{
  "version": 1,
  "article": {
    "path": "iv-surface-sampling-replication",
    "title": "对齐目标误差最低 0.45 倍，用错设计最高放大 2.03 倍：IV 曲面抽样协议的合成复测",
    "locale": "zh-cn",
    "canonical": "https://vxi.me/posts/iv-surface-sampling-replication/",
    "markdown": "https://vxi.me/posts/iv-surface-sampling-replication.md"
  },
  "verification": {
    "level": "first-party",
    "verified_at": "2026-11-03T00:00:00.000Z",
    "summary": "预注册冻结协议的一手复测：等权侧排序复现（等格 0.45、错配 2.03 倍），PPS 在其 LW 目标上被等格全档击败，归因层级错配；时钟方向复现。限合成总体与单次运行，论文 NVDA 量级未测。",
    "environments": [
      "Python 3.12 / NumPy 自行实现（Madow 系统抽样、Hájek 与 HT 比率估计）",
      "冻结合成期权链总体：200 日 / 111,573 份合约 / 种子 20261103",
      "日内过程 320 日（GARCH 波动聚集）/ 种子 20261104",
      "整日聚类 bootstrap 1,000 次给出全部区间"
    ],
    "evidence_mode": "mixed"
  },
  "public_sources": {
    "repository": {
      "url": "https://github.com/vxime/iv-surface-sampling-replication",
      "commit": "58f4505494eb9f646367403de0d4e076c93ddf89",
      "label": "公开复现代码、回执与逐项数据"
    },
    "run": null
  }
}