{ "experiment": "R-MATH-INK-06-P-PROXY-SHIFT-001", "generated_at": "2026-07-23T21:35:33.958856+00:00", "seed": 17, "reference": "CROHME writer-validation truth groups", "candidate": "approved P isolated trajectories with synthetic row layout", "validation_role_counts": { "identifier_lower": 401, "multiply_operator": 44, "identifier_upper": 22 }, "proxy": { "samples": 1590, "role_counts": { "identifier_upper": 781, "identifier_lower": 809 }, "ambiguous_size_samples": 389, "sources": { "hwrt": 438, "uci-uji-pen-v1": 256, "uci-uji-pen-v2": 896 }, "track": "P_approved_synthetic_layout_proxy", "actual_continuous_formula": false, "target_bases": [ "c", "o", "s", "u", "v", "w", "x", "z" ], "lowercase_ratio": 1.0 }, "role_shift": { "identifier_lower": { "reference_samples": 401, "candidate_samples": 809, "features": { "teacher_family_mass": { "reference_samples": 401, "candidate_samples": 809, "reference_mean": 0.6060622334480286, "candidate_mean": 0.8209525942802429, "reference_std": 0.35282278060913086, "candidate_std": 0.2961852550506592, "standardized_mean_difference": 0.6597044282657114, "mean_absolute_quantile_distance": 0.21285369992256165 }, "teacher_top1": { "reference_samples": 401, "candidate_samples": 809, "reference_mean": 0.6049194931983948, "candidate_mean": 0.726399838924408, "reference_std": 0.21185898780822754, "candidate_std": 0.16862668097019196, "standardized_mean_difference": 0.634471608631743, "mean_absolute_quantile_distance": 0.12024392187595367 }, "teacher_entropy": { "reference_samples": 401, "candidate_samples": 809, "reference_mean": 0.21543195843696594, "candidate_mean": 0.11704465746879578, "reference_std": 0.11910299956798553, "candidate_std": 0.06580780446529388, "standardized_mean_difference": -1.0225350682468304, "mean_absolute_quantile_distance": 0.097715824842453 }, "bbox_width": { "reference_samples": 401, "candidate_samples": 809, "reference_mean": 0.06869771331548691, "candidate_mean": 0.2510845959186554, "reference_std": 0.03588062524795532, "candidate_std": 0.0840773656964302, "standardized_mean_difference": 2.8216191770723147, "mean_absolute_quantile_distance": 0.18182258307933807 }, "bbox_height": { "reference_samples": 401, "candidate_samples": 809, "reference_mean": 0.3474024832248688, "candidate_mean": 0.7116909027099609, "reference_std": 0.1750456988811493, "candidate_std": 0.09534971415996552, "standardized_mean_difference": 2.584562648818895, "mean_absolute_quantile_distance": 0.3640669286251068 }, "local_height_ratio": { "reference_samples": 401, "candidate_samples": 809, "reference_mean": 1.139341950416565, "candidate_mean": 0.741969108581543, "reference_std": 0.4629884362220764, "candidate_std": 0.10287872701883316, "standardized_mean_difference": -1.184888789842866, "mean_absolute_quantile_distance": 0.42106547951698303 }, "local_width_ratio": { "reference_samples": 401, "candidate_samples": 809, "reference_mean": 1.2371164560317993, "candidate_mean": 1.0617985725402832, "reference_std": 0.44012874364852905, "candidate_std": 0.3945411443710327, "standardized_mean_difference": -0.41946397378970135, "mean_absolute_quantile_distance": 0.17371578514575958 }, "left_gap": { "reference_samples": 401, "candidate_samples": 809, "reference_mean": 1.1788383722305298, "candidate_mean": 0.33475786447525024, "reference_std": 1.080208420753479, "candidate_std": 0.12264561653137207, "standardized_mean_difference": -1.0980192135440987, "mean_absolute_quantile_distance": 0.8872766494750977 }, "right_gap": { "reference_samples": 401, "candidate_samples": 809, "reference_mean": 1.0703167915344238, "candidate_mean": 0.33759987354278564, "reference_std": 1.267625331878662, "candidate_std": 0.12626643478870392, "standardized_mean_difference": -0.8134229411954296, "mean_absolute_quantile_distance": 0.8366667628288269 } }, "largest_shifts": [ { "feature": "bbox_width", "absolute_standardized_mean_difference": 2.8216191770723147 }, { "feature": "bbox_height", "absolute_standardized_mean_difference": 2.584562648818895 }, { "feature": "local_height_ratio", "absolute_standardized_mean_difference": 1.184888789842866 }, { "feature": "left_gap", "absolute_standardized_mean_difference": 1.0980192135440987 }, { "feature": "teacher_entropy", "absolute_standardized_mean_difference": 1.0225350682468304 } ] }, "identifier_upper": { "reference_samples": 22, "candidate_samples": 781, "features": { "teacher_family_mass": { "reference_samples": 22, "candidate_samples": 781, "reference_mean": 0.8246449828147888, "candidate_mean": 0.803795337677002, "reference_std": 0.31858986616134644, "candidate_std": 0.31461936235427856, "standardized_mean_difference": -0.06585259444348719, "mean_absolute_quantile_distance": 0.03567245602607727 }, "teacher_top1": { "reference_samples": 22, "candidate_samples": 781, "reference_mean": 0.7868841290473938, "candidate_mean": 0.8251377940177917, "reference_std": 0.1799909770488739, "candidate_std": 0.1476837545633316, "standardized_mean_difference": 0.23235894038924026, "mean_absolute_quantile_distance": 0.050301019102334976 }, "teacher_entropy": { "reference_samples": 22, "candidate_samples": 781, "reference_mean": 0.0961635410785675, "candidate_mean": 0.08980369567871094, "reference_std": 0.06707010418176651, "candidate_std": 0.05952510982751846, "standardized_mean_difference": -0.10029731532573508, "mean_absolute_quantile_distance": 0.018942177295684814 }, "bbox_width": { "reference_samples": 22, "candidate_samples": 781, "reference_mean": 0.08788423985242844, "candidate_mean": 0.21138401329517365, "reference_std": 0.02424454316496849, "candidate_std": 0.06387500464916229, "standardized_mean_difference": 2.5563739498452787, "mean_absolute_quantile_distance": 0.12389104068279266 }, "bbox_height": { "reference_samples": 22, "candidate_samples": 781, "reference_mean": 0.4731973707675934, "candidate_mean": 0.9367628693580627, "reference_std": 0.2780466675758362, "candidate_std": 0.07178061455488205, "standardized_mean_difference": 2.2829589500284175, "mean_absolute_quantile_distance": 0.4673648476600647 }, "local_height_ratio": { "reference_samples": 22, "candidate_samples": 781, "reference_mean": 1.4559041261672974, "candidate_mean": 0.9755271077156067, "reference_std": 0.544471025466919, "candidate_std": 0.07961194962263107, "standardized_mean_difference": -1.2346071269543992, "mean_absolute_quantile_distance": 0.4669833183288574 }, "local_width_ratio": { "reference_samples": 22, "candidate_samples": 781, "reference_mean": 1.4032069444656372, "candidate_mean": 0.8815558552742004, "reference_std": 0.6612966656684875, "candidate_std": 0.2695997357368469, "standardized_mean_difference": -1.0330258254896256, "mean_absolute_quantile_distance": 0.4956299662590027 }, "left_gap": { "reference_samples": 22, "candidate_samples": 781, "reference_mean": 0.5899125337600708, "candidate_mean": 0.33634957671165466, "reference_std": 0.3308897912502289, "candidate_std": 0.12277354300022125, "standardized_mean_difference": -1.016035941005507, "mean_absolute_quantile_distance": 0.28364264965057373 }, "right_gap": { "reference_samples": 22, "candidate_samples": 781, "reference_mean": 1.9250835180282593, "candidate_mean": 0.3313145339488983, "reference_std": 0.34331008791923523, "candidate_std": 0.12159264832735062, "standardized_mean_difference": -6.188598761520088, "mean_absolute_quantile_distance": 1.6213083267211914 } }, "largest_shifts": [ { "feature": "right_gap", "absolute_standardized_mean_difference": 6.188598761520088 }, { "feature": "bbox_width", "absolute_standardized_mean_difference": 2.5563739498452787 }, { "feature": "bbox_height", "absolute_standardized_mean_difference": 2.2829589500284175 }, { "feature": "local_height_ratio", "absolute_standardized_mean_difference": 1.2346071269543992 }, { "feature": "local_width_ratio", "absolute_standardized_mean_difference": 1.0330258254896256 } ] } }, "maximum_layout_absolute_smd": 2.8216191770723147, "decision": { "layout_distribution_compatible": false, "threshold_absolute_smd": 0.5, "expand_to_three_seeds": false, "product_validation": false }, "track": "diagnostic_R_reference_plus_P_proxy", "product_validation": false }