poster_embed.html

<!DOCTYPE html>
<!-- DESIGN DIRECTION (Step 2.5, locked)
  concept: diagnostic instrument panel — precise gauges for conditional coverage
  axis1 layout: single full-flow evidence card; focal content is the headline result figure
  axis2 canvas: cool light tint, flat, full
  axis3 palette: accent + emphasis register; steel blue #245C80 and rust #A2521C
  axis4 type: grotesque display + serif body
  axis5 density: balanced
  axis6 card: white surface, hairline frame, shadow, blue diagnostic rail
  axis7 heading: numbered underline joint
  axis8 masthead: centered clinical rule; footer contact strip
  devices: metric scoreboard, verdict chips
  hero moment: 77.6% recovered deviation — the one loud element
  soft-clash notes: single-flow skeleton chosen for robust logbook embedding and equal-height evidence panels
-->
<html lang="en">
<head>
<meta charset="UTF-8">
<meta name="generator" content="posterly">
<title>Conditional Coverage Diagnostics for Conformal Prediction</title>
<style>
  @page { size: 60in 36in; margin: 0; }
  :root {
    --accent: #245C80;
    --accent-deep: #153B57;
    --accent-soft: #DCEBF4;
    --emph: #A2521C;
    --emph-soft: #F4E4D9;
    --ink-on-accent: #FFFFFF;
    --text: #16222B;
    --muted: #64717C;
    --page: #EDF4F8;
    --card: #FFFFFF;
    --line: #BDD2DF;
    --ps-mark-ink: #64717C;
    --u: 1.6px;
    --font-sans: "Inter", "Helvetica Neue", Arial, sans-serif;
    --font-serif: "Charter", "Source Serif Pro", Georgia, serif;
  }
  * { box-sizing: border-box; margin: 0; padding: 0; }
  p, li, figcaption, .lede, .panel-copy, .caption { text-wrap: pretty; }
  .title, .hero-copy { text-wrap: balance; }
  html, body { background: #2b2b2b; color: var(--text); font-family: var(--font-serif); }
  .poster {
    width: calc(1524 * var(--u));
    height: calc(914.4 * var(--u));
    margin: 20px auto;
    padding: calc(16 * var(--u)) calc(22 * var(--u)) calc(13 * var(--u));
    background: var(--page);
    display: grid;
    grid-template-columns: minmax(0, 1fr);
    grid-template-rows: auto minmax(0, 1fr) auto auto;
    gap: calc(10 * var(--u));
    position: relative;
    overflow: hidden;
  }
  .header {
    display: grid;
    grid-template-columns: calc(170 * var(--u)) minmax(0, 1fr) calc(250 * var(--u));
    align-items: center;
    gap: calc(16 * var(--u));
    padding-bottom: calc(9 * var(--u));
    border-bottom: calc(2 * var(--u)) solid var(--accent);
  }
  .venue {
    font-family: var(--font-sans);
    font-weight: 900;
    color: var(--ink-on-accent);
    background: var(--accent);
    padding: calc(10 * var(--u));
    text-align: center;
    letter-spacing: calc(1 * var(--u));
    border-radius: calc(3 * var(--u));
  }
  .venue strong { display: block; font-size: calc(17 * var(--u)); line-height: 1; }
  .venue span { display: block; font-size: calc(7 * var(--u)); margin-top: calc(3 * var(--u)); }
  .title { font-family: var(--font-sans); font-size: calc(24 * var(--u)); line-height: 1.02; color: var(--accent-deep); font-weight: 850; text-align: center; }
  .byline { font-family: var(--font-sans); font-size: calc(7 * var(--u)); color: var(--muted); text-align: center; margin-top: calc(4 * var(--u)); }
  .status { font-family: var(--font-sans); color: var(--accent-deep); text-align: right; }
  .status .small { font-size: calc(6 * var(--u)); text-transform: uppercase; letter-spacing: calc(1 * var(--u)); }
  .status .large { font-size: calc(13 * var(--u)); font-weight: 850; margin-top: calc(2 * var(--u)); }
  .body { min-height: 0; display: grid; grid-template-columns: minmax(0, 1fr); }
  .column { min-height: 0; display: flex; flex-direction: column; }
  .card {
    min-height: 0;
    flex: 1;
    background: var(--card);
    border: calc(1 * var(--u)) solid var(--line);
    border-left: calc(7 * var(--u)) solid var(--accent);
    border-radius: calc(5 * var(--u));
    box-shadow: 0 calc(2 * var(--u)) calc(6 * var(--u)) rgba(0,0,0,.10);
    padding: calc(12 * var(--u)) calc(14 * var(--u));
    display: grid;
    grid-template-columns: 1fr 1.58fr;
    grid-template-rows: minmax(0, 1fr) auto auto;
    gap: calc(12 * var(--u));
    align-items: stretch;
  }
  .panel { min-width: 0; display: flex; flex-direction: column; justify-content: center; padding: calc(6 * var(--u)); }
  .panel + .panel { border-left: calc(1 * var(--u)) solid var(--line); padding-left: calc(18 * var(--u)); }
  .audit { grid-column: 1 / -1; display: grid; grid-template-columns: repeat(3, 1fr); gap: calc(8 * var(--u)); justify-content: initial; border-left: 0 !important; border-top: calc(1 * var(--u)) solid var(--line); padding: calc(5 * var(--u)) calc(9 * var(--u)) 0 !important; }
  .audit > .eyebrow { grid-column: 1 / -1; }
  .audit-item .section { margin-top: 0; }
  .eyebrow { font-family: var(--font-sans); color: var(--accent); font-size: calc(7 * var(--u)); font-weight: 800; letter-spacing: calc(1.2 * var(--u)); text-transform: uppercase; }
  .hero-stat { font-family: var(--font-sans); color: var(--emph); font-weight: 900; font-size: calc(28 * var(--u)); line-height: 1; margin-top: calc(6 * var(--u)); white-space: nowrap; }
  .hero-copy { font-family: var(--font-sans); font-size: calc(13 * var(--u)); line-height: 1.18; color: var(--accent-deep); font-weight: 760; margin-top: calc(6 * var(--u)); }
  .lede { font-size: calc(10 * var(--u)); line-height: 1.35; margin-top: calc(8 * var(--u)); text-wrap: pretty; }
  .section { font-family: var(--font-sans); font-size: calc(12 * var(--u)); color: var(--accent-deep); font-weight: 800; margin-top: calc(9 * var(--u)); padding-bottom: calc(3 * var(--u)); border-bottom: calc(1 * var(--u)) solid var(--accent-soft); }
  .panel-copy, .panel li { font-size: calc(9.5 * var(--u)); line-height: 1.33; text-wrap: pretty; }
  .panel ul { padding-left: calc(13 * var(--u)); margin-top: calc(5 * var(--u)); }
  .panel li { margin-bottom: calc(3 * var(--u)); }
  .chip-row { display: flex; flex-wrap: wrap; gap: calc(4 * var(--u)); margin-top: calc(8 * var(--u)); }
  .chip { font-family: var(--font-sans); font-size: calc(6 * var(--u)); padding: calc(3 * var(--u)) calc(6 * var(--u)); color: var(--accent-deep); background: var(--accent-soft); border: calc(.5 * var(--u)) solid var(--line); border-radius: calc(10 * var(--u)); font-weight: 700; }
  .chip.emph { background: var(--emph-soft); color: var(--text); }
  .figure { display: flex; flex-direction: column; align-items: center; justify-content: center; margin-top: calc(5 * var(--u)); flex: 1; }
  .figure img { display: block; width: 96%; max-height: calc(430 * var(--u)); object-fit: contain; background: var(--page); border: calc(1 * var(--u)) solid var(--line); border-radius: calc(3 * var(--u)); }
  .caption { font-family: var(--font-sans); font-size: calc(7.5 * var(--u)); color: var(--muted); text-align: center; margin-top: calc(3 * var(--u)); }
  .bottom-band { grid-column: 1 / -1; display: grid; grid-template-columns: repeat(3, 1fr); gap: calc(8 * var(--u)); padding: calc(6 * var(--u)) calc(9 * var(--u)); background: var(--page); border-top: calc(1 * var(--u)) solid var(--line); font-family: var(--font-sans); font-size: calc(7.5 * var(--u)); line-height: 1.25; }
  .bottom-band strong { display: block; color: var(--accent-deep); margin-bottom: calc(2 * var(--u)); }
  .scoreboard { display: grid; grid-template-columns: repeat(3, 1fr); gap: calc(8 * var(--u)); background: var(--accent-deep); color: var(--ink-on-accent); border-radius: calc(4 * var(--u)); padding: calc(7 * var(--u)) calc(12 * var(--u)); }
  .metric { text-align: center; font-family: var(--font-sans); display: flex; flex-direction: column; justify-content: center; }
  .metric strong { font-size: calc(15 * var(--u)); color: var(--emph-soft); white-space: nowrap; }
  .metric span { font-size: calc(5.8 * var(--u)); margin-top: calc(2 * var(--u)); }
  .footer { display: grid; grid-template-columns: 1fr auto; gap: calc(12 * var(--u)); font-family: var(--font-sans); font-size: calc(5.7 * var(--u)); color: var(--muted); padding: 0 calc(3 * var(--u)); }
  .footer .right { color: var(--accent-deep); font-weight: 750; }
  .ps-sprite { position: absolute; width: 0; height: 0; overflow: hidden; }
  .corner-sig { position: absolute; right: calc(4 * var(--u)); bottom: calc(3 * var(--u)); width: calc(7 * var(--u)); height: calc(7 * var(--u)); color: var(--ps-mark-ink); opacity: .5; }
  .corner-sig svg, [data-ps-mark="woven"] svg { width: 100%; height: 100%; display: block; }
  [data-ps-mark="woven"] { display: inline-block; width: .82em; height: .82em; vertical-align: -.14em; line-height: 0; }
  @media print {
    :root { --u: 1mm; }
    html, body { width: 100%; height: 100%; overflow: hidden; margin: 0; padding: 0; background: white; }
    .poster { margin: 0; box-shadow: none; page-break-after: avoid; }
  }
</style>
</head>
<body>
<main class="poster" data-measure-role="poster" data-posterly-contract="identity-v1" data-ps-identity="on">
  <svg class="ps-sprite" data-color-exempt="logo" aria-hidden="true"><defs><symbol id="psReg" viewBox="0 0 100 100"><circle cx="50" cy="50" r="25" fill="none" stroke="currentColor" stroke-width="8"/><line x1="50" y1="7" x2="50" y2="93" stroke="currentColor" stroke-width="8"/><line x1="7" y1="50" x2="93" y2="50" stroke="currentColor" stroke-width="8"/></symbol></defs></svg>
  <header class="header" data-measure-role="header">
    <div class="venue"><strong>ICML 2026</strong><span>REPRODUCTION</span></div>
      <div><h1 class="title">Conditional Coverage Diagnostics for Conformal Prediction</h1><div class="byline">Independent challenge reproduction · SabaPivot · OpenReview vaApZm6MKM</div></div>
    <div class="status"><div class="small">Outcome</div><div class="large">4 verified · 2 partial</div></div>
  </header>
  <section class="body" data-measure-role="body">
    <div class="column" data-measure-role="column">
      <article class="card" data-measure-role="card">
        <section class="panel">
          <div class="eyebrow">Independent headline</div>
          <div class="hero-stat">77.6%</div>
          <div class="hero-copy">of the true L1 conditional-coverage deviation recovered by cross-fitted LightGBM at n=5,000.</div>
          <p class="lede">A functional classifier detects heteroskedastic miscoverage missed by coarse ten-group partitioning.</p>
          <div class="chip-row"><span class="chip">5-fold CV</span><span class="chip">8 fresh seeds</span><span class="chip emph">independent code <span data-ps-mark="woven" data-color-exempt="logo" aria-hidden="true"><svg viewBox="0 0 100 100"><use href="#psReg"/></svg></span></span></div>
          <div class="section">What was tested</div>
          <ul><li>Eight-dimensional heteroskedastic Gaussian regression with standard split conformal intervals.</li><li>L1-, L2-, and KL-ERT from baseline-risk improvement.</li><li>Over/under decomposition and an explicit in-sample leakage stress test.</li></ul>
        </section>
        <section class="panel">
          <div class="eyebrow">Evidence</div>
          <figure class="figure"><img data-fig-layout="beside-text" src="results/ert_convergence.png" alt="ERT convergence against test sample size"><figcaption class="caption">Independent convergence study. Dashed line is the true L1 deviation.</figcaption></figure>
        </section>
        <section class="panel audit">
          <div class="eyebrow">Claim audit</div>
          <div class="audit-item">
          <div class="section">Verified</div>
          <p class="panel-copy">Definitions, classifier ordering, asymmetric decomposition, and cross-fitting all reproduced independently.</p>
          </div><div class="audit-item">
          <div class="section">Scoped or contradicted</div>
          <p class="panel-copy">Current arXiv Table 2 says 68.9% versus 33.7%; the supplied claim says 68.4% versus 38.3%. Table 4 was not rerun.</p>
          </div><div class="audit-item">
          <div class="section">Reproducibility boundary</div>
          <p class="panel-copy">No Hugging Face Job: the token lacks job.write. Public artifacts preserve seeds, outputs, and environment.</p>
          </div>
        </section>
        <aside class="bottom-band"><div><strong>Scope</strong>Synthetic diagnostics, definitions, and cross-fitting leakage.</div><div><strong>Version audit</strong>Current arXiv values checked against the supplied claim text.</div><div><strong>Evidence</strong>Code, CSV metrics, figure, and environment bundled publicly.</div></aside>
      </article>
    </div>
  </section>
  <section class="scoreboard" data-measure-role="footer-strip">
    <div class="metric"><strong>0.07194</strong><span>LightGBM L1-ERT at n=5,000</span></div>
    <div class="metric"><strong>0.00730</strong><span>PartitionWise L1-ERT at n=5,000</span></div>
    <div class="metric"><strong>0.0282</strong><span>five-fold value versus 0.168 in-sample</span></div>
  </section>
  <footer class="footer" data-measure-role="footer"><div>Rerun: uv run --with-requirements requirements.txt python reproduce.py</div><div class="right">hf.co/datasets/SabaPivot/icml26-repro-conditional-coverage-diagnostics</div></footer>
  <span class="corner-sig" data-ps-mark="corner" data-color-exempt="logo" aria-hidden="true"><svg viewBox="0 0 100 100"><use href="#psReg"/></svg></span>
</main>
</body>
</html>