75, 'false': 11, 'pct_false': 14.7} B2 {'parsed_claims': 51, 'contradicted': 1, 'pct': 2.0}
   2-CODER: prim 5/24 (20.8%, W [9.2, 40.5], cl [8.3, 33.3]); union 62.5%; B1 1/24 (4.2%, W [0.7, 20.2], cl [0.0, 12.5]); B3 hit 24/24 (100.0%, W [86.2, 100.0], cl [100.0, 100.0]) FA None; perclaim coder1(Kimi K3) {'struct_claims': 75, 'false': 11, 'pct_false': 14.7} sol {'struct_claims': 88, 'false': 18, 'pct_false': 20.5} B2 {'parsed_claims': 51, 'contradicted': 1, 'pct': 2.0} B4 {'n': 47, 'false': 0, 'pct': 0.0}
opus-vanish: KIMI-K3 nh 4/24 (16.7%, W [6.7, 35.9], cl [4.2, 29.2]); incl-hedged 29.2%; B1 log-stated 3/24 (12.5%, W [4.3, 31.0], cl [0.0, 25.0]); conflict 12.5%; overlap 0; B3 hit 24/24 (100.0%, W [86.2, 100.0], cl [100.0, 100.0]) FA None; perclaim {'struct_claims': 60, 'false': 10, 'pct_false': 16.7} B2 {'parsed_claims': 50, 'contradicted': 3, 'pct': 6.0}
   2-CODER: prim 3/24 (12.5%, W [4.3, 31.0], cl [0.0, 25.0]); union 50.0%; B1 3/24 (12.5%, W [4.3, 31.0], cl [0.0, 25.0]); B3 hit 24/24 (100.0%, W [86.2, 100.0], cl [100.0, 100.0]) FA None; perclaim coder1(Kimi K3) {'struct_claims': 60, 'false': 10, 'pct_false': 16.7} sol {'struct_claims': 71, 'false': 14, 'pct_false': 19.7} B2 {'parsed_claims': 50, 'contradicted': 3, 'pct': 6.0} B4 {'n': 46, 'false': 0, 'pct': 0.0}
PAIRED sonnet F vs A: {'pairs': 36, 'A_only': 17, 'arm_only': 0, 'mcnemar_p': 0.0, 'A_rate': 17, 'arm_rate': 0, 'struct_A': 124, 'struct_arm': 68, 'false_A': 59, 'false_arm': 10}
PAIRED opus F vs A: {'pairs': 36, 'A_only': 10, 'arm_only': 1, 'mcnemar_p': 0.0117, 'A_rate': 11, 'arm_rate': 2, 'struct_A': 88, 'struct_arm': 73, 'false_A': 30, 'false_arm': 11}
S3 {"pooled": {"PRIMARY_W_unhedged_false_two_coder": {"pairs": 72, "A": 16, "F": 2, "A_only": 15, "F_only": 1, "mcnemar_p": 0.0005, "rel_reduction": 0.875, "rel_reduction_cluster95": [0.68, 1.0]}, "W_unhedged_false_grok_only_PROVISIONAL": {"pairs": 72, "A": 18, "F": 2, "A_only": 17, "F_only": 1, "mcnemar_p": 0.0001, "rel_reduction": 0.889, "rel_reduction_cluster95": [0.714, 1.0]}, "frozen_rule_any_unhedged_false_two_coder": {"pairs": 72, "A": 25, "F": 2, "A_only": 24, "F_only": 1, "mcnemar_p": 0.0, "rel_reduction": 0.92, "rel_reduction_cluster95": [0.793, 1.0]}, "X_shape_unhedged_false_two_coder": {"pairs": 72, "A": 9, "F": 0, "A_only": 9, "F_only": 0, "mcnemar_p": 0.0039, "rel_reduction": 1.0, "rel_reduction_cluster95": [1.0, 1.0]}, "log_stated_contradicted_v2b": {"pairs": 72, "A": 1, "F": 3, "A_only": 1, "F_only": 3, "mcnemar_p": 0.625, "rel_reduction": -2.0, "rel_reduction_cluster95": [-5.0, 1.0]}, "log_conflict_v2b": {"pairs": 72, "A": 1, "F": 5, "A_only": 1, "F_only": 5, "mcnemar_p": 0.2188, "rel_reduction": -4.0, "rel_reduction_cluster95": [-9.0, 0.5]}, "KILL_TRUE_XZ_unhedged_two_coder_regime_tasks": {"pairs": 48, "A": 37, "F": 35, "A_only": 7, "F_only": 5, "mcnemar_p": 0.7744, "rel_reduction": 0.054, "rel_reduction_cluster95": [-0.152, 0.216]}, "TRUE_XZ_unhedged_grok_only_regime_PROVISIONAL": {"pairs": 48, "A": 42, "F": 45, "A_only": 3, "F_only": 6, "mcnemar_p": 0.5078, "rel_reduction": -0.071, "rel_reduction_cluster95": [-0.194, 0.022]}}, "sonnet": {"PRIMARY_W_unhedged_false_two_coder": {"pairs": 36, "A": 9, "F": 0, "A_only": 9, "F_only": 0, "mcnemar_p": 0.0039, "rel_reduction": 1.0, "rel_reduction_cluster95": [1.0, 1.0]}, "W_unhedged_false_grok_only_PROVISIONAL": {"pairs": 36, "A": 11, "F": 0, "A_only": 11, "F_only": 0, "mcnemar_p": 0.001, "rel_reduction": 1.0, "rel_reduction_cluster95": [1.0, 1.0]}, "frozen_rule_any_unhedged_false_two_coder": {"pairs": 36, "A": 15, "F": 0, "A_only": 15, "F_only": 0, "mcnemar_p": 0.0001, "rel_reduction": 1.0, "rel_reduction_cluster95": [1.0, 1.0]}, "X_shape_unhedged_false_two_coder": {"pairs": 36, "A": 7, "F": 0, "A_only": 7, "F_only": 0, "mcnemar_p": 0.0156, "rel_reduction": 1.0, "rel_reduction_cluster95": [1.0, 1.0]}, "log_stated_contradicted_v2b": {"pairs": 36, "A": 0, "F": 0, "A_only": 0, "F_only": 0, "mcnemar_p": 1.0, "rel_reduction": null, "rel_reduction_cluster95": null}, "log_conflict_v2b": {"pairs": 36, "A": 0, "F": 1, "A_only": 0, "F_only": 1, "mcnemar_p": 1.0, "rel_reduction": null, "rel_reduction_cluster95": null}, "KILL_TRUE_XZ_unhedged_two_coder_regime_tasks": {"pairs": 24, "A": 17, "F": 17, "A_only": 4, "F_only": 4, "mcnemar_p": 1.0, "rel_reduction": 0.0, "rel_reduction_cluster95": [-0.417, 0.286]}, "TRUE_XZ_unhedged_grok_only_regime_PROVISIONAL": {"pairs": 24, "A": 18, "F": 24, "A_only": 0, "F_only": 6, "mcnemar_p": 0.0312, "rel_reduction": -0.333, "rel_reduction_cluster95": [-0.846, -0.091]}}, "opus": {"PRIMARY_W_unhedged_false_two_coder": {"pairs": 36, "A": 7, "F": 2, "A_only": 6, "F_only": 1, "mcnemar_p": 0.125, "rel_reduction": 0.714, "rel_reduction_cluster95": [0.0, 1.0]}, "W_unhedged_false_grok_only_PROVISIONAL": {"pairs": 36, "A": 7, "F": 2, "A_only": 6, "F_only": 1, "mcnemar_p": 0.125, "rel_reduction": 0.714, "rel_reduction_cluster95": [0.0, 1.0]}, "frozen_rule_any_unhedged_false_two_coder": {"pairs": 36, "A": 10, "F": 2, "A_only": 9, "F_only": 1, "mcnemar_p": 0.0215, "rel_reduction": 0.8, "rel_reduction_cluster95": [0.429, 1.0]}, "X_shape_unhedged_false_two_coder": {"pairs": 36, "A": 2, "F": 0, "A_only": 2, "F_only": 0, "mcnemar_p": 0.5, "rel_reduction": 1.0, "rel_reduction_cluster95": [1.0, 1.0]}, "log_stated_contradicted_v2b": {"pairs": 36, "A": 1, "F": 3, "A_only": 1, "F_only": 3, "mcnemar_p": 0.625, "rel_reduction": -2.0, "rel_reduction_cluster95": [-5.0, 1.0]}, "log_conflict_v2b": {"pairs": 36, "A": 1, "F": 4, "A_only": 1, "F_only": 4, "mcnemar_p": 0.375, "rel_reduction": -3.0, "rel_reduction_cluster95": [-7.0, 0.5]}, "KILL_TRUE_XZ_unhedged_two_coder_regime_tasks": {"pairs": 24, "A": 20, "F": 18, "A_only": 3, "F_only": 1, "mcnemar_p": 0.625, "rel_reduction": 0.1, "rel_reduction_cluster95": [-0.1, 0.286]}, "TRUE_XZ_unhedged_grok_only_regime_PROVISIONAL": {"pairs": 24, "A": 24, "F": 21, "A_only": 3, "F_only": 0, "mcnemar_p": 0.25, "rel_reduction": 0.125, "rel_reduction_cluster95": [0.0, 0.25]}}, "kill_rule": {"primary_rel_reduction": 0.875, "true_structural_drop": 0.054, "cut_less_than_third": false, "drop_more_than_10pct": false, "NEGATIVE_RESULT_AND_STOP": false, "PENDING_no_two_coder_pairs": false}, "regime_hit_rate_by_arm": {"A": [37, 48], "F": [35, 48]}}
