# Эксперимент: GQL Safe Residual Layer (#1256)

{
  "worked": true,
  "confidence": 8,
  "verdict": "Built a NumPy GQL-style residual safety layer for U=(D,m,E), with an exact GQL identity check and robust scalar bisection to find the admissible interpolation boundary. Across two fixed seeds, the limiter produced 100% admissible states versus about 52–53% for the componentwise clipping control, while retaining about 0.67 of the learned residual; the identity error was below 2e-15. The claimed safety phenomenon is real in this toy setting, but the proposed delta-derived worst-normal shortcut was not safe here, so the MVP uses convex-margin bisection instead.",
  "metrics": {
    "baseline": "Componentwise D/E clipping: valid rate 0.52505 (seed 1234), 0.52385 (seed 2025); mean residual retention 0.9238.",
    "idea": "GQL residual limiter: valid rate 1.0 for both seeds; mean theta/residual retention 0.67179 (seed 1234), 0.66942 (seed 2025); minimum admissibility margin approximately 1.0e-6."
  },
  "how_to_run": "/home/maxwelhelp/main/bin/python3 gql_experiment.py",
  "files": [
    "gql_experiment.py",
    "gql_results.json"
  ],
  "limitations": "No neural operator, training loop, PDE shock-tube rollout, GPU implementation, EOS pressure inversion, runtime comparison, or multidimensional flux update was tested. The implementation uses numerical bisection rather than the paper's generalized eigenvalue/stereographic solver, and the clipping control only clips density and energy independently."
}