| <!DOCTYPE html> |
| <html lang="en"> |
| <head> |
| <meta charset="UTF-8"> |
| <meta name="viewport" content="width=device-width, initial-scale=1.0"> |
| <title>Recurrent-Depth Parallel Sampler Reproduction (h7WBYYJF1Q)</title> |
| <style> |
| :root { |
| --bg-dark: #0b0f19; |
| --bg-card: #131b2e; |
| --bg-card-hover: #1c2744; |
| --border-color: #233152; |
| --text-main: #f1f5f9; |
| --text-muted: #94a3b8; |
| --accent-blue: #38bdf8; |
| --accent-purple: #c084fc; |
| --status-partial-bg: #451a03; |
| --status-partial-text: #fbbf24; |
| --status-partial-border: #92400e; |
| --status-unavail-bg: #311b92; |
| --status-unavail-text: #d8b4fe; |
| --status-unavail-border: #6b21a8; |
| --font-sans: system-ui, -apple-system, BlinkMacSystemFont, 'Segoe UI', Roboto, sans-serif; |
| --font-mono: 'SFMono-Regular', Consolas, 'Liberation Mono', Menlo, monospace; |
| } |
| |
| * { box-sizing: border-box; margin: 0; padding: 0; } |
| body { |
| background-color: var(--bg-dark); |
| color: var(--text-main); |
| font-family: var(--font-sans); |
| line-height: 1.6; |
| padding: 2rem 1rem; |
| } |
| |
| .container { |
| max-width: 1100px; |
| margin: 0 auto; |
| } |
| |
| header { |
| margin-bottom: 2.5rem; |
| border-bottom: 1px solid var(--border-color); |
| padding-bottom: 1.5rem; |
| } |
| |
| h1 { |
| font-size: 2.2rem; |
| font-weight: 700; |
| background: linear-gradient(135deg, #38bdf8 0%, #c084fc 100%); |
| -webkit-background-clip: text; |
| -webkit-text-fill-color: transparent; |
| margin-bottom: 0.5rem; |
| } |
| |
| .meta-badges { |
| display: flex; |
| flex-wrap: wrap; |
| gap: 0.75rem; |
| margin-top: 1rem; |
| } |
| |
| .badge { |
| font-family: var(--font-mono); |
| font-size: 0.8rem; |
| padding: 0.35rem 0.75rem; |
| border-radius: 6px; |
| background: var(--bg-card); |
| border: 1px solid var(--border-color); |
| color: var(--accent-blue); |
| } |
| |
| .claims-grid { |
| display: grid; |
| grid-template-columns: 1fr; |
| gap: 1.5rem; |
| margin-bottom: 2.5rem; |
| } |
| |
| @media (min-width: 768px) { |
| .claims-grid { |
| grid-template-columns: 1fr 1fr; |
| } |
| } |
| |
| .claim-card { |
| background: var(--bg-card); |
| border: 1px solid var(--border-color); |
| border-radius: 12px; |
| padding: 1.5rem; |
| transition: transform 0.2s ease, border-color 0.2s ease; |
| } |
| |
| .claim-card:hover { |
| transform: translateY(-2px); |
| border-color: var(--accent-blue); |
| } |
| |
| .claim-header { |
| display: flex; |
| justify-content: space-between; |
| align-items: center; |
| margin-bottom: 1rem; |
| } |
| |
| .claim-title { |
| font-size: 1.1rem; |
| font-weight: 600; |
| } |
| |
| .status-badge { |
| font-family: var(--font-mono); |
| font-size: 0.75rem; |
| font-weight: 600; |
| text-transform: uppercase; |
| padding: 0.25rem 0.6rem; |
| border-radius: 4px; |
| } |
| |
| .status-partial { |
| background: var(--status-partial-bg); |
| color: var(--status-partial-text); |
| border: 1px solid var(--status-partial-border); |
| } |
| |
| .status-unavailable { |
| background: var(--status-unavail-bg); |
| color: var(--status-unavail-text); |
| border: 1px solid var(--status-unavail-border); |
| } |
| |
| .claim-text { |
| font-size: 0.95rem; |
| color: var(--text-muted); |
| margin-bottom: 1rem; |
| font-style: italic; |
| background: rgba(0,0,0,0.2); |
| padding: 0.75rem; |
| border-radius: 6px; |
| border-left: 3px solid var(--accent-blue); |
| } |
| |
| .section-title { |
| font-size: 1.5rem; |
| font-weight: 600; |
| margin: 2rem 0 1rem 0; |
| color: var(--text-main); |
| display: flex; |
| align-items: center; |
| gap: 0.5rem; |
| } |
| |
| .table-container { |
| overflow-x: auto; |
| background: var(--bg-card); |
| border: 1px solid var(--border-color); |
| border-radius: 12px; |
| margin-bottom: 2rem; |
| } |
| |
| table { |
| width: 100%; |
| border-collapse: collapse; |
| font-size: 0.9rem; |
| text-align: left; |
| } |
| |
| th, td { |
| padding: 0.85rem 1.25rem; |
| border-bottom: 1px solid var(--border-color); |
| } |
| |
| th { |
| background: rgba(255,255,255,0.03); |
| font-weight: 600; |
| color: var(--accent-blue); |
| } |
| |
| code { |
| font-family: var(--font-mono); |
| font-size: 0.85rem; |
| color: var(--accent-purple); |
| } |
| |
| .provenance-card { |
| background: var(--bg-card); |
| border: 1px solid var(--border-color); |
| border-radius: 12px; |
| padding: 1.5rem; |
| font-family: var(--font-mono); |
| font-size: 0.85rem; |
| margin-bottom: 2rem; |
| word-break: break-all; |
| } |
| |
| .download-links { |
| display: flex; |
| flex-wrap: wrap; |
| gap: 1rem; |
| margin-bottom: 2.5rem; |
| } |
| |
| .download-btn { |
| display: inline-flex; |
| align-items: center; |
| gap: 0.5rem; |
| padding: 0.6rem 1.2rem; |
| background: var(--bg-card); |
| border: 1px solid var(--border-color); |
| border-radius: 8px; |
| color: var(--text-main); |
| text-decoration: none; |
| font-size: 0.9rem; |
| font-weight: 500; |
| transition: all 0.2s ease; |
| } |
| |
| .download-btn:hover { |
| background: var(--bg-card-hover); |
| border-color: var(--accent-blue); |
| color: var(--accent-blue); |
| } |
| |
| .limitations-box { |
| background: rgba(239, 68, 68, 0.08); |
| border: 1px solid rgba(239, 68, 68, 0.3); |
| border-radius: 12px; |
| padding: 1.5rem; |
| margin-bottom: 2.5rem; |
| } |
| |
| .limitations-box h3 { |
| color: #f87171; |
| font-size: 1.1rem; |
| margin-bottom: 0.75rem; |
| } |
| |
| .limitations-box ul { |
| list-style-type: square; |
| padding-left: 1.25rem; |
| color: var(--text-muted); |
| } |
| |
| footer { |
| text-align: center; |
| color: var(--text-muted); |
| font-size: 0.85rem; |
| border-top: 1px solid var(--border-color); |
| padding-top: 1.5rem; |
| } |
| </style> |
| </head> |
| <body> |
| <div class="container"> |
| <header> |
| <h1>Efficient Parallel Samplers for Recurrent-Depth Models</h1> |
| <p style="color: var(--text-muted);">ICML 2026 Reproduction Challenge • Independent Artifact Evidence Presentation</p> |
| <div class="meta-badges"> |
| <span class="badge">Paper: h7WBYYJF1Q</span> |
| <span class="badge">arXiv: 2510.14961v1</span> |
| <span class="badge">Attempt: 534db42c-5b16-4f00-9a7d-a47056fc9dd4</span> |
| <span class="badge">Mode: Deterministic CPU Audit</span> |
| </div> |
| </header> |
|
|
| <section class="claims-grid"> |
| <div class="claim-card"> |
| <div class="claim-header"> |
| <div class="claim-title">Claim 1: Wavefront Sampler</div> |
| <span class="status-badge status-partial">PARTIAL</span> |
| </div> |
| <div class="claim-text">"The sampler decodes new tokens every forward pass while refining latent states for those tokens in parallel through recurrent depth (Section 3.1)."</div> |
| <p style="font-size: 0.9rem; color: var(--text-muted);"> |
| <strong>Findings:</strong> Released sampler AST (<code>recpre/raven_modeling_minimal.py</code>) confirms <code>generate_diffusion_style</code> applies <code>inner_recurrence</code> steps per outer iteration across active states, decodes logits, and appends new tokens. However, decoding occurs after the inner loop per outer step rather than after every individual inner step. |
| </p> |
| </div> |
|
|
| <div class="claim-card"> |
| <div class="claim-header"> |
| <div class="claim-title">Claim 2: Expressiveness Theorem</div> |
| <span class="status-badge status-unavailable">UNAVAILABLE</span> |
| </div> |
| <div class="claim-text">"The paper proves the sampler is strictly more expressive than baseline autoregressive generation under the same time budget on modern hardware (Theorem 4.2)."</div> |
| <p style="font-size: 0.9rem; color: var(--text-muted);"> |
| <strong>Findings:</strong> Source TeX audit reveals a citation mismatch: Theorem 4.2 is the prefilling result, while same-runtime decoding expressiveness is Theorem 4.4. No complete independently checkable proof of the decoding theorem is included in the v1 source. |
| </p> |
| </div> |
| </section> |
|
|
| <h2 class="section-title">⚡ Wavefront Mechanism Schedule Simulation</h2> |
| <div class="table-container"> |
| <table> |
| <thead> |
| <tr> |
| <th>Outer Step</th> |
| <th>Active Positions Before</th> |
| <th>Recurrence Updates</th> |
| <th>Decoded Position</th> |
| <th>Candidate Positions</th> |
| <th>Retained Positions</th> |
| <th>Active Width / Bound</th> |
| </tr> |
| </thead> |
| <tbody> |
| <tr> |
| <td>Step 1</td> |
| <td><code>[0]</code></td> |
| <td>+4 updates</td> |
| <td>Position 0</td> |
| <td><code>[1]</code></td> |
| <td><code>[1]</code></td> |
| <td>2 / 8</td> |
| </tr> |
| <tr> |
| <td>Step 2</td> |
| <td><code>[0, 1]</code></td> |
| <td>+4 updates</td> |
| <td>Position 1</td> |
| <td><code>[2]</code></td> |
| <td><code>[2]</code></td> |
| <td>3 / 8</td> |
| </tr> |
| <tr> |
| <td>Step 3</td> |
| <td><code>[0, 1, 2]</code></td> |
| <td>+4 updates</td> |
| <td>Position 2</td> |
| <td><code>[3]</code></td> |
| <td><code>[3]</code></td> |
| <td>4 / 8</td> |
| </tr> |
| <tr> |
| <td>Step 4</td> |
| <td><code>[0, 1, 2, 3]</code></td> |
| <td>+4 updates</td> |
| <td>Position 3</td> |
| <td><code>[4]</code></td> |
| <td><code>[4]</code></td> |
| <td>5 / 8</td> |
| </tr> |
| <tr> |
| <td>Step 5</td> |
| <td><code>[0, 1, 2, 3, 4]</code></td> |
| <td>+4 updates</td> |
| <td>Position 4</td> |
| <td><code>[5]</code></td> |
| <td><code>[5]</code></td> |
| <td>6 / 8</td> |
| </tr> |
| <tr> |
| <td>Step 6</td> |
| <td><code>[0, 1, 2, 3, 4, 5]</code></td> |
| <td>+4 updates</td> |
| <td>Position 5</td> |
| <td><code>[6]</code></td> |
| <td><code>[6]</code></td> |
| <td>7 / 8</td> |
| </tr> |
| <tr> |
| <td>Step 7</td> |
| <td><code>[0, 1, 2, 3, 4, 5, 6]</code></td> |
| <td>+4 updates</td> |
| <td>Position 6</td> |
| <td><code>[7]</code></td> |
| <td><code>[7]</code></td> |
| <td>8 / 8</td> |
| </tr> |
| <tr> |
| <td>Step 8</td> |
| <td><code>[0, 1, 2, 3, 4, 5, 6, 7]</code></td> |
| <td>+4 updates</td> |
| <td>Position 7</td> |
| <td><code>[8]</code></td> |
| <td><code>[]</code></td> |
| <td>8 / 8</td> |
| </tr> |
| </tbody> |
| </table> |
| </div> |
|
|
| <h2 class="section-title">📜 Provenance & Pinned Input Hashes</h2> |
| <div class="provenance-card"> |
| <p style="margin-bottom: 0.5rem; color: var(--accent-blue);"><strong>Provenance Token:</strong></p> |
| <p style="margin-bottom: 1rem;"><code>arxiv:2510.14961v1+pdf-sha256:74e7985abe41ee2a75914a65e3778a15353fb0c0964d6ea34e7bfeb1f18312c8+source-sha256:60a795d123a2d2d642971834b6e0cba6dda80b5dfcd539f78d01639582d9c41d+github:seal-rg/recurrent-pretraining@1ea7220ec7eb42d13e89db0663df254d0bcdc28e+git-blob:recpre/raven_modeling_minimal.py@0e83a0766644df9113a8923f43350c6a1b5a182c</code></p> |
|
|
| <p style="margin-bottom: 0.5rem; color: var(--accent-purple);"><strong>Input & Claim Hashes:</strong></p> |
| <p>• Claim 1 SHA-256: <code>d0da87ee16f7485d3dff369e7465f66299c55ac003a54e1cf8c00b3a0ad8b265</code></p> |
| <p>• Claim 2 SHA-256: <code>2e15221c8b5516b0ab705e29a3d7c5d924ed5f0187c970a0caf60a1402757804</code></p> |
| <p>• arxiv_submission.tex SHA-256: <code>cdc058830d1e51f631e4fb8d1f2de0b79de91670fd4111646fe624f8c258d3b8</code></p> |
| <p>• raven_modeling_minimal.py SHA-256: <code>18fcacd53fb5696a76c0d3bda44480f2f3900aa9659c137a08962c593a9a9e42</code></p> |
| <p>• raven_modeling_minimal.py Git Blob: <code>0e83a0766644df9113a8923f43350c6a1b5a182c</code></p> |
| <p>• LICENSE (Apache-2.0) SHA-256: <code>bc6c264d8ba4450599cf95c4699c6b82142f32ca1ecd91011c17b50a5a36a2f5</code></p> |
| <p>• ATTRIBUTION.md (CC-BY-4.0) SHA-256: <code>79775b50c72988b90eae75ef87e9d4df9dbd0bfceefaed60b398656a88d8a735</code></p> |
| </div> |
|
|
| <h2 class="section-title">📥 Download Evidence Bundles</h2> |
| <div class="download-links"> |
| <a class="download-btn" href="evidence/manifest.json" download>📄 manifest.json</a> |
| <a class="download-btn" href="evidence/claim-1-wavefront.json" download>📄 claim-1-wavefront.json</a> |
| <a class="download-btn" href="evidence/claim-2-theorem-audit.json" download>📄 claim-2-theorem-audit.json</a> |
| <a class="download-btn" href="evidence/results.json" download>📄 results.json</a> |
| <a class="download-btn" href="REPORT.md" download>📝 REPORT.md</a> |
| </div> |
|
|
| <div class="limitations-box"> |
| <h3>⚠️ Scope & Limitations</h3> |
| <ul> |
| <li><strong>No Model Execution:</strong> 3.5B Huginn-0125 weights were not loaded or run.</li> |
| <li><strong>No GPU Benchmarking:</strong> Modern GPU hardware timing and reported ~5x speedups were not measured.</li> |
| <li><strong>Audit Only:</strong> Evidence statuses reflect static AST, schedule, and TeX audits.</li> |
| </ul> |
| </div> |
|
|
| <footer> |
| <p>ICML 2026 Reproduction Challenge • Generated deterministically</p> |
| </footer> |
| </div> |
| </body> |
| </html> |
|
|
