{ "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": 0.9300114277989874, "mse_channels": [ 0.9300114277989874 ], "mse_channel_mean": 0.9300114277989874, "conditional_kl": 60.03192832650445, "m30_total": 13.866408520898865 }, "similarity_shift": { "mse": 1.8300038997859625, "mse_channels": [ 1.8300038997859625 ], "mse_channel_mean": 1.8300038997859625, "conditional_kl": 162.36338692533366, "m30_total": 36.13268518463865 }, "random": { "mse": { "mean": 1.9990614652633667, "std": 0.02397168055176735, "min": 1.9353325366973877, "true_z": 44.59637451171875, "true_rank_among_random": 0, "count": 300 }, "mse_channel_mean": { "mean": 1.9990614652633667, "std": 0.02397168055176735, "min": 1.9353325366973877, "true_z": 44.59637451171875, "true_rank_among_random": 0, "count": 300 }, "conditional_kl": { "mean": 162.61036682128906, "std": 2.2810802459716797, "min": 157.043212890625, "true_z": 44.96923828125, "true_rank_among_random": 0, "count": 300 }, "m30_total": { "mean": 36.52019500732422, "std": 0.4689028859138489, "min": 35.337093353271484, "true_z": 48.31231689453125, "true_rank_among_random": 0, "count": 300 } } }, "gate_b": { "pairs": 130816, "improving_pairs": 271, "improving_fraction": 0.002071612035225049, "delta_mean": 0.007848642884402391, "delta_min": -0.004467133777118928, "identity_is_local_min_mse": false, "worst_improving_swaps": [ { "pair": [ 131, 334 ], "delta_mse": -0.004467133777118928, "text_cosine": 1.881744439051865, "visual_cosine": 0.5073372484880835 }, { "pair": [ 64, 158 ], "delta_mse": -0.0037848839739655677, "text_cosine": -0.2983393359569729, "visual_cosine": 0.037696856097792844 }, { "pair": [ 28, 158 ], "delta_mse": -0.003683242093005895, "text_cosine": 1.81501344963681, "visual_cosine": 1.1801674150832302 }, { "pair": [ 321, 434 ], "delta_mse": -0.003557930908854528, "text_cosine": 0.11369066429120464, "visual_cosine": 1.1804349132769365 }, { "pair": [ 64, 131 ], "delta_mse": -0.003415828825072732, "text_cosine": -0.6801621159938575, "visual_cosine": 0.1908215735263468 }, { "pair": [ 131, 135 ], "delta_mse": -0.003248027291974044, "text_cosine": 0.4751710629893898, "visual_cosine": -1.1564625017213115 }, { "pair": [ 75, 334 ], "delta_mse": -0.003207364128212948, "text_cosine": 0.7635089594066743, "visual_cosine": 0.651351723243795 }, { "pair": [ 64, 232 ], "delta_mse": -0.0031709159645822264, "text_cosine": -0.20208163164159137, "visual_cosine": 0.2191522404011218 }, { "pair": [ 64, 434 ], "delta_mse": -0.0030453962821109628, "text_cosine": 1.5636026078680862, "visual_cosine": -0.10044839358424484 }, { "pair": [ 334, 417 ], "delta_mse": -0.0030047048091416725, "text_cosine": 0.9137843127526165, "visual_cosine": 0.6338860958145965 }, { "pair": [ 76, 434 ], "delta_mse": -0.0028493854746996844, "text_cosine": 0.22779805194149536, "visual_cosine": 0.7330565784579498 }, { "pair": [ 135, 478 ], "delta_mse": -0.002832835651214184, "text_cosine": 2.6751970440626534, "visual_cosine": 1.6885325359105137 }, { "pair": [ 131, 467 ], "delta_mse": -0.0026876937084237334, "text_cosine": 1.969557408356569, "visual_cosine": 0.7886515837914926 }, { "pair": [ 227, 321 ], "delta_mse": -0.002645039040365607, "text_cosine": 1.2132723778483088, "visual_cosine": -0.04055466425627424 }, { "pair": [ 74, 227 ], "delta_mse": -0.0026219241542046385, "text_cosine": 0.30340194317623004, "visual_cosine": 0.6245900485946351 }, { "pair": [ 158, 334 ], "delta_mse": -0.0026154480751618726, "text_cosine": 0.30340194317623004, "visual_cosine": 0.4119367444167878 }, { "pair": [ 227, 335 ], "delta_mse": -0.0024302299461121768, "text_cosine": 2.315658428495852, "visual_cosine": 0.6727119174448595 }, { "pair": [ 64, 75 ], "delta_mse": -0.0024262499654758617, "text_cosine": -0.43885310809176603, "visual_cosine": 0.10928614101988993 }, { "pair": [ 360, 442 ], "delta_mse": -0.0023668063618040346, "text_cosine": 1.0687989439452257, "visual_cosine": 0.47053242852114885 }, { "pair": [ 227, 351 ], "delta_mse": -0.0023586359870804486, "text_cosine": 1.661659234815383, "visual_cosine": -0.12923645865970063 } ] }, "descent_from_true": { "start_accuracy": 1.0, "final_accuracy": 0.748046875, "final_objective": 0.8423136428012938, "final_energy": { "mse": 0.8423136428012938, "mse_channels": [ 0.8423136428012938 ], "mse_channel_mean": 0.8423136428012938, "conditional_kl": 70.76533524844979, "m30_total": 15.837694335292547 }, "accepted_moves": 110, "moved_fraction": 0.251953125, "mse_local_min_certificate": true, "trajectory": [ { "step": 0, "objective": 0.9255442940218683, "accuracy": 0.99609375 }, { "step": 50, "objective": 0.8654292720529905, "accuracy": 0.845703125 }, { "step": 100, "objective": 0.8458247698725851, "accuracy": 0.763671875 } ] }, "descent_from_random": [ { "start_accuracy": 0.00390625, "final_accuracy": 0.0, "final_objective": 0.7580243074667786, "final_energy": { "mse": 0.7580243074667786, "mse_channels": [ 0.7580243074667786 ], "mse_channel_mean": 0.7580243074667786, "conditional_kl": 93.14555737902761, "m30_total": 20.14516009073908 }, "accepted_moves": 754, "moved_fraction": 0.984375, "mse_local_min_certificate": true }, { "start_accuracy": 0.0, "final_accuracy": 0.0, "final_objective": 0.7838500894389513, "final_energy": { "mse": 0.7838500894389513, "mse_channels": [ 0.7838500894389513 ], "mse_channel_mean": 0.7838500894389513, "conditional_kl": 96.32116994840675, "m30_total": 20.831934168559254 }, "accepted_moves": 745, "moved_fraction": 0.982421875, "mse_local_min_certificate": true }, { "start_accuracy": 0.001953125, "final_accuracy": 0.001953125, "final_objective": 0.7770707750509355, "final_energy": { "mse": 0.7770707750509355, "mse_channels": [ 0.7770707750509355 ], "mse_channel_mean": 0.7770707750509355, "conditional_kl": 91.93526106496762, "m30_total": 19.941193763095395 }, "accepted_moves": 769, "moved_fraction": 0.98046875, "mse_local_min_certificate": true }, { "start_accuracy": 0.0, "final_accuracy": 0.00390625, "final_objective": 0.7816937178687126, "final_energy": { "mse": 0.7816937178687126, "mse_channels": [ 0.7816937178687126 ], "mse_channel_mean": 0.7816937178687126, "conditional_kl": 98.96646900258722, "m30_total": 21.35668123625487 }, "accepted_moves": 774, "moved_fraction": 0.98046875, "mse_local_min_certificate": true }, { "start_accuracy": 0.005859375, "final_accuracy": 0.0, "final_objective": 0.7881194886877553, "final_energy": { "mse": 0.7881194886877553, "mse_channels": [ 0.7881194886877553 ], "mse_channel_mean": 0.7881194886877553, "conditional_kl": 91.99377283194835, "m30_total": 19.97499354376518 }, "accepted_moves": 766, "moved_fraction": 0.982421875, "mse_local_min_certificate": true } ], "verdict": { "true_z_mse": 44.59637451171875, "improving_fraction": 0.002071612035225049, "descent_keeps": 0.748046875, "true_mse": 0.9300114277989874, "best_random_descent": 0.7580243074667786, "counterfeit_found": true } }