Spectral-Ordering Block Optimizer / bench_report.json

Mechanism confirmed, baseline not beaten

Raw ⬇ ZIP
  1{
  2  "bench_version": 1,
  3  "track": "sequence",
  4  "model": "transformer_tiny",
  5  "metric_direction": "lower is better",
  6  "n_seeds": 8,
  7  "baseline": {
  8    "best_cfg": {
  9      "lr": 0.0015,
 10      "weight_decay": 0.0
 11    },
 12    "sweep": [
 13      {
 14        "cfg": {
 15          "lr": 0.0015,
 16          "weight_decay": 0.0
 17        },
 18        "mean": 0.3047138750553131
 19      },
 20      {
 21        "cfg": {
 22          "lr": 0.003,
 23          "weight_decay": 0.0
 24        },
 25        "mean": 0.347607359290123
 26      },
 27      {
 28        "cfg": {
 29          "lr": 0.006,
 30          "weight_decay": 0.0
 31        },
 32        "mean": 0.47413551807403564
 33      },
 34      {
 35        "cfg": {
 36          "lr": 0.003,
 37          "weight_decay": 0.0001
 38        },
 39        "mean": 0.34758908301591873
 40      }
 41    ],
 42    "full": {
 43      "mean": 0.3249584026634693,
 44      "std": 0.05372834754753049,
 45      "per_seed": [
 46        0.2926701307296753,
 47        0.2746025025844574,
 48        0.2790817320346832,
 49        0.3725011348724365,
 50        0.32657936215400696,
 51        0.2697966992855072,
 52        0.43131738901138306,
 53        0.35311827063560486
 54      ],
 55      "n": 8
 56    }
 57  },
 58  "idea": {
 59    "best_cfg": {
 60      "lr": 0.0015,
 61      "weight_decay": 0.0
 62    },
 63    "sweep": [
 64      {
 65        "cfg": {
 66          "lr": 0.0015,
 67          "weight_decay": 0.0
 68        },
 69        "mean": 1.981750339269638
 70      },
 71      {
 72        "cfg": {
 73          "lr": 0.0015,
 74          "weight_decay": 0.0
 75        },
 76        "mean": 1.981750339269638
 77      },
 78      {
 79        "cfg": {
 80          "lr": 0.006,
 81          "weight_decay": 0.0
 82        },
 83        "mean": 20.26469349861145
 84      }
 85    ],
 86    "mean": 1.7784757614135742,
 87    "std": 0.4284677525253833,
 88    "per_seed": [
 89      2.11492919921875,
 90      2.2909317016601562,
 91      1.8699382543563843,
 92      1.6512022018432617,
 93      2.2055816650390625,
 94      1.2742841243743896,
 95      1.835213541984558,
 96      0.9857254028320312
 97    ],
 98    "n": 8
 99  },
100  "comparison": {
101    "delta_mean": 1.453517358750105,
102    "idea_wins": 0,
103    "n_pairs": 8,
104    "per_seed_diffs": [
105      1.8222590684890747,
106      2.016329199075699,
107      1.590856522321701,
108      1.2787010669708252,
109      1.8790023028850555,
110      1.0044874250888824,
111      1.403896152973175,
112      0.6326071321964264
113    ],
114    "p_value": 0.0081,
115    "mde": 0.3948775559757818,
116    "mde_rel_pct": 121.51633954968739,
117    "verdict": "idea worse (significant)",
118    "system_worked": false
119  },
120  "mechanism_signature": {
121    "predicted_rho": 0.9985000165639166,
122    "observed_local_perturbation_ratio": 1.0000219345092773,
123    "relative_gap_pct": 0.15242042264536065,
124    "selected_order": [
125      0,
126      1,
127      2
128    ],
129    "trained_test_mse": 0.2926701307296753,
130    "confirmed": true
131  }
132}