Эксперимент: Multi-output BBL mass constraint (#1181)

{ "worked": false, "confidence": 9, "verdict": "Built bbl_experiment.py with stable weighted power means, Q_d(p), soft-min BBL penalty, positive heads, Monte Carlo mass ratios, Gaussian equality verification, and a fixed-seed two-view classifier comparison. The Gaussian sanity check reproduced the claimed equality (minimum ratio 1.0 and RHS 1.0), but the practical penalty had a negative gap on every training step, zero penalty, and zero gradient. Baseline and idea were bit-identical: 99.50% clean accuracy, 88.06% corrupted accuracy, ECE 0.00518, and disagreement 0.0991; therefore no observed regularization or robustness win.", "metrics": { "baseline": "clean_acc=0.9950, corrupt_acc=0.880625, ECE=0.005185, disagreement=0.099064, active_steps=0, max_penalty=0", "idea": "clean_acc=0.9950, corrupt_acc=0.880625, ECE=0.005185, disagreement=0.099064, active_steps=0, max_penalty=0; Gaussian check min=1.0, RHS=1.0" }, "how_to_run": "/home/maxwelhelp/main/bin/python3 bbl_experiment.py", "files": [ "bbl_experiment.py", "results.json" ], "limitations": "Only a small synthetic two-view binary classifier was tested, with one p=0 setting, fixed equal weights, approximate shared-sample mass estimates, and a single seed. No MNIST/CIFAR benchmark, hyperparameter sweep, multi-branch study, or comparison of alternative mass-normalization schemes was run." }