{ "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": 1.1685381217592508, "mse_channels": [ 1.1685381217592508 ], "mse_channel_mean": 1.1685381217592508, "conditional_kl": 90.21964607032946, "m30_total": 20.381005457584397 }, "similarity_shift": { "mse": 1.820171062855486, "mse_channels": [ 1.820171062855486 ], "mse_channel_mean": 1.820171062855486, "conditional_kl": 163.65696765471284, "m30_total": 36.37173565665354 }, "random": { "mse": { "mean": 2.0001020431518555, "std": 0.01599620096385479, "min": 1.9564861059188843, "true_z": 51.98508834838867, "true_rank_among_random": 0, "count": 300 }, "mse_channel_mean": { "mean": 2.0001020431518555, "std": 0.01599620096385479, "min": 1.9564861059188843, "true_z": 51.98508834838867, "true_rank_among_random": 0, "count": 300 }, "conditional_kl": { "mean": 168.61756896972656, "std": 1.9571372270584106, "min": 162.2455291748047, "true_z": 40.05744934082031, "true_rank_among_random": 0, "count": 300 }, "m30_total": { "mean": 37.723716735839844, "std": 0.39688318967819214, "min": 36.460758209228516, "true_z": 43.697265625, "true_rank_among_random": 0, "count": 300 } } }, "gate_b": { "pairs": 130816, "improving_pairs": 952, "improving_fraction": 0.007277397260273972, "delta_mean": 0.005820310517985544, "delta_min": -0.004950545727676812, "identity_is_local_min_mse": false, "worst_improving_swaps": [ { "pair": [ 118, 285 ], "delta_mse": -0.004950545727676812, "text_cosine": -0.3778797085142215, "visual_cosine": 0.5629043970742635 }, { "pair": [ 22, 118 ], "delta_mse": -0.004110298358545636, "text_cosine": 1.4886701098260828, "visual_cosine": 1.1755596888665971 }, { "pair": [ 275, 285 ], "delta_mse": -0.003997771726107714, "text_cosine": 0.6431330501638224, "visual_cosine": -0.35426136562362853 }, { "pair": [ 285, 328 ], "delta_mse": -0.003920223173239507, "text_cosine": 0.622401316067121, "visual_cosine": 0.020093502573850473 }, { "pair": [ 0, 285 ], "delta_mse": -0.003881395322844788, "text_cosine": 0.622401316067121, "visual_cosine": 1.1398397353817846 }, { "pair": [ 328, 451 ], "delta_mse": -0.0037328074890283936, "text_cosine": 0.8356371118132723, "visual_cosine": 1.0445104350149643 }, { "pair": [ 379, 423 ], "delta_mse": -0.003638480926693669, "text_cosine": 0.7564135468895015, "visual_cosine": 2.8252193791341633 }, { "pair": [ 0, 371 ], "delta_mse": -0.003422593628962994, "text_cosine": 1.4886701098260828, "visual_cosine": -0.2552556649452302 }, { "pair": [ 328, 511 ], "delta_mse": -0.0034050755307425302, "text_cosine": 2.155523988523928, "visual_cosine": 1.8346337396256396 }, { "pair": [ 118, 289 ], "delta_mse": -0.003323807321889081, "text_cosine": 1.4886701098260828, "visual_cosine": -0.4707792267404456 }, { "pair": [ 28, 142 ], "delta_mse": -0.003287436419471128, "text_cosine": 1.3546576724691275, "visual_cosine": 0.5281095508573972 }, { "pair": [ 64, 328 ], "delta_mse": -0.0032450945384208413, "text_cosine": 2.086914441940284, "visual_cosine": 0.5220891449876855 }, { "pair": [ 116, 328 ], "delta_mse": -0.0032181367581151105, "text_cosine": 1.3993286226111625, "visual_cosine": 0.04304217000896547 }, { "pair": [ 28, 285 ], "delta_mse": -0.003169727206331315, "text_cosine": -0.3778797085142215, "visual_cosine": 0.4854524943009991 }, { "pair": [ 28, 286 ], "delta_mse": -0.0031311470178413295, "text_cosine": 0.35437685442236, "visual_cosine": -0.07060785857721866 }, { "pair": [ 28, 371 ], "delta_mse": -0.0031263730376454592, "text_cosine": 0.35437685442236, "visual_cosine": 0.5445452228385695 }, { "pair": [ 117, 118 ], "delta_mse": -0.003097616312192148, "text_cosine": 2.488951340942, "visual_cosine": 2.1951732721424535 }, { "pair": [ 64, 246 ], "delta_mse": -0.0030759089883498993, "text_cosine": 1.220645648181322, "visual_cosine": 1.2374709909763775 }, { "pair": [ 118, 147 ], "delta_mse": -0.0030357080498330073, "text_cosine": 0.35437685442236, "visual_cosine": 0.4836881825828448 }, { "pair": [ 116, 379 ], "delta_mse": -0.00303135304348194, "text_cosine": 0.15496193936124306, "visual_cosine": 2.326468820786202 } ] }, "descent_from_true": { "start_accuracy": 1.0, "final_accuracy": 0.373046875, "final_objective": 1.0257766597059934, "final_energy": { "mse": 1.0257766597059934, "mse_channels": [ 1.0257766597059934 ], "mse_channel_mean": 1.0257766597059934, "conditional_kl": 97.69344305411153, "m30_total": 21.590241930234292 }, "accepted_moves": 336, "moved_fraction": 0.626953125, "mse_local_min_certificate": true, "trajectory": [ { "step": 0, "objective": 1.163587576031574, "accuracy": 0.99609375 }, { "step": 50, "objective": 1.090487209832174, "accuracy": 0.818359375 }, { "step": 100, "objective": 1.0715917664809553, "accuracy": 0.689453125 }, { "step": 150, "objective": 1.061964304088665, "accuracy": 0.58203125 }, { "step": 200, "objective": 1.0488727695218916, "accuracy": 0.515625 }, { "step": 250, "objective": 1.0370732598148131, "accuracy": 0.4765625 }, { "step": 300, "objective": 1.0281099129801385, "accuracy": 0.408203125 } ] }, "descent_from_random": [ { "start_accuracy": 0.00390625, "final_accuracy": 0.0078125, "final_objective": 1.046290397297093, "final_energy": { "mse": 1.046290397297093, "mse_channels": [ 1.046290397297093 ], "mse_channel_mean": 1.046290397297093, "conditional_kl": 104.96701714460303, "m30_total": 23.085984223514792 }, "accepted_moves": 734, "moved_fraction": 0.984375, "mse_local_min_certificate": true }, { "start_accuracy": 0.0, "final_accuracy": 0.001953125, "final_objective": 1.0457940736685236, "final_energy": { "mse": 1.0457940736685236, "mse_channels": [ 1.0457940736685236 ], "mse_channel_mean": 1.0457940736685236, "conditional_kl": 105.09482191757168, "m30_total": 23.110552530851386 }, "accepted_moves": 792, "moved_fraction": 0.986328125, "mse_local_min_certificate": true }, { "start_accuracy": 0.001953125, "final_accuracy": 0.00390625, "final_objective": 1.0408587960800706, "final_energy": { "mse": 1.0408587960800706, "mse_channels": [ 1.0408587960800706 ], "mse_channel_mean": 1.0408587960800706, "conditional_kl": 107.9008315007676, "m30_total": 23.661883892313664 }, "accepted_moves": 818, "moved_fraction": 0.99609375, "mse_local_min_certificate": true }, { "start_accuracy": 0.0, "final_accuracy": 0.0078125, "final_objective": 1.0480423677324506, "final_energy": { "mse": 1.0480423677324506, "mse_channels": [ 1.0480423677324506 ], "mse_channel_mean": 1.0480423677324506, "conditional_kl": 104.89994002511857, "m30_total": 23.076072740488616 }, "accepted_moves": 803, "moved_fraction": 0.986328125, "mse_local_min_certificate": true }, { "start_accuracy": 0.005859375, "final_accuracy": 0.005859375, "final_objective": 1.0604355486128925, "final_energy": { "mse": 1.0604355486128925, "mse_channels": [ 1.0604355486128925 ], "mse_channel_mean": 1.0604355486128925, "conditional_kl": 106.12575109921288, "m30_total": 23.34602131706836 }, "accepted_moves": 834, "moved_fraction": 0.986328125, "mse_local_min_certificate": true } ], "verdict": { "true_z_mse": 51.98508834838867, "improving_fraction": 0.007277397260273972, "descent_keeps": 0.373046875, "true_mse": 1.1685381217592508, "best_random_descent": 1.0408587960800706, "counterfeit_found": true } }