Delay-Kernel Bifurcation Scheduler / bench_report.json

Mechanism confirmed, baseline not beaten

Raw ⬇ ZIP
  1{
  2  "bench_version": 1,
  3  "track": "dynamics",
  4  "model": "rnn_small",
  5  "metric_direction": "lower is better",
  6  "n_seeds": 8,
  7  "baseline": {
  8    "best_cfg": {
  9      "lr": 0.006,
 10      "epochs": 18,
 11      "wd": 0.0001
 12    },
 13    "sweep": [
 14      {
 15        "cfg": {
 16          "lr": 0.0015,
 17          "epochs": 18,
 18          "wd": 0.0
 19        },
 20        "mean": 0.0016401863540522754
 21      },
 22      {
 23        "cfg": {
 24          "lr": 0.0015,
 25          "epochs": 18,
 26          "wd": 0.0001
 27        },
 28        "mean": 0.0016229975590249524
 29      },
 30      {
 31        "cfg": {
 32          "lr": 0.003,
 33          "epochs": 18,
 34          "wd": 0.0
 35        },
 36        "mean": 0.00105226032610517
 37      },
 38      {
 39        "cfg": {
 40          "lr": 0.003,
 41          "epochs": 18,
 42          "wd": 0.0001
 43        },
 44        "mean": 0.0010288678167853504
 45      },
 46      {
 47        "cfg": {
 48          "lr": 0.006,
 49          "epochs": 18,
 50          "wd": 0.0
 51        },
 52        "mean": 0.0007705322786932811
 53      },
 54      {
 55        "cfg": {
 56          "lr": 0.006,
 57          "epochs": 18,
 58          "wd": 0.0001
 59        },
 60        "mean": 0.0007483311783289537
 61      }
 62    ],
 63    "full": {
 64      "mean": 0.0007712390506640077,
 65      "std": 0.00015769195775709773,
 66      "per_seed": [
 67        0.0006971143302507699,
 68        0.0008527055033482611,
 69        0.0005140206776559353,
 70        0.0009294842020608485,
 71        0.0005480053368955851,
 72        0.0008613452082499862,
 73        0.0008049127063713968,
 74        0.0009623244404792786
 75      ],
 76      "n": 8
 77    }
 78  },
 79  "idea": {
 80    "mean": 0.0007767367897031363,
 81    "std": 0.00015170973680853395,
 82    "per_seed": [
 83      0.0007888000109232962,
 84      0.0009194124140776694,
 85      0.0004823579511139542,
 86      0.0008317303727380931,
 87      0.0005752710276283324,
 88      0.0008502130513079464,
 89      0.0008313465514220297,
 90      0.000934762938413769
 91    ],
 92    "n": 8
 93  },
 94  "comparison": {
 95    "delta_mean": 5.497739039128646e-06,
 96    "idea_wins": 4,
 97    "n_pairs": 8,
 98    "per_seed_diffs": [
 99      9.168568067252636e-05,
100      6.670691072940826e-05,
101      -3.16627265419811e-05,
102      -9.775382932275534e-05,
103      2.7265690732747316e-05,
104      -1.1132156942039728e-05,
105      2.6433845050632954e-05,
106      -2.7561502065509558e-05
107    ],
108    "p_value": 0.8404,
109    "mde": 5.046532793104354e-05,
110    "mde_rel_pct": 6.5434093213504685,
111    "verdict": "no measurable effect",
112    "system_worked": false
113  },
114  "mechanism_signature": {
115    "predicted_vs_observed": {
116      "final_tau": 0.7026815647163405,
117      "root_trace": [
118        [
119          0.41413663615422736,
120          -3.976707951927841
121        ],
122        [
123          0.5210281416079688,
124          -1.3861438181717687
125        ],
126        [
127          0.5849097575396691,
128          -0.8485201991462532
129        ],
130        [
131          0.6324862316350907,
132          -0.6447059261927688
133        ],
134        [
135          0.6707034118207508,
136          -0.5277147523207517
137        ],
138        [
139          0.7026815647163405,
140          -0.44972691119487185
141        ]
142      ],
143      "observed_sensitivity_peak_token": 6,
144      "observed_sensitivity_profile": [
145        0.04500379413366318,
146        0.06951507180929184,
147        0.10417859256267548,
148        0.14784090220928192,
149        0.20971933007240295,
150        0.2682212293148041,
151        0.31737804412841797,
152        0.09851673245429993
153      ],
154      "predicted_tau_c": 0.6045997880780726,
155      "predicted_omega": 1.7320508075688772,
156      "observed_final_root": -0.39328237958536605,
157      "confirmed": true,
158      "trained_model_probe_seed": 0
159    },
160    "note": "root schedule and test-time perturbation-compatible delayed system measured on trained benchmark model"
161  },
162  "idea_sweep": [
163    {
164      "cfg": {
165        "lr": 0.006,
166        "wd": 0.0001,
167        "epochs": 18,
168        "tau0": 0.1,
169        "gamma": 0.08,
170        "target": -0.05
171      },
172      "mean": 0.0007767367897031363,
173      "per_seed": [
174        0.0007888000109232962,
175        0.0009194124140776694,
176        0.0004823579511139542,
177        0.0008317303727380931,
178        0.0005752710276283324,
179        0.0008502130513079464,
180        0.0008313465514220297,
181        0.000934762938413769
182      ]
183    },
184    {
185      "cfg": {
186        "lr": 0.006,
187        "wd": 0.0001,
188        "epochs": 18,
189        "tau0": 0.6,
190        "gamma": 0.08,
191        "target": -0.05
192      },
193      "mean": 0.0007824423082638532,
194      "per_seed": [
195        0.0008003599941730499,
196        0.0009204379166476429,
197        0.0004860293702222407,
198        0.000844792346470058,
199        0.0005833087489008904,
200        0.0008501522243022919,
201        0.0007866035448387265,
202        0.0009878543205559254
203      ]
204    },
205    {
206      "cfg": {
207        "lr": 0.006,
208        "wd": 0.0001,
209        "epochs": 18,
210        "tau0": 0.6,
211        "gamma": 0.16,
212        "target": -0.05
213      },
214      "mean": 0.0007843925777706318,
215      "per_seed": [
216        0.0008042265544645488,
217        0.0009266391862183809,
218        0.0004910334246233106,
219        0.0008239991730079055,
220        0.0005916950758546591,
221        0.0008566270698793232,
222        0.0007817440782673657,
223        0.0009991760598495603
224      ]
225    }
226  ],
227  "custom_track": null
228}