artifacts/mlp_analysis.json

{
  "provenance": "fresh independent CPU training; paired statistics recomputed from raw epoch CSV",
  "paper": "arXiv:2605.21648v2, Table 9 / Figure 6 scale",
  "raw_csv_sha256": "535cf185807d6d1b3546755d5217acf7e9c580ac9491987aba7c3579fe5c40a1",
  "integrity": {
    "schedules": [
      "constant",
      "early_step",
      "late_step"
    ],
    "seeds": [
      42,
      43,
      44,
      45,
      46,
      47,
      48,
      49,
      50,
      51,
      52,
      53,
      54,
      55,
      56,
      57,
      58,
      59,
      60,
      61
    ],
    "epochs": 50,
    "raw_rows": 3000,
    "complete_rectangular_panel": true
  },
  "early_step_vs_constant": {
    "candidate": "early_step",
    "baseline": "constant",
    "paired_seeds": [
      42,
      43,
      44,
      45,
      46,
      47,
      48,
      49,
      50,
      51,
      52,
      53,
      54,
      55,
      56,
      57,
      58,
      59,
      60,
      61
    ],
    "n": 20,
    "candidate_loss_mean": 1.6676626688539984,
    "baseline_loss_mean": 1.8574766079485419,
    "relative_loss_reduction_percent": 10.218914105420698,
    "candidate_better_loss_seeds": 20,
    "loss_paired_t": -35.925582256598155,
    "loss_paired_p_two_sided": 6.236102812906037e-19,
    "candidate_acc_mean": 42.753,
    "baseline_acc_mean": 42.114999999999995,
    "accuracy_difference_pp": 0.6380000000000052,
    "candidate_better_accuracy_seeds": 17,
    "accuracy_paired_t": 4.044973710624573,
    "accuracy_paired_p_two_sided": 0.0006913917615950649
  },
  "late_step_vs_constant": {
    "candidate": "late_step",
    "baseline": "constant",
    "paired_seeds": [
      42,
      43,
      44,
      45,
      46,
      47,
      48,
      49,
      50,
      51,
      52,
      53,
      54,
      55,
      56,
      57,
      58,
      59,
      60,
      61
    ],
    "n": 20,
    "candidate_loss_mean": 2.818823402583599,
    "baseline_loss_mean": 1.8574766079485419,
    "relative_loss_reduction_percent": -51.75552631571495,
    "candidate_better_loss_seeds": 0,
    "loss_paired_t": 77.64343361199273,
    "loss_paired_p_two_sided": 3.0245176013266884e-25,
    "candidate_acc_mean": 39.88799999999999,
    "baseline_acc_mean": 42.114999999999995,
    "accuracy_difference_pp": -2.227000000000004,
    "candidate_better_accuracy_seeds": 0,
    "accuracy_paired_t": -12.837213515069436,
    "accuracy_paired_p_two_sided": 8.226236187177159e-11
  },
  "early_step_vs_late_step": {
    "candidate": "early_step",
    "baseline": "late_step",
    "paired_seeds": [
      42,
      43,
      44,
      45,
      46,
      47,
      48,
      49,
      50,
      51,
      52,
      53,
      54,
      55,
      56,
      57,
      58,
      59,
      60,
      61
    ],
    "n": 20,
    "candidate_loss_mean": 1.6676626688539984,
    "baseline_loss_mean": 2.818823402583599,
    "relative_loss_reduction_percent": 40.83834172351846,
    "candidate_better_loss_seeds": 20,
    "loss_paired_t": -80.50977259918412,
    "loss_paired_p_two_sided": 1.5219000524190717e-25,
    "candidate_acc_mean": 42.753,
    "baseline_acc_mean": 39.88799999999999,
    "accuracy_difference_pp": 2.865000000000009,
    "candidate_better_accuracy_seeds": 20,
    "accuracy_paired_t": 19.359526948238305,
    "accuracy_paired_p_two_sided": 5.750138070626599e-14
  }
}