{ "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": 1.0234778236916264, "mse_channels": [ 1.0234778236916264 ], "mse_channel_mean": 1.0234778236916264, "conditional_kl": 61.452794797661305, "m30_total": 14.337514606915516 }, "similarity_shift": { "mse": 1.8890656518537374, "mse_channels": [ 1.8890656518537374 ], "mse_channel_mean": 1.8890656518537374, "conditional_kl": 163.40293254243772, "m30_total": 36.45871781219502 }, "random": { "mse": { "mean": 1.9994356632232666, "std": 0.02549530379474163, "min": 1.9340183734893799, "true_z": 38.2799072265625, "true_rank_among_random": 0, "count": 300 }, "mse_channel_mean": { "mean": 1.9994356632232666, "std": 0.02549530379474163, "min": 1.9340183734893799, "true_z": 38.2799072265625, "true_rank_among_random": 0, "count": 300 }, "conditional_kl": { "mean": 162.80917358398438, "std": 2.4263155460357666, "min": 155.34698486328125, "true_z": 41.77378463745117, "true_rank_among_random": 0, "count": 300 }, "m30_total": { "mean": 36.56070327758789, "std": 0.4999410808086395, "min": 35.01566696166992, "true_z": 44.45161437988281, "true_rank_among_random": 0, "count": 300 } } }, "gate_b": { "pairs": 130816, "improving_pairs": 277, "improving_fraction": 0.002117477984344423, "delta_mean": 0.007344444874908608, "delta_min": -0.004573651561417804, "identity_is_local_min_mse": false, "worst_improving_swaps": [ { "pair": [ 76, 434 ], "delta_mse": -0.004573651561417804, "text_cosine": 0.22779805194149536, "visual_cosine": 0.437904948436021 }, { "pair": [ 64, 152 ], "delta_mse": -0.003568867892357639, "text_cosine": -0.2020814242968897, "visual_cosine": 0.7395285222772097 }, { "pair": [ 28, 158 ], "delta_mse": -0.003291008296808442, "text_cosine": 1.81501344963681, "visual_cosine": 1.1188605778667853 }, { "pair": [ 152, 348 ], "delta_mse": -0.0030259614049693863, "text_cosine": -0.3299953772713672, "visual_cosine": 0.9293350397802795 }, { "pair": [ 34, 434 ], "delta_mse": -0.0029870880418199457, "text_cosine": 0.4695723413541927, "visual_cosine": 0.08561089781840142 }, { "pair": [ 76, 417 ], "delta_mse": -0.002938015375904725, "text_cosine": 0.16892190186079242, "visual_cosine": -0.2141006289898463 }, { "pair": [ 351, 434 ], "delta_mse": -0.002816824743549316, "text_cosine": 0.4736752783113544, "visual_cosine": -0.2223358314441135 }, { "pair": [ 101, 425 ], "delta_mse": -0.0026902432578103986, "text_cosine": -0.14024252577365168, "visual_cosine": 0.4914744795589924 }, { "pair": [ 64, 265 ], "delta_mse": -0.002688631983166332, "text_cosine": 0.05783179430958286, "visual_cosine": 0.9278845059132398 }, { "pair": [ 12, 417 ], "delta_mse": -0.0026708115768626106, "text_cosine": -0.43885310809176603, "visual_cosine": -0.38894655852896093 }, { "pair": [ 358, 448 ], "delta_mse": -0.002603549458818946, "text_cosine": 1.881744439051865, "visual_cosine": -1.0678405909745763 }, { "pair": [ 348, 417 ], "delta_mse": -0.0025937574842425995, "text_cosine": 0.8786679993839674, "visual_cosine": 2.1060904826442735 }, { "pair": [ 254, 417 ], "delta_mse": -0.0025639006242814608, "text_cosine": 0.7635089594066743, "visual_cosine": -1.0790167933491146 }, { "pair": [ 404, 417 ], "delta_mse": -0.0025468484837988985, "text_cosine": 0.7635089594066743, "visual_cosine": -0.7047770422653113 }, { "pair": [ 321, 434 ], "delta_mse": -0.002506076350820478, "text_cosine": 0.11369066429120464, "visual_cosine": 1.2546953341585219 }, { "pair": [ 351, 417 ], "delta_mse": -0.0024631315092251997, "text_cosine": -0.7412090614289673, "visual_cosine": -0.7265300168020604 }, { "pair": [ 74, 417 ], "delta_mse": -0.002460720593981292, "text_cosine": -0.43885310809176603, "visual_cosine": 0.5522611101706392 }, { "pair": [ 358, 440 ], "delta_mse": -0.002458688028614438, "text_cosine": 1.920949175248557, "visual_cosine": -0.3837262920595943 }, { "pair": [ 334, 417 ], "delta_mse": -0.002377982827399627, "text_cosine": 0.9137843127526165, "visual_cosine": 0.9491351402589897 }, { "pair": [ 28, 152 ], "delta_mse": -0.002314779314826676, "text_cosine": -0.8038227034300927, "visual_cosine": -0.11000513757380417 } ] }, "descent_from_true": { "start_accuracy": 1.0, "final_accuracy": 0.7578125, "final_objective": 0.9412325438384209, "final_energy": { "mse": 0.9412325438384209, "mse_channels": [ 0.9412325438384209 ], "mse_channel_mean": 0.9412325438384209, "conditional_kl": 71.34334407156564, "m30_total": 16.151133901989972 }, "accepted_moves": 107, "moved_fraction": 0.2421875, "mse_local_min_certificate": true, "trajectory": [ { "step": 0, "objective": 1.0189041721302086, "accuracy": 0.99609375 }, { "step": 50, "objective": 0.9561541044364367, "accuracy": 0.849609375 }, { "step": 100, "objective": 0.9423329921888384, "accuracy": 0.76953125 } ] }, "descent_from_random": [ { "start_accuracy": 0.00390625, "final_accuracy": 0.0, "final_objective": 0.8308049734351112, "final_energy": { "mse": 0.8308049734351112, "mse_channels": [ 0.8308049734351112 ], "mse_channel_mean": 0.8308049734351112, "conditional_kl": 97.00303823192738, "m30_total": 21.0622175932557 }, "accepted_moves": 750, "moved_fraction": 0.98828125, "mse_local_min_certificate": true }, { "start_accuracy": 0.0, "final_accuracy": 0.001953125, "final_objective": 0.8469482682810084, "final_energy": { "mse": 0.8469482682810084, "mse_channels": [ 0.8469482682810084 ], "mse_channel_mean": 0.8469482682810084, "conditional_kl": 92.92078639378623, "m30_total": 20.278053815319264 }, "accepted_moves": 787, "moved_fraction": 0.98046875, "mse_local_min_certificate": true }, { "start_accuracy": 0.001953125, "final_accuracy": 0.001953125, "final_objective": 0.8738041951201377, "final_energy": { "mse": 0.8738041951201377, "mse_channels": [ 0.8738041951201377 ], "mse_channel_mean": 0.8738041951201377, "conditional_kl": 94.70207071997939, "m30_total": 20.688022534236154 }, "accepted_moves": 765, "moved_fraction": 0.984375, "mse_local_min_certificate": true }, { "start_accuracy": 0.0, "final_accuracy": 0.005859375, "final_objective": 0.8173758976213201, "final_energy": { "mse": 0.8173758976213201, "mse_channels": [ 0.8173758976213201 ], "mse_channel_mean": 0.8173758976213201, "conditional_kl": 94.02453522241312, "m30_total": 20.439658839725265 }, "accepted_moves": 768, "moved_fraction": 0.982421875, "mse_local_min_certificate": true }, { "start_accuracy": 0.005859375, "final_accuracy": 0.001953125, "final_objective": 0.8150717953409379, "final_energy": { "mse": 0.8150717953409379, "mse_channels": [ 0.8150717953409379 ], "mse_channel_mean": 0.8150717953409379, "conditional_kl": 91.35303557635255, "m30_total": 19.900750705952387 }, "accepted_moves": 766, "moved_fraction": 0.990234375, "mse_local_min_certificate": true } ], "verdict": { "true_z_mse": 38.2799072265625, "improving_fraction": 0.002117477984344423, "descent_keeps": 0.7578125, "true_mse": 1.0234778236916264, "best_random_descent": 0.8150717953409379, "counterfeit_found": true } }