{ "protocol": "Vision sets are model-free connected-component sprites (declared minimal world knowledge); text sets are per-group phrase states. Hidden pairs score orderings only.", "split": "test", "samples": 512, "mean_vision_set_size": 2.59375, "gate_a": { "true": { "mse": 1.0645479078007105, "mse_channels": [ 1.0645479078007105 ], "mse_channel_mean": 1.0645479078007105, "conditional_kl": 83.16008245981592, "m30_total": 18.761112307564606 }, "similarity_shift": { "mse": 1.7999073443357338, "mse_channels": [ 1.7999073443357338 ], "mse_channel_mean": 1.7999073443357338, "conditional_kl": 166.8408662120983, "m30_total": 36.96798793109112 }, "random": { "mse": { "mean": 2.0003468990325928, "std": 0.015627698972821236, "min": 1.9611443281173706, "true_z": 59.880794525146484, "true_rank_among_random": 0, "count": 300 }, "mse_channel_mean": { "mean": 2.0003468990325928, "std": 0.015627698972821236, "min": 1.9611443281173706, "true_z": 59.880794525146484, "true_rank_among_random": 0, "count": 300 }, "conditional_kl": { "mean": 169.95213317871094, "std": 2.3629813194274902, "min": 163.24522399902344, "true_z": 36.72989273071289, "true_rank_among_random": 0, "count": 300 }, "m30_total": { "mean": 37.991119384765625, "std": 0.47560393810272217, "min": 36.62797927856445, "true_z": 40.43281555175781, "true_rank_among_random": 0, "count": 300 } } }, "gate_b": { "pairs": 130816, "improving_pairs": 167, "improving_fraction": 0.0012766022504892367, "delta_mean": 0.006539553594596856, "delta_min": -0.0015065925098377733, "identity_is_local_min_mse": false, "worst_improving_swaps": [ { "pair": [ 124, 385 ], "delta_mse": -0.0015065925098377733, "text_cosine": 0.7564135468895015, "visual_cosine": 2.246878920343339 }, { "pair": [ 77, 440 ], "delta_mse": -0.0014186607433094318, "text_cosine": 0.4883890852447405, "visual_cosine": 2.984984633484502 }, { "pair": [ 68, 385 ], "delta_mse": -0.001391416543140343, "text_cosine": -0.3778797085142215, "visual_cosine": 2.61311960069059 }, { "pair": [ 108, 417 ], "delta_mse": -0.0012825281237332689, "text_cosine": -0.3778797085142215, "visual_cosine": 3.2643594089122288 }, { "pair": [ 101, 494 ], "delta_mse": -0.0011948795578246374, "text_cosine": 0.9111575118085833, "visual_cosine": 2.2291918030503415 }, { "pair": [ 108, 234 ], "delta_mse": -0.0010799808819813863, "text_cosine": 1.756694571470844, "visual_cosine": 1.297622499673109 }, { "pair": [ 56, 260 ], "delta_mse": -0.0010664745881310415, "text_cosine": 0.4883890852447405, "visual_cosine": 2.1197652933728626 }, { "pair": [ 260, 481 ], "delta_mse": -0.0010313083395566951, "text_cosine": 1.4886701098260828, "visual_cosine": 2.147599620414076 }, { "pair": [ 101, 388 ], "delta_mse": -0.0010304864979527825, "text_cosine": 0.06562045214632273, "visual_cosine": 2.565632317087539 }, { "pair": [ 168, 260 ], "delta_mse": -0.0010197932778717887, "text_cosine": -0.5118919393366019, "visual_cosine": 1.3875163270889872 }, { "pair": [ 481, 490 ], "delta_mse": -0.001008812479874572, "text_cosine": 2.0661829143781576, "visual_cosine": 3.0109978581266517 }, { "pair": [ 293, 481 ], "delta_mse": -0.0009861524532252704, "text_cosine": 2.488951340942, "visual_cosine": 3.0926043105951484 }, { "pair": [ 205, 307 ], "delta_mse": -0.0009481864816338404, "text_cosine": 0.4883890852447405, "visual_cosine": 2.8881114990757935 }, { "pair": [ 43, 211 ], "delta_mse": -0.0009097075572110447, "text_cosine": 1.220645648181322, "visual_cosine": 1.6912755277370144 }, { "pair": [ 39, 75 ], "delta_mse": -0.0008812790380505994, "text_cosine": -0.3778797085142215, "visual_cosine": -0.3587863294807634 }, { "pair": [ 108, 284 ], "delta_mse": -0.0008779164232221449, "text_cosine": 2.7569753895176112, "visual_cosine": 1.218323682931187 }, { "pair": [ 314, 417 ], "delta_mse": -0.0008603789379163277, "text_cosine": 0.4883890852447405, "visual_cosine": 1.4976617645046042 }, { "pair": [ 168, 207 ], "delta_mse": -0.0008595632241604462, "text_cosine": -0.5118919393366019, "visual_cosine": 1.093601272133213 }, { "pair": [ 205, 396 ], "delta_mse": -0.0007980022855767631, "text_cosine": -0.24386747769184097, "visual_cosine": 2.8737961392843467 }, { "pair": [ 87, 481 ], "delta_mse": -0.000781281400662361, "text_cosine": 1.4886701098260828, "visual_cosine": 1.780833674337543 } ] }, "descent_from_true": { "start_accuracy": 1.0, "final_accuracy": 0.689453125, "final_objective": 1.0316476013748346, "final_energy": { "mse": 1.0316476013748346, "mse_channels": [ 1.0316476013748346 ], "mse_channel_mean": 1.0316476013748346, "conditional_kl": 89.56823150514602, "m30_total": 19.97694150377887 }, "accepted_moves": 105, "moved_fraction": 0.310546875, "mse_local_min_certificate": true, "trajectory": [ { "step": 0, "objective": 1.0630413152908729, "accuracy": 0.99609375 }, { "step": 50, "objective": 1.036917262542022, "accuracy": 0.828125 }, { "step": 100, "objective": 1.0317017596029638, "accuracy": 0.6953125 } ] }, "descent_from_random": [ { "start_accuracy": 0.00390625, "final_accuracy": 0.00390625, "final_objective": 1.0369360654493502, "final_energy": { "mse": 1.0369360654493502, "mse_channels": [ 1.0369360654493502 ], "mse_channel_mean": 1.0369360654493502, "conditional_kl": 107.85022707689544, "m30_total": 23.643917546277788 }, "accepted_moves": 813, "moved_fraction": 0.986328125, "mse_local_min_certificate": true }, { "start_accuracy": 0.0, "final_accuracy": 0.005859375, "final_objective": 1.0348565130892038, "final_energy": { "mse": 1.0348565130892038, "mse_channels": [ 1.0348565130892038 ], "mse_channel_mean": 1.0348565130892038, "conditional_kl": 103.74262607392889, "m30_total": 22.818238240964188 }, "accepted_moves": 741, "moved_fraction": 0.990234375, "mse_local_min_certificate": true }, { "start_accuracy": 0.001953125, "final_accuracy": 0.00390625, "final_objective": 0.9921776112924858, "final_energy": { "mse": 0.9921776112924858, "mse_channels": [ 0.9921776112924858 ], "mse_channel_mean": 0.9921776112924858, "conditional_kl": 102.96093832764919, "m30_total": 22.57654288811481 }, "accepted_moves": 807, "moved_fraction": 0.99609375, "mse_local_min_certificate": true }, { "start_accuracy": 0.0, "final_accuracy": 0.001953125, "final_objective": 1.0080896835981785, "final_energy": { "mse": 1.0080896835981785, "mse_channels": [ 1.0080896835981785 ], "mse_channel_mean": 1.0080896835981785, "conditional_kl": 103.27214949469742, "m30_total": 22.67060926613584 }, "accepted_moves": 842, "moved_fraction": 0.998046875, "mse_local_min_certificate": true }, { "start_accuracy": 0.005859375, "final_accuracy": 0.001953125, "final_objective": 1.0259326022810817, "final_energy": { "mse": 1.0259326022810817, "mse_channels": [ 1.0259326022810817 ], "mse_channel_mean": 1.0259326022810817, "conditional_kl": 104.45691373996016, "m30_total": 22.943247952554195 }, "accepted_moves": 854, "moved_fraction": 0.998046875, "mse_local_min_certificate": true } ], "verdict": { "true_z_mse": 59.880794525146484, "improving_fraction": 0.0012766022504892367, "descent_keeps": 0.689453125, "true_mse": 1.0645479078007105, "best_random_descent": 0.9921776112924858, "counterfeit_found": true } }