# Эксперимент: Effective-Exploration Bias Correction (#614)

{
  "worked": false,
  "confidence": 8,
  "verdict": "Built a reproducible UCB Gaussian-bandit simulation and a small contextual neural reward-model comparison in experiment.py. The suboptimal-arm bias was negative and adding 1/f_T moved the standardized statistic from -0.416 to -0.044, supporting the correction sign and local effect. However, f_T times the measured bias ranged from -0.38 to -2.50 in the exploration sweep, and the neural correction worsened held-out MSE (0.527 to 0.581) and rare-arm signed bias (0.365 to 0.406); therefore the idea did not produce a validated overall win.",
  "metrics": {
    "baseline": "Neural held-out MSE 0.5268; rare-arm signed bias 0.3651. Mechanism horizon sweep f_T*Z: -0.886, -0.973, -1.105, -1.276 for T=400,800,1400,2400.",
    "idea": "Neural held-out MSE 0.5806; rare-arm signed bias 0.4056. Direct known-gate correction changed Z from -0.4155 to -0.0440 at T=1400, but exploration sweep scaled bias ranged from -0.379 to -2.501 and was not consistently near predicted -1."
  },
  "how_to_run": "/home/maxwelhelp/main/bin/python3 experiment.py",
  "files": [
    "experiment.py",
    "results.json",
    "REPORT.json"
  ],
  "limitations": "Reduced Monte Carlo repetitions were used after the initial run exceeded the runtime budget. Only stationary synthetic Gaussian rewards were tested; the contextual policy used an oracle nonlinear mean for arm selection rather than a fully learned online neural UCB policy. The soft optimality gate was not separately calibrated, and no repeated neural seeds or FLOP-normalized study was performed."
}