Bifurcation-Calibrated Stale-Gradient Controller / bench_report.json

Failed on benchmark

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_config": {
  9      "lr": 0.006
 10    },
 11    "full": {
 12      "mean": 0.0020863086683675647,
 13      "std": 0.00045500750701757853,
 14      "per_seed": [
 15        0.0018287990242242813,
 16        0.0014274335699155927,
 17        0.0014271392719820142,
 18        0.002110177418217063,
 19        0.002619170816615224,
 20        0.0024625908117741346,
 21        0.002196874236688018,
 22        0.00261828419752419
 23      ],
 24      "n": 8
 25    },
 26    "sweep": {
 27      "best_cfg": {
 28        "lr": 0.006
 29      },
 30      "sweep": [
 31        {
 32          "cfg": {
 33            "lr": 0.001
 34          },
 35          "mean": 0.004955001320922747
 36        },
 37        {
 38          "cfg": {
 39            "lr": 0.003
 40          },
 41          "mean": 0.00297580796177499
 42        },
 43        {
 44          "cfg": {
 45            "lr": 0.006
 46          },
 47          "mean": 0.0020863086683675647
 48        }
 49      ],
 50      "full": {
 51        "mean": 0.0020863086683675647,
 52        "std": 0.00045500750701757853,
 53        "per_seed": [
 54          0.0018287990242242813,
 55          0.0014274335699155927,
 56          0.0014271392719820142,
 57          0.002110177418217063,
 58          0.002619170816615224,
 59          0.0024625908117741346,
 60          0.002196874236688018,
 61          0.00261828419752419
 62        ],
 63        "n": 8
 64      }
 65    }
 66  },
 67  "idea": {
 68    "per_seed": [
 69      0.003938712179660797,
 70      0.0016923155635595322,
 71      0.0017820197390392423,
 72      0.000892799231223762,
 73      0.05783765763044357,
 74      0.07740116119384766,
 75      0.003681593108922243,
 76      0.0014422490494325757
 77    ],
 78    "mean": 0.018583563462016173,
 79    "config": {
 80      "lr": 0.006,
 81      "delay": 4
 82    }
 83  },
 84  "comparison": {
 85    "delta_mean": 0.016497254793648608,
 86    "idea_wins": 2,
 87    "n_pairs": 8,
 88    "per_seed_diffs": [
 89      0.002109913155436516,
 90      0.0002648819936439395,
 91      0.0003548804670572281,
 92      -0.001217378186993301,
 93      0.05521848681382835,
 94      0.07493857038207352,
 95      0.0014847188722342253,
 96      -0.0011760351480916142
 97    ],
 98    "p_value": 0.1486,
 99    "mde": 0.025474402834414687,
100    "mde_rel_pct": 1221.027512402907,
101    "verdict": "no significant win",
102    "system_worked": false
103  },
104  "mechanism_signature": {
105    "prediction": "cycle amplitude should increase approximately as mu^(1/M), M=4 for n=q=1",
106    "predicted_exponent": 0.25,
107    "observed_delay_plus_one_loglog_slope": -0.0009415027279748768,
108    "amplitude_by_delay": {
109      "0": 0.00010812481520919455,
110      "2": 0.00010821067043065237,
111      "4": 0.00010790639975175029
112    },
113    "return_slope_mean": -0.04708371831531154,
114    "confirmed": false
115  },
116  "idea_sweep": [
117    {
118      "per_seed": [
119        0.003938712179660797,
120        0.0016923155635595322,
121        0.0017820197390392423,
122        0.000892799231223762,
123        0.06468966603279114,
124        0.0856112390756607,
125        0.0035924564581364393,
126        0.0014422490494325757
127      ],
128      "mean": 0.020455182166188024,
129      "config": {
130        "lr": 0.006,
131        "delay": 0
132      }
133    },
134    {
135      "per_seed": [
136        0.003938712179660797,
137        0.0016923155635595322,
138        0.0017820197390392423,
139        0.000892799231223762,
140        0.06427709013223648,
141        0.08493981510400772,
142        0.003083865623921156,
143        0.0014422490494325757
144      ],
145      "mean": 0.02025610832788516,
146      "config": {
147        "lr": 0.006,
148        "delay": 2
149      }
150    },
151    {
152      "per_seed": [
153        0.003938712179660797,
154        0.0016923155635595322,
155        0.0017820197390392423,
156        0.000892799231223762,
157        0.05783765763044357,
158        0.07740116119384766,
159        0.003681593108922243,
160        0.0014422490494325757
161      ],
162      "mean": 0.018583563462016173,
163      "config": {
164        "lr": 0.006,
165        "delay": 4
166      }
167    }
168  ]
169}