1{
  2  "bench_version": 1,
  3  "track": "dynamics",
  4  "model": "su11_scan_matched",
  5  "metric_direction": "lower is better",
  6  "n_seeds": 8,
  7  "baseline": {
  8    "best_cfg": {
  9      "lr": 0.006
 10    },
 11    "sweep": [
 12      {
 13        "cfg": {
 14          "lr": 0.001
 15        },
 16        "mean": 0.6121751964092255
 17      },
 18      {
 19        "cfg": {
 20          "lr": 0.003
 21        },
 22        "mean": 0.31023339182138443
 23      },
 24      {
 25        "cfg": {
 26          "lr": 0.006
 27        },
 28        "mean": 0.02995142457075417
 29      }
 30    ],
 31    "full": {
 32      "mean": 0.035023413598537445,
 33      "std": 0.022148416496165235,
 34      "per_seed": [
 35        0.007818580605089664,
 36        0.020558837801218033,
 37        0.0673612505197525,
 38        0.024067029356956482,
 39        0.04443933442234993,
 40        0.010387144051492214,
 41        0.0372326523065567,
 42        0.06832247972488403
 43      ],
 44      "n": 8
 45    }
 46  },
 47  "idea": {
 48    "mean": 0.03177711000898853,
 49    "std": 0.020838650231134917,
 50    "per_seed": [
 51      0.006415394600480795,
 52      0.017402278259396553,
 53      0.061642639338970184,
 54      0.02086794376373291,
 55      0.04112565889954567,
 56      0.008823038078844547,
 57      0.0346921943128109,
 58      0.06324773281812668
 59    ],
 60    "n": 8
 61  },
 62  "comparison": {
 63    "delta_mean": -0.0032463035895489156,
 64    "idea_wins": 8,
 65    "n_pairs": 8,
 66    "per_seed_diffs": [
 67      -0.0014031860046088696,
 68      -0.00315655954182148,
 69      -0.005718611180782318,
 70      -0.0031990855932235718,
 71      -0.0033136755228042603,
 72      -0.001564105972647667,
 73      -0.002540457993745804,
 74      -0.005074746906757355
 75    ],
 76    "p_value": 0.0081,
 77    "mde": 0.0012718422855006327,
 78    "mde_rel_pct": 3.6314058363338515,
 79    "verdict": "idea better (significant)",
 80    "system_worked": true
 81  },
 82  "idea_sweep": [
 83    {
 84      "cfg": {
 85        "lr": 0.001
 86      },
 87      "mean": 0.60923171043396
 88    },
 89    {
 90      "cfg": {
 91        "lr": 0.003
 92      },
 93      "mean": 0.30314522609114647
 94    },
 95    {
 96      "cfg": {
 97        "lr": 0.006
 98      },
 99      "mean": 0.02658206399064511
100    }
101  ],
102  "idea_best_cfg": {
103    "lr": 0.006
104  },
105  "baseline_definition": "Euler discretization of the same learned SU(1,1) generator; exact exponential is the sole intervention.",
106  "protocol_note": "Identical trained systems, data, parameters, optimizer, epochs, and learning-rate grid; lower MSE is better.",
107  "mechanism_signature": {
108    "prediction": "Exact exponential preserves |a|^2-|b|^2=1 while Euler drift accumulates.",
109    "predicted_exact_invariant_error": 0.0,
110    "observed_exact_mean_error": 2.8064474122402316e-07,
111    "observed_exact_max_error": 1.430511474609375e-06,
112    "observed_euler_mean_error": 0.029417473822832108,
113    "observed_euler_max_error": 0.10198694467544556,
114    "confirmed": true
115  },
116  "timing_seconds": 95.12729549407959,
117  "parameter_counts": {
118    "baseline": 193,
119    "idea": 193
120  }
121}