Ultra-Local Neural Safety Shield / results.json

Unverified

Raw ⬇ ZIP
  1{
  2  "seed": 2347,
  3  "estimator_convergence": {
  4    "initial_error": 2.1555888661801905,
  5    "final_error": 2.1347138238807114e-05,
  6    "error_ratio": 9.903158516788637e-06,
  7    "converged": true
  8  },
  9  "boundary_scaling": {
 10    "kcs": [
 11      0.5,
 12      1.0,
 13      2.0,
 14      4.0,
 15      8.0
 16    ],
 17    "observed_action_boundary": [
 18      2.6,
 19      1.3,
 20      0.65,
 21      0.325,
 22      0.1625
 23    ],
 24    "predicted_action_boundary": [
 25      2.6,
 26      1.3,
 27      0.65,
 28      0.325,
 29      0.1625
 30    ],
 31    "observed_uncertainty_shift": [
 32      0.6000000000000001,
 33      0.30000000000000004,
 34      0.15000000000000002,
 35      0.07500000000000001,
 36      0.037500000000000006
 37    ],
 38    "predicted_uncertainty_shift": [
 39      0.6,
 40      0.3,
 41      0.15,
 42      0.075,
 43      0.0375
 44    ],
 45    "max_relative_shift_error": 1.8503717077055104e-16
 46  },
 47  "rejection_transition": {
 48    "deltas": [
 49      0.0,
 50      0.05,
 51      0.1,
 52      0.15000000000000002,
 53      0.2,
 54      0.25,
 55      0.30000000000000004,
 56      0.35000000000000003,
 57      0.4,
 58      0.45,
 59      0.5,
 60      0.55,
 61      0.6000000000000001,
 62      0.65,
 63      0.7000000000000001,
 64      0.75,
 65      0.8
 66    ],
 67    "observed_rejection": [
 68      0,
 69      0,
 70      1,
 71      1,
 72      1,
 73      1,
 74      1,
 75      1,
 76      1,
 77      1,
 78      1,
 79      1,
 80      1,
 81      1,
 82      1,
 83      1,
 84      1
 85    ],
 86    "predicted_rejection": [
 87      0,
 88      0,
 89      1,
 90      1,
 91      1,
 92      1,
 93      1,
 94      1,
 95      1,
 96      1,
 97      1,
 98      1,
 99      1,
100      1,
101      1,
102      1,
103      1
104    ],
105    "predicted_transition_delta": 0.09999999999999998,
106    "first_grid_rejection_delta": 0.1,
107    "classification_match": true
108  },
109  "comparison": {
110    "0.0": {
111      "none": {
112        "violation_rate": 0.8,
113        "min_y": -5.512533586373504,
114        "intervention_rate": 0.0,
115        "mean_envelope": 0.0,
116        "mean_abs_parameter_error": 0.7017936884469176,
117        "feasible_fraction": 0.0
118      },
119      "nominal": {
120        "violation_rate": 0.0,
121        "min_y": 6.249159913934024e-06,
122        "intervention_rate": 0.846,
123        "mean_envelope": 0.0,
124        "mean_abs_parameter_error": 0.4018022779395717,
125        "feasible_fraction": 1.0
126      },
127      "robust": {
128        "violation_rate": 0.0,
129        "min_y": 0.0002012089185122062,
130        "intervention_rate": 0.896,
131        "mean_envelope": 0.11467836639810068,
132        "mean_abs_parameter_error": 0.41695593532423597,
133        "feasible_fraction": 1.0
134      }
135    },
136    "0.4": {
137      "none": {
138        "violation_rate": 0.754,
139        "min_y": -5.3020792291599355,
140        "intervention_rate": 0.0,
141        "mean_envelope": 0.0,
142        "mean_abs_parameter_error": 0.7856583007615411,
143        "feasible_fraction": 0.0
144      },
145      "nominal": {
146        "violation_rate": 0.266,
147        "min_y": -0.05278730147661559,
148        "intervention_rate": 0.784,
149        "mean_envelope": 0.0,
150        "mean_abs_parameter_error": 0.4028288879612133,
151        "feasible_fraction": 1.0
152      },
153      "robust": {
154        "violation_rate": 0.0,
155        "min_y": 0.07663705359803767,
156        "intervention_rate": 0.806,
157        "mean_envelope": 0.3661973974228208,
158        "mean_abs_parameter_error": 0.3633387232740152,
159        "feasible_fraction": 1.0
160      }
161    },
162    "0.8": {
163      "none": {
164        "violation_rate": 0.724,
165        "min_y": -5.091624871946351,
166        "intervention_rate": 0.0,
167        "mean_envelope": 0.0,
168        "mean_abs_parameter_error": 0.8762396697011174,
169        "feasible_fraction": 0.0
170      },
171      "nominal": {
172        "violation_rate": 0.264,
173        "min_y": -0.10945973078773451,
174        "intervention_rate": 0.762,
175        "mean_envelope": 0.0,
176        "mean_abs_parameter_error": 0.5164801355918727,
177        "feasible_fraction": 1.0
178      },
179      "robust": {
180        "violation_rate": 0.0,
181        "min_y": 0.23924526758390846,
182        "intervention_rate": 0.668,
183        "mean_envelope": 0.7138660683830571,
184        "mean_abs_parameter_error": 0.4496627596910146,
185        "feasible_fraction": 1.0
186      }
187    },
188    "1.2": {
189      "none": {
190        "violation_rate": 0.704,
191        "min_y": -4.881170514732767,
192        "intervention_rate": 0.0,
193        "mean_envelope": 0.0,
194        "mean_abs_parameter_error": 0.9716611347332816,
195        "feasible_fraction": 0.0
196      },
197      "nominal": {
198        "violation_rate": 0.242,
199        "min_y": -0.13937602662428075,
200        "intervention_rate": 0.518,
201        "mean_envelope": 0.0,
202        "mean_abs_parameter_error": 0.5604387272209128,
203        "feasible_fraction": 0.96
204      },
205      "robust": {
206        "violation_rate": 0.0,
207        "min_y": 0.4461218594406183,
208        "intervention_rate": 0.562,
209        "mean_envelope": 1.106924317763119,
210        "mean_abs_parameter_error": 0.5431385545778623,
211        "feasible_fraction": 0.994
212      }
213    }
214  },
215  "checks": {
216    "sgd_estimator_convergence": true,
217    "inverse_kc_scaling": true,
218    "rejection_transition": true,
219    "robust_nonworse_all_amplitudes": true,
220    "mechanism_pass": true
221  }
222}