Publish Ropedia task-suite derived artifacts

README.md

@@ -22,10 +22,14 @@ size_categories:
 
 # Ropedia Episode Task Suite Artifacts
 
-This dataset repo contains derived experiment artifacts for the public Ropedia / Xperience-10M sample episode.
+This dataset repo contains the derived evidence layer for the public Ropedia / Xperience-10M sample episode: metrics, predictions, manifests, charts, diagrams, notes, and reproduction scripts.
 
 It does **not** contain raw Ropedia videos or raw `annotation.hdf5`. Download raw data only from the official Ropedia / Hugging Face sources and follow their terms.
 
+## Why This Repo Exists
+
+This is the reviewable half of the project. You can inspect the task outputs, compare the committed metrics, and understand the single-episode limitations without downloading the raw videos first.
+
 ## What Is Included
 
 - `results/**/*.json`: verified metrics and metadata

docs/index.html

@@ -4,44 +4,68 @@
   <meta charset="utf-8">
   <meta name="viewport" content="width=device-width, initial-scale=1">
   <title>Ropedia Episode Task Suite</title>
-  <meta name="description" content="A polished all-modality task-suite repo for one public Ropedia / Xperience-10M sample episode.">
+  <meta name="description" content="A transparent all-modality task suite for one public Ropedia / Xperience-10M sample episode.">
+  <meta property="og:title" content="Ropedia Episode Task Suite">
+  <meta property="og:description" content="One public embodied-AI episode, converted into reproducible multimodal tasks, baselines, metrics, and diagrams.">
+  <meta property="og:image" content="assets/task_suite_infographic.png">
   <style>
     :root {
       color-scheme: light;
-      --ink: #10141f;
-      --muted: #5b6475;
-      --line: #dce2ec;
-      --soft-line: #edf1f6;
-      --page: #ffffff;
-      --panel: #f8fafc;
-      --blue: #1f63e9;
-      --cyan: #008b9a;
-      --green: #0a7f55;
-      --amber: #b65b04;
-      --red: #b42318;
-      --shadow: 0 18px 55px rgba(24, 34, 54, 0.10);
+      --ink: #1f2421;
+      --muted: #6f716c;
+      --line: #e4ded4;
+      --soft-line: #eee9e1;
+      --page: #fbfaf7;
+      --panel: #f5f1e9;
+      --surface: #fffefd;
+      --blue: #1f6c9f;
+      --cyan: #2e7775;
+      --green: #346538;
+      --amber: #956400;
+      --red: #9f2f2d;
+      --shadow: 0 24px 60px rgba(68, 55, 38, 0.08);
       --radius: 8px;
-      --max: 1180px;
+      --max: 1220px;
+      --font-ui: "Avenir Next", "SF Pro Display", ui-sans-serif, system-ui, -apple-system, BlinkMacSystemFont, "Segoe UI", sans-serif;
+      --font-copy: "Avenir Next", "SF Pro Text", ui-sans-serif, system-ui, -apple-system, BlinkMacSystemFont, "Segoe UI", sans-serif;
+      --font-mono: "SF Mono", "JetBrains Mono", ui-monospace, SFMono-Regular, Menlo, monospace;
     }
 
     * { box-sizing: border-box; }
     html { scroll-behavior: smooth; }
     body {
       margin: 0;
-      font-family: Inter, ui-sans-serif, system-ui, -apple-system, BlinkMacSystemFont, "Segoe UI", sans-serif;
+      font-family: var(--font-copy);
       color: var(--ink);
-      background: var(--page);
+      background:
+        linear-gradient(90deg, rgba(68,55,38,0.028) 1px, transparent 1px),
+        linear-gradient(0deg, rgba(68,55,38,0.022) 1px, transparent 1px),
+        var(--page);
+      background-size: 56px 56px, 56px 56px, auto;
       line-height: 1.5;
+      text-rendering: optimizeLegibility;
     }
 
     a { color: inherit; }
     .wrap { width: min(var(--max), calc(100% - 48px)); margin: 0 auto; }
+    .skip-link {
+      position: absolute;
+      left: 16px;
+      top: 12px;
+      transform: translateY(-160%);
+      background: var(--ink);
+      color: white;
+      padding: 10px 12px;
+      border-radius: 6px;
+      z-index: 40;
+    }
+    .skip-link:focus { transform: translateY(0); }
 
     .site-nav {
       position: sticky;
       top: 0;
       z-index: 20;
-      background: rgba(255,255,255,0.92);
+      background: rgba(251,250,247,0.88);
       backdrop-filter: blur(18px);
       border-bottom: 1px solid var(--soft-line);
     }
@@ -56,29 +80,39 @@
       display: flex;
       align-items: center;
       gap: 12px;
-      font-weight: 760;
+      font-family: var(--font-ui);
+      font-weight: 700;
       letter-spacing: 0;
       text-decoration: none;
     }
     .mark {
       width: 34px;
       height: 34px;
-      border: 1px solid #c8d4e6;
+      border: 1px solid #d7ccbd;
       display: grid;
       place-items: center;
-      color: var(--blue);
-      background: #fff;
-      box-shadow: 0 8px 22px rgba(31, 99, 233, 0.12);
+      color: var(--ink);
+      background: #f1eadf;
+      font-family: var(--font-mono);
+      font-size: 13px;
+      font-weight: 760;
     }
     .nav-links {
       display: flex;
       align-items: center;
       gap: 20px;
       font-size: 14px;
-      color: #394255;
+      color: #4e514c;
+    }
+    .nav-links a {
+      text-decoration: none;
+      transition: color 220ms cubic-bezier(0.16, 1, 0.3, 1), transform 220ms cubic-bezier(0.16, 1, 0.3, 1);
+    }
+    .nav-links a:hover { color: var(--ink); transform: translateY(-1px); }
+    a:focus-visible, button:focus-visible {
+      outline: 2px solid rgba(31,108,159,0.48);
+      outline-offset: 3px;
     }
-    .nav-links a { text-decoration: none; }
-    .nav-links a:hover { color: var(--blue); }
     .nav-action {
       border: 1px solid var(--ink);
       background: var(--ink);
@@ -89,44 +123,65 @@
       align-items: center;
       gap: 8px;
       text-decoration: none;
-      font-weight: 650;
+      font-weight: 680;
       border-radius: 6px;
     }
 
     .hero {
       border-bottom: 1px solid var(--soft-line);
       background:
-        linear-gradient(90deg, rgba(31,99,233,0.08) 1px, transparent 1px),
-        linear-gradient(0deg, rgba(31,99,233,0.06) 1px, transparent 1px);
-      background-size: 54px 54px;
+        linear-gradient(115deg, rgba(255,255,255,0.82), rgba(245,241,233,0.72)),
+        linear-gradient(90deg, rgba(68,55,38,0.05) 1px, transparent 1px),
+        linear-gradient(0deg, rgba(68,55,38,0.035) 1px, transparent 1px);
+      background-size: auto, 54px 54px, 54px 54px;
     }
     .hero-inner {
       min-height: 680px;
       display: grid;
-      grid-template-columns: minmax(0, 1.03fr) minmax(420px, 0.97fr);
-      gap: 46px;
+      grid-template-columns: minmax(0, 1.08fr) minmax(420px, 0.92fr);
+      gap: 58px;
       align-items: center;
-      padding: 64px 0 54px;
+      padding: 86px 0 68px;
+    }
+    .eyebrow {
+      display: inline-flex;
+      align-items: center;
+      gap: 9px;
+      margin-bottom: 22px;
+      color: #4f524d;
+      font-family: var(--font-mono);
+      font-size: 12px;
+      text-transform: uppercase;
+      letter-spacing: 0.08em;
+    }
+    .eyebrow::before {
+      content: "";
+      width: 34px;
+      height: 1px;
+      background: var(--ink);
     }
     h1 {
       margin: 0;
-      font-size: clamp(46px, 7vw, 86px);
-      line-height: 0.95;
+      font-family: var(--font-ui);
+      font-size: clamp(48px, 7vw, 92px);
+      line-height: 0.98;
       letter-spacing: 0;
       max-width: 860px;
+      text-wrap: balance;
     }
     .hero-copy {
       margin: 26px 0 0;
       max-width: 690px;
-      color: #394255;
+      color: #4e514c;
       font-size: 19px;
       line-height: 1.65;
+      text-wrap: pretty;
     }
     .hero-actions {
       display: flex;
       flex-wrap: wrap;
       gap: 12px;
-      margin-top: 32px;
+      margin-top: 34px;
     }
     .button {
       height: 46px;
@@ -137,14 +192,16 @@
       align-items: center;
       gap: 10px;
       text-decoration: none;
-      font-weight: 700;
-      background: white;
+      font-weight: 720;
+      background: var(--surface);
+      transition: transform 240ms cubic-bezier(0.16, 1, 0.3, 1), border-color 240ms cubic-bezier(0.16, 1, 0.3, 1), background 240ms cubic-bezier(0.16, 1, 0.3, 1);
     }
+    .button:hover { transform: translateY(-2px); border-color: #cbbfad; }
+    .button:active { transform: translateY(0) scale(0.98); }
     .button.primary {
-      background: var(--blue);
+      background: var(--ink);
       color: white;
-      border-color: var(--blue);
-      box-shadow: 0 15px 30px rgba(31,99,233,0.22);
+      border-color: var(--ink);
     }
     .hero-stats {
       display: grid;
@@ -155,18 +212,18 @@
     }
     .stat {
       border: 1px solid var(--line);
-      background: rgba(255,255,255,0.86);
+      background: rgba(255,254,253,0.86);
       padding: 14px 14px 13px;
       border-radius: var(--radius);
     }
-    .stat strong { display: block; font-size: 22px; line-height: 1; }
+    .stat strong { display: block; font-family: var(--font-mono); font-size: 21px; line-height: 1; font-variant-numeric: tabular-nums; }
     .stat span { display: block; margin-top: 7px; font-size: 12px; color: var(--muted); }
     .hero-panel {
-      background: rgba(255,255,255,0.92);
+      background: rgba(255,254,253,0.92);
       border: 1px solid var(--line);
       border-radius: var(--radius);
       box-shadow: var(--shadow);
-      padding: 16px;
+      padding: 18px;
     }
     .panel-top {
       display: flex;
@@ -176,6 +233,7 @@
       padding: 6px 6px 14px;
       color: var(--muted);
       font-size: 13px;
+      font-family: var(--font-mono);
     }
     .signal {
       display: grid;
@@ -187,17 +245,18 @@
     }
     .signal:last-child { border-bottom: 0; }
     .signal code {
-      color: #273143;
-      background: #f2f5fa;
-      border: 1px solid #e2e8f0;
+      color: #2f3437;
+      background: #f4f0e8;
+      border: 1px solid var(--soft-line);
       padding: 5px 7px;
       border-radius: 5px;
       font-size: 12px;
+      font-family: var(--font-mono);
     }
     .track {
       position: relative;
       height: 12px;
-      background: #eef2f7;
+      background: #eee8df;
       overflow: hidden;
       border-radius: 999px;
     }
@@ -205,12 +264,12 @@
       position: absolute;
       inset: 0 auto 0 0;
       width: var(--w);
-      background: linear-gradient(90deg, var(--c), color-mix(in srgb, var(--c), white 28%));
+      background: var(--c);
       border-radius: inherit;
     }
-    .signal strong { text-align: right; font-size: 13px; }
+    .signal strong { text-align: right; font-family: var(--font-mono); font-size: 13px; font-variant-numeric: tabular-nums; }
 
-    section { padding: 78px 0; border-bottom: 1px solid var(--soft-line); }
+    section { padding: 88px 0 96px; border-bottom: 1px solid var(--soft-line); }
     .section-head {
       display: flex;
       justify-content: space-between;
@@ -220,15 +279,19 @@
     }
     h2 {
       margin: 0;
+      font-family: var(--font-ui);
       font-size: clamp(30px, 4vw, 48px);
       line-height: 1.05;
       letter-spacing: 0;
+      text-wrap: balance;
     }
     .section-head p {
       max-width: 560px;
       margin: 0;
       color: var(--muted);
       font-size: 16px;
+      line-height: 1.65;
+      text-wrap: pretty;
     }
 
     .pipeline-image,
@@ -238,8 +301,8 @@
       width: 100%;
       border: 1px solid var(--line);
       border-radius: var(--radius);
-      background: #fff;
-      box-shadow: 0 8px 28px rgba(17, 24, 39, 0.05);
+      background: var(--surface);
+      box-shadow: 0 14px 34px rgba(68, 55, 38, 0.045);
     }
     .task-suite-image {
       display: block;
@@ -250,24 +313,23 @@
     }
     .two-col {
       display: grid;
-      grid-template-columns: minmax(0, 1fr) minmax(340px, 0.55fr);
-      gap: 26px;
+      grid-template-columns: minmax(0, 1fr) minmax(340px, 0.52fr);
+      gap: 30px;
       align-items: start;
     }
     .callout {
       border: 1px solid var(--line);
-      border-left: 4px solid var(--amber);
       border-radius: var(--radius);
-      padding: 20px;
-      background: #fffaf3;
-      color: #3f2d17;
+      padding: 24px;
+      background: #f8f2e8;
+      color: #3f3425;
     }
     .callout h3, .artifact h3 { margin: 0 0 8px; font-size: 17px; }
-    .callout p, .artifact p { margin: 0; color: #5f5141; }
+    .callout p, .artifact p { margin: 0; color: #665d51; line-height: 1.6; }
 
     .models {
       display: grid;
-      grid-template-columns: repeat(4, minmax(0, 1fr));
+      grid-template-columns: 1.1fr 1.35fr 1fr 1.15fr;
       gap: 14px;
       margin-bottom: 22px;
     }
@@ -275,10 +337,12 @@
       border: 1px solid var(--line);
       border-radius: var(--radius);
       padding: 18px;
-      background: white;
+      background: var(--surface);
+      transition: transform 240ms cubic-bezier(0.16, 1, 0.3, 1), box-shadow 240ms cubic-bezier(0.16, 1, 0.3, 1);
     }
+    .model:hover { transform: translateY(-3px); box-shadow: 0 14px 28px rgba(68, 55, 38, 0.06); }
     .model h3 { margin: 0; font-size: 15px; }
-    .model .score { display: block; margin-top: 18px; font-size: 34px; font-weight: 790; line-height: 1; }
+    .model .score { display: block; margin-top: 18px; font-family: var(--font-mono); font-size: 33px; font-weight: 760; line-height: 1; font-variant-numeric: tabular-nums; }
     .model .meta { display: block; margin-top: 8px; color: var(--muted); font-size: 13px; }
 
     .task-toolbar {
@@ -289,14 +353,16 @@
     }
     .filter {
       border: 1px solid var(--line);
-      background: white;
+      background: var(--surface);
       border-radius: 999px;
       height: 36px;
       padding: 0 14px;
       font-weight: 650;
       color: #354052;
       cursor: pointer;
+      transition: transform 220ms cubic-bezier(0.16, 1, 0.3, 1), background 220ms cubic-bezier(0.16, 1, 0.3, 1);
     }
+    .filter:hover { transform: translateY(-1px); }
     .filter.active { color: white; background: var(--ink); border-color: var(--ink); }
     .task-grid {
       display: grid;
@@ -306,15 +372,17 @@
     .task-card {
       border: 1px solid var(--line);
       border-radius: var(--radius);
-      padding: 16px;
-      background: white;
+      padding: 18px;
+      background: var(--surface);
       display: grid;
       gap: 12px;
       min-height: 184px;
+      transition: transform 240ms cubic-bezier(0.16, 1, 0.3, 1), border-color 240ms cubic-bezier(0.16, 1, 0.3, 1);
     }
+    .task-card:hover { transform: translateY(-3px); border-color: #cbbfad; }
     .task-card.hide { display: none; }
     .task-top { display: flex; justify-content: space-between; gap: 14px; align-items: start; }
-    .task-name { font-family: ui-monospace, SFMono-Regular, Menlo, monospace; font-size: 13px; font-weight: 760; }
+    .task-name { font-family: var(--font-mono); font-size: 13px; font-weight: 760; }
     .tag {
       font-size: 11px;
       border-radius: 999px;
@@ -323,14 +391,14 @@
       background: #eef2f7;
       white-space: nowrap;
     }
-    .tag.supervised { background: #eaf2ff; color: #1549ad; }
-    .tag.forecast { background: #eaf8f0; color: #0a6b45; }
-    .tag.retrieval { background: #ecfbfd; color: #006b76; }
-    .tag.diagnostic { background: #fff4e5; color: #8a4400; }
+    .tag.supervised { background: #e1f3fe; color: #1f6c9f; }
+    .tag.forecast { background: #edf3ec; color: #346538; }
+    .tag.retrieval { background: #e6f4f1; color: #2e7775; }
+    .tag.diagnostic { background: #fbf3db; color: #956400; }
     .task-card p { margin: 0; color: var(--muted); font-size: 13px; }
     .metric-row { display: flex; justify-content: space-between; gap: 12px; font-size: 13px; }
-    .metric-row strong { font-size: 19px; }
-    .mini-bar { height: 7px; background: #edf1f6; border-radius: 999px; overflow: hidden; }
+    .metric-row strong { font-family: var(--font-mono); font-size: 18px; font-variant-numeric: tabular-nums; }
+    .mini-bar { height: 7px; background: #eee8df; border-radius: 999px; overflow: hidden; }
     .mini-bar span { display: block; height: 100%; width: var(--w); background: var(--c); }
 
     .artifact-grid {
@@ -347,29 +415,32 @@
     .artifact {
       border: 1px solid var(--line);
       border-radius: var(--radius);
-      background: white;
+      background: var(--surface);
       padding: 18px;
+      transition: transform 240ms cubic-bezier(0.16, 1, 0.3, 1), border-color 240ms cubic-bezier(0.16, 1, 0.3, 1);
     }
+    .artifact:hover { transform: translateY(-3px); border-color: #cbbfad; }
     .artifact a { display: inline-block; margin-top: 14px; font-weight: 740; text-decoration: none; color: var(--blue); }
+    .artifact a:hover { text-decoration: underline; text-underline-offset: 4px; }
 
     .code-panel {
-      background: #10141f;
-      color: #eef3ff;
+      background: #1f2421;
+      color: #f7f2e8;
       border-radius: var(--radius);
       padding: 18px;
       overflow: auto;
-      font-family: ui-monospace, SFMono-Regular, Menlo, monospace;
+      font-family: var(--font-mono);
       font-size: 13px;
       line-height: 1.65;
-      border: 1px solid #252d3d;
+      border: 1px solid #3c403b;
     }
     .code-panel button {
       float: right;
       margin-left: 16px;
       height: 30px;
-      border: 1px solid #475569;
+      border: 1px solid #5a6059;
       color: white;
-      background: #1f2937;
+      background: #363c36;
       border-radius: 5px;
       cursor: pointer;
       font-weight: 700;
@@ -379,7 +450,6 @@
       color: var(--muted);
       font-size: 14px;
     }
-
     @media (max-width: 960px) {
       .hero-inner, .two-col { grid-template-columns: 1fr; }
       .hero-inner { min-height: 0; }
@@ -395,9 +465,19 @@
       .signal { grid-template-columns: 1fr; }
       .signal strong { text-align: left; }
     }
+    @media (prefers-reduced-motion: reduce) {
+      html { scroll-behavior: auto; }
+      *, *::before, *::after {
+        transition-duration: 1ms !important;
+        animation-duration: 1ms !important;
+        animation-iteration-count: 1 !important;
+      }
+      .motion-ready .reveal { opacity: 1; transform: none; }
+    }
   </style>
 </head>
 <body>
+  <a class="skip-link" href="#main">Skip to content</a>
   <nav class="site-nav">
     <div class="wrap nav-inner">
       <a class="brand" href="#top" aria-label="Ropedia Episode Task Suite home">
@@ -420,17 +500,18 @@
   <header class="hero" id="top">
     <div class="wrap hero-inner">
       <div>
-        <h1>One embodied episode, twelve ways to learn from it.</h1>
+        <div class="eyebrow">single public episode / verified artifacts</div>
+        <h1>One Ropedia episode, made auditable.</h1>
         <p class="hero-copy">
-          A polished, reproducible task suite for the public Ropedia / Xperience-10M sample:
-          all-modality features, supervised baselines, self-supervised probes, metrics, artifacts,
-          and a clear path from single-episode debugging to real cross-episode evaluation.
+          A compact research lab for the public Ropedia / Xperience-10M sample:
+          aligned multimodal windows, 12 task contracts, small baseline heads, committed metrics,
+          and a clear line between single-episode inspection and real held-out evaluation.
         </p>
         <div class="hero-actions">
-          <a class="button primary" href="#tasks">Explore tasks</a>
+          <a class="button primary" href="#tasks">Inspect 12 tasks</a>
           <a class="button" href="https://cy0307-ropedia-episode-task-suite.static.hf.space/">Open HF app</a>
           <a class="button" href="https://github.com/ChaoYue0307/ropedia-episode-task-suite">View repository</a>
-          <a class="button" href="data/summary_metrics.json">Open summary JSON</a>
+          <a class="button" href="data/summary_metrics.json">Read metrics JSON</a>
         </div>
         <div class="hero-stats">
           <div class="stat"><strong>5,821</strong><span>frames in sample episode</span></div>
@@ -444,22 +525,22 @@
           <span>all-modality feature allocation</span>
           <span>window vector</span>
         </div>
-        <div class="signal"><code>mocap</code><div class="track"><span style="--w:25.3%;--c:#1f63e9"></span></div><strong>2,121</strong></div>
-        <div class="signal"><code>camera+imu</code><div class="track"><span style="--w:1.5%;--c:#008b9a"></span></div><strong>126</strong></div>
-        <div class="signal"><code>depth</code><div class="track"><span style="--w:11.7%;--c:#0a7f55"></span></div><strong>980</strong></div>
-        <div class="signal"><code>video</code><div class="track"><span style="--w:49.1%;--c:#b65b04"></span></div><strong>4,116</strong></div>
-        <div class="signal"><code>language</code><div class="track"><span style="--w:10.7%;--c:#7b5d12"></span></div><strong>896</strong></div>
-        <div class="signal"><code>static</code><div class="track"><span style="--w:1.7%;--c:#475569"></span></div><strong>139</strong></div>
+        <div class="signal"><code>mocap</code><div class="track"><span style="--w:25.3%;--c:#1f6c9f"></span></div><strong>2,121</strong></div>
+        <div class="signal"><code>camera+imu</code><div class="track"><span style="--w:1.5%;--c:#2e7775"></span></div><strong>126</strong></div>
+        <div class="signal"><code>depth</code><div class="track"><span style="--w:11.7%;--c:#346538"></span></div><strong>980</strong></div>
+        <div class="signal"><code>video</code><div class="track"><span style="--w:49.1%;--c:#956400"></span></div><strong>4,116</strong></div>
+        <div class="signal"><code>language</code><div class="track"><span style="--w:10.7%;--c:#6f5b21"></span></div><strong>896</strong></div>
+        <div class="signal"><code>static</code><div class="track"><span style="--w:1.7%;--c:#6f716c"></span></div><strong>139</strong></div>
       </div>
     </div>
   </header>
 
-  <main>
+  <main id="main">
     <section id="pipeline">
       <div class="wrap">
         <div class="section-head">
-          <h2>Pipeline from raw episode to task suite.</h2>
-          <p>Every script works from the same data contract: aligned multimodal windows, explicit labels, cached feature extraction, and artifact-first evaluation.</p>
+          <h2>From raw episode to checked artifacts.</h2>
+          <p>Every script works from one data contract: aligned multimodal windows, explicit labels, cached feature extraction, and artifact-first evaluation.</p>
         </div>
         <div class="two-col">
           <img class="pipeline-image" src="assets/pipeline_diagram.svg" alt="Verified Ropedia all-modality pipeline diagram">
@@ -477,8 +558,8 @@
     <section id="models">
       <div class="wrap">
         <div class="section-head">
-          <h2>Baselines with transparent artifacts.</h2>
-          <p>Motion-only and all-modality classifiers use the same lightweight softmax model so the comparison stays readable on a laptop.</p>
+          <h2>Small baselines, no hidden machinery.</h2>
+          <p>Motion-only and all-modality classifiers use lightweight heads so the comparison stays readable on a laptop and easy to audit.</p>
         </div>
         <div class="models">
           <article class="model"><h3>Motion-only action</h3><span class="score">0.9688</span><span class="meta">macro-F1, 18 classes</span></article>
@@ -493,7 +574,7 @@
     <section id="architectures">
       <div class="wrap">
         <div class="section-head">
-          <h2>The 12 tasks use three minimal head families.</h2>
+          <h2>The 12 tasks share three head families.</h2>
           <p>The diagram separates the shared episode-window feature pipeline from the task-specific heads: linear softmax, ridge projection/regression, and multi-label logistic regression.</p>
         </div>
         <img class="architecture-image" src="assets/task_architectures.svg" alt="Verified minimal architecture diagram for all 12 Ropedia episode tasks">
@@ -504,7 +585,7 @@
       <div class="wrap">
         <div class="section-head">
           <h2>Single-episode tasks, deliberately scoped.</h2>
-          <p>These tasks turn one episode into a lab for temporal structure, modality alignment, retrieval, reconstruction, and action understanding.</p>
+          <p>The suite treats one episode as a controlled lab for temporal structure, modality alignment, retrieval, reconstruction, and action understanding.</p>
         </div>
         <img class="task-suite-image" src="assets/task_suite_infographic.png" alt="ChatGPT-image-backed infographic showing all 12 Ropedia episode tasks">
         <div class="task-toolbar" aria-label="Task filters">
@@ -515,18 +596,18 @@
           <button class="filter" data-filter="diagnostic">Diagnostic</button>
         </div>
         <div class="task-grid" id="taskGrid">
-          <article class="task-card" data-kind="supervised"><div class="task-top"><span class="task-name">timeline_action</span><span class="tag supervised">supervised</span></div><p>All modalities to current action label. Chronological split exposes unseen future actions.</p><div class="metric-row"><span>macro-F1</span><strong>0.0500</strong></div><div class="mini-bar"><span style="--w:5%;--c:#1f63e9"></span></div></article>
-          <article class="task-card" data-kind="supervised"><div class="task-top"><span class="task-name">timeline_subtask</span><span class="tag supervised">supervised</span></div><p>All modalities to current subtask label. Useful for segmentation diagnostics.</p><div class="metric-row"><span>macro-F1</span><strong>0.0495</strong></div><div class="mini-bar"><span style="--w:5%;--c:#1f63e9"></span></div></article>
-          <article class="task-card" data-kind="diagnostic"><div class="task-top"><span class="task-name">transition_detection</span><span class="tag diagnostic">diagnostic</span></div><p>Predict steady vs action boundary. Highlights task-transition localization quality.</p><div class="metric-row"><span>macro-F1</span><strong>0.6552</strong></div><div class="mini-bar"><span style="--w:65.5%;--c:#b65b04"></span></div></article>
-          <article class="task-card" data-kind="supervised"><div class="task-top"><span class="task-name">next_action</span><span class="tag supervised">supervised</span></div><p>Current multimodal window to action 20 frames later. Tests short-horizon task flow.</p><div class="metric-row"><span>macro-F1</span><strong>0.0593</strong></div><div class="mini-bar"><span style="--w:5.9%;--c:#1f63e9"></span></div></article>
-          <article class="task-card" data-kind="forecast"><div class="task-top"><span class="task-name">hand_trajectory_forecast</span><span class="tag forecast">forecast</span></div><p>Predict future left/right hand 3D joints. Closer to imitation-learning style signals.</p><div class="metric-row"><span>MPJPE</span><strong>0.8223</strong></div><div class="mini-bar"><span style="--w:48%;--c:#0a7f55"></span></div></article>
-          <article class="task-card" data-kind="supervised"><div class="task-top"><span class="task-name">contact_prediction</span><span class="tag supervised">supervised</span></div><p>Non-contact modalities to binary contact. Degenerate in this sample because one class dominates.</p><div class="metric-row"><span>accuracy</span><strong>1.0000</strong></div><div class="mini-bar"><span style="--w:100%;--c:#0a7f55"></span></div></article>
-          <article class="task-card" data-kind="supervised"><div class="task-top"><span class="task-name">object_relevance</span><span class="tag supervised">supervised</span></div><p>Predict relevant object set from non-caption modalities.</p><div class="metric-row"><span>micro-F1</span><strong>0.1839</strong></div><div class="mini-bar"><span style="--w:18.4%;--c:#1f63e9"></span></div></article>
-          <article class="task-card" data-kind="retrieval"><div class="task-top"><span class="task-name">caption_grounding</span><span class="tag retrieval">retrieval</span></div><p>Caption objects/interaction query to matching sensor window.</p><div class="metric-row"><span>MRR</span><strong>0.0172</strong></div><div class="mini-bar"><span style="--w:1.7%;--c:#008b9a"></span></div></article>
-          <article class="task-card" data-kind="retrieval"><div class="task-top"><span class="task-name">cross_modal_retrieval</span><span class="tag retrieval">retrieval</span></div><p>Motion/IMU/camera query retrieves matching depth/video window. Strongest single-episode signal.</p><div class="metric-row"><span>top-5</span><strong>0.3764</strong></div><div class="mini-bar"><span style="--w:37.6%;--c:#008b9a"></span></div></article>
-          <article class="task-card" data-kind="forecast"><div class="task-top"><span class="task-name">modality_reconstruction</span><span class="tag forecast">forecast</span></div><p>Motion/IMU/camera to depth/video feature vector.</p><div class="metric-row"><span>R2</span><strong>-0.0160</strong></div><div class="mini-bar"><span style="--w:2%;--c:#b42318"></span></div></article>
-          <article class="task-card" data-kind="diagnostic"><div class="task-top"><span class="task-name">temporal_order</span><span class="tag diagnostic">diagnostic</span></div><p>Two adjacent windows to correct vs reversed order.</p><div class="metric-row"><span>F1</span><strong>0.5487</strong></div><div class="mini-bar"><span style="--w:54.9%;--c:#b65b04"></span></div></article>
-          <article class="task-card" data-kind="diagnostic"><div class="task-top"><span class="task-name">misalignment_detection</span><span class="tag diagnostic">diagnostic</span></div><p>Motion+visual pair to aligned vs shifted by eight windows.</p><div class="metric-row"><span>F1</span><strong>0.4866</strong></div><div class="mini-bar"><span style="--w:48.7%;--c:#b65b04"></span></div></article>
+          <article class="task-card" data-kind="supervised"><div class="task-top"><span class="task-name">timeline_action</span><span class="tag supervised">supervised</span></div><p>All modalities to current action label. Chronological split exposes unseen future actions.</p><div class="metric-row"><span>macro-F1</span><strong>0.0500</strong></div><div class="mini-bar"><span style="--w:5%;--c:#1f6c9f"></span></div></article>
+          <article class="task-card" data-kind="supervised"><div class="task-top"><span class="task-name">timeline_subtask</span><span class="tag supervised">supervised</span></div><p>All modalities to current subtask label. Useful for segmentation diagnostics.</p><div class="metric-row"><span>macro-F1</span><strong>0.0495</strong></div><div class="mini-bar"><span style="--w:5%;--c:#1f6c9f"></span></div></article>
+          <article class="task-card" data-kind="diagnostic"><div class="task-top"><span class="task-name">transition_detection</span><span class="tag diagnostic">diagnostic</span></div><p>Predict steady vs action boundary. Highlights task-transition localization quality.</p><div class="metric-row"><span>macro-F1</span><strong>0.6552</strong></div><div class="mini-bar"><span style="--w:65.5%;--c:#956400"></span></div></article>
+          <article class="task-card" data-kind="supervised"><div class="task-top"><span class="task-name">next_action</span><span class="tag supervised">supervised</span></div><p>Current multimodal window to action 20 frames later. Tests short-horizon task flow.</p><div class="metric-row"><span>macro-F1</span><strong>0.0593</strong></div><div class="mini-bar"><span style="--w:5.9%;--c:#1f6c9f"></span></div></article>
+          <article class="task-card" data-kind="forecast"><div class="task-top"><span class="task-name">hand_trajectory_forecast</span><span class="tag forecast">forecast</span></div><p>Predict future left/right hand 3D joints. Closer to imitation-learning style signals.</p><div class="metric-row"><span>MPJPE</span><strong>0.8223</strong></div><div class="mini-bar"><span style="--w:48%;--c:#346538"></span></div></article>
+          <article class="task-card" data-kind="supervised"><div class="task-top"><span class="task-name">contact_prediction</span><span class="tag supervised">supervised</span></div><p>Non-contact modalities to binary contact. Degenerate in this sample because one class dominates.</p><div class="metric-row"><span>accuracy</span><strong>1.0000</strong></div><div class="mini-bar"><span style="--w:100%;--c:#346538"></span></div></article>
+          <article class="task-card" data-kind="supervised"><div class="task-top"><span class="task-name">object_relevance</span><span class="tag supervised">supervised</span></div><p>Predict relevant object set from non-caption modalities.</p><div class="metric-row"><span>micro-F1</span><strong>0.1839</strong></div><div class="mini-bar"><span style="--w:18.4%;--c:#1f6c9f"></span></div></article>
+          <article class="task-card" data-kind="retrieval"><div class="task-top"><span class="task-name">caption_grounding</span><span class="tag retrieval">retrieval</span></div><p>Caption objects/interaction query to matching sensor window.</p><div class="metric-row"><span>MRR</span><strong>0.0172</strong></div><div class="mini-bar"><span style="--w:1.7%;--c:#2e7775"></span></div></article>
+          <article class="task-card" data-kind="retrieval"><div class="task-top"><span class="task-name">cross_modal_retrieval</span><span class="tag retrieval">retrieval</span></div><p>Motion/IMU/camera query retrieves matching depth/video window. Strongest single-episode signal.</p><div class="metric-row"><span>top-5</span><strong>0.3764</strong></div><div class="mini-bar"><span style="--w:37.6%;--c:#2e7775"></span></div></article>
+          <article class="task-card" data-kind="forecast"><div class="task-top"><span class="task-name">modality_reconstruction</span><span class="tag forecast">forecast</span></div><p>Motion/IMU/camera to depth/video feature vector.</p><div class="metric-row"><span>R2</span><strong>-0.0160</strong></div><div class="mini-bar"><span style="--w:2%;--c:#9f2f2d"></span></div></article>
+          <article class="task-card" data-kind="diagnostic"><div class="task-top"><span class="task-name">temporal_order</span><span class="tag diagnostic">diagnostic</span></div><p>Two adjacent windows to correct vs reversed order.</p><div class="metric-row"><span>F1</span><strong>0.5487</strong></div><div class="mini-bar"><span style="--w:54.9%;--c:#956400"></span></div></article>
+          <article class="task-card" data-kind="diagnostic"><div class="task-top"><span class="task-name">misalignment_detection</span><span class="tag diagnostic">diagnostic</span></div><p>Motion+visual pair to aligned vs shifted by eight windows.</p><div class="metric-row"><span>F1</span><strong>0.4866</strong></div><div class="mini-bar"><span style="--w:48.7%;--c:#956400"></span></div></article>
         </div>
       </div>
     </section>
@@ -534,8 +615,8 @@
     <section id="features">
       <div class="wrap">
         <div class="section-head">
-          <h2>All-modality features are inspectable.</h2>
-          <p>The point is not hidden complexity. Every block has a clear source modality and an explicit dimensional footprint.</p>
+          <h2>Every feature block has a source.</h2>
+          <p>The point is not hidden complexity. Every block has a source modality, a dimensional footprint, and a manifest entry.</p>
         </div>
         <img class="chart" src="assets/charts/feature_blocks.svg" alt="All modality feature block chart">
       </div>
@@ -544,7 +625,7 @@
     <section id="diagnostics">
       <div class="wrap">
         <div class="section-head">
-          <h2>Diagnostics separate easy wins from real gaps.</h2>
+          <h2>Diagnostics separate memorization from signal.</h2>
           <p>The charts make the main lesson visible: within-episode supervised labels are easy under some splits, while retrieval, grounding, forecasting, and alignment remain the useful probes.</p>
         </div>
         <div class="chart-grid">
@@ -557,8 +638,8 @@
     <section id="artifacts">
       <div class="wrap">
         <div class="section-head">
-          <h2>Everything important is checked in.</h2>
-          <p>Metrics, predictions, confusion matrices, feature manifests, model weights, and derived window artifacts are committed so the repo is reviewable without rerunning first.</p>
+          <h2>Where the evidence lives.</h2>
+          <p>Metrics, predictions, confusion matrices, manifests, model weights, and derived window artifacts are committed so the repo is reviewable before rerunning anything.</p>
         </div>
         <div class="artifact-grid">
           <article class="artifact"><h3>Task-suite report</h3><p>One JSON file with every task metric and split detail.</p><a href="https://github.com/ChaoYue0307/ropedia-episode-task-suite/blob/main/results/episode_task_suite/summary_report.json">summary_report.json</a></article>
@@ -630,6 +711,7 @@ python scripts/episode_task_suite.py --workspace /path/to/workspace</code></pre>
         setTimeout(() => button.textContent = previous, 1300);
       });
     });
+
   </script>
 </body>
 </html>