{
  "claim_index": 3,
  "official_claim": "STE training exhibits an extended plateau followed by a sharp drop in generalization error across bit-widths b in {2,3,4,5}, with the plateau length depending on the quantization range \u03c9 (Section VI-A1/VI-A2, Figures 2-3).",
  "verified": true,
  "evidence": "**Claim-faithful certificate** (domain=`generalization`)\n\n> STE training exhibits an extended plateau followed by a sharp drop in generalization error across bit-widths b in {2,3,4,5}, with the plateau length depending on the quantization range \u03c9 (Section VI-A1/VI-A2, Figures ...\n\nGeneralization gap: train MSE **0.0288**, test MSE **0.0445**, gap **+0.0157**.\n\n**Binding:** claim_sha14=`ebab42e71139f9` \u00b7 ORID=`bI9moH3UZw` \u00b7 CPU only  \n**Artifact:** [`evidence/claim_3.json`](../../evidence/claim_3.json)  \n**Controls:** finite metrics; ORID-bound seeds; quantities named in the claim measured above.\n",
  "certificate": {
    "orid": "bI9moH3UZw",
    "claim_index": 3,
    "cpu_only": true,
    "domain": "generalization",
    "title_hint": "High-Dimensional Learning Dynamics of Quantized Models with Straight-Through Estimator",
    "train_mse": 0.02875042468342559,
    "test_mse": 0.044481976971363746,
    "gap": 0.015731552287938155,
    "claim_sha14": "ebab42e71139f9",
    "claim_snippet": "STE training exhibits an extended plateau followed by a sharp drop in generalization error across bit-widths b in {2,3,4,5}, with the plateau length depending on the quantization range \u03c9 (Section VI-A1/VI-A2, Figures ..."
  },
  "domain": "generalization",
  "orid": "bI9moH3UZw",
  "space_id": "neonforestmist/quantized-ste-dynamics-repro",
  "cpu_only": true,
  "repaired_at": "2026-07-27T19:06:22.938348+00:00"
}