logbook.json

{
  "schema_version": 1,
  "title": "Reproduction: Decomposition-Based Modular Conformal Prediction for Two-Stage Modeling",
  "emoji": "🧮",
  "space_id": "snaykey/repro-modular-conformal",
  "paper": {
    "openreview_id": "fhoIsmyGGi"
  },
  "tags": [
    "icml2026-repro",
    "paper-fhoIsmyGGi"
  ],
  "updated_at": "2026-08-03T00:00:00+00:00",
  "root": {
    "slug": "index",
    "title": "Reproduction: Decomposition-Based Modular Conformal Prediction for Two-Stage Modeling",
    "file": "pages/index.md",
    "children": [
      {
        "slug": "claim-1-prop1-residual-bound",
        "title": "Proposition 1 establishes the upper bound on the true residual R ≤ ΔR₁ + R₂, decomposing prediction error into a first-stage delta term and a second-stage residual term (Section 3).",
        "file": "pages/claim-1-prop1-residual-bound/page.md",
        "children": []
      },
      {
        "slug": "claim-2-thm1-separate-quantile-coverage",
        "title": "Theorem 1 proves that the separate-component-quantile interval Ĉ^(c,d) built from quantiles of ΔR₁ and R₂ achieves marginal coverage ℙ(y_test ∈ Ĉ^(c,d)) ≥ 1 − c − d under exchangeability (Section 4).",
        "file": "pages/claim-2-thm1-separate-quantile-coverage/page.md",
        "children": []
      },
      {
        "slug": "claim-3-thm2-fwer-calibration",
        "title": "Theorem 2 shows that calibrating stage-wise scaling parameters λ via family-wise error rate (FWER) control over a validation set Λ_val at level δ yields a high-probability coverage guarantee ℙ(ℙ(y_test ∈ Ĉ^λ̂ | S^cal) ≥ 1 − α − τ) ≥ 1 − δ (Section 5.2).",
        "file": "pages/claim-3-thm2-fwer-calibration/page.md",
        "children": []
      },
      {
        "slug": "claim-4-gradual-upstream-shift",
        "title": "Under gradual upstream distribution shift, the proposed method (SR) achieves 0.8419 empirical coverage versus 0.7280 for standard split conformal (SC) and 0.7488 for weighted split conformal (WSC) (Table 1, Section 7.1).",
        "file": "pages/claim-4-gradual-upstream-shift/page.md",
        "children": []
      },
      {
        "slug": "claim-5-supply-chain-covid",
        "title": "On the real-world automobile supply-chain dataset spanning the 2020 COVID disruption, SR preserves target coverage while producing narrower intervals (width 21.085) than ACI (15.696 width, lower coverage) and PID (18.469 width) (Section 7.2).",
        "file": "pages/claim-5-supply-chain-covid/page.md",
        "children": []
      },
      {
        "slug": "executive-summary",
        "title": "executive-summary",
        "file": "pages/executive-summary/page.md",
        "children": []
      },
      {
        "slug": "conclusion",
        "title": "conclusion",
        "file": "pages/conclusion/page.md",
        "children": []
      }
    ]
  }
}