Spaces:
Running
Running
| <html lang="en"> | |
| <head> | |
| <meta charset="utf-8"> | |
| <meta name="viewport" content="width=device-width, initial-scale=1"> | |
| <title>Linear Contracts Reproduction Evidence</title> | |
| <style> | |
| :root { color-scheme: light; --ink: #17202a; --muted: #5d6d7e; --paper: #f7f4ec; --card: #ffffff; --blue: #174a7e; --green: #18794e; --gold: #a56400; } | |
| * { box-sizing: border-box; } | |
| body { margin: 0; background: var(--paper); color: var(--ink); font-family: ui-sans-serif, system-ui, sans-serif; } | |
| main { max-width: 1180px; margin: auto; padding: 42px; } | |
| header { border-left: 8px solid var(--blue); padding-left: 22px; } | |
| h1 { margin: 0 0 8px; font-size: clamp(2rem, 5vw, 4rem); line-height: 1.02; } | |
| h2 { margin: 0 0 12px; color: var(--blue); } | |
| p { line-height: 1.5; } | |
| .sub { color: var(--muted); font-size: 1.12rem; } | |
| .grid { display: grid; grid-template-columns: repeat(3, minmax(0, 1fr)); gap: 18px; margin: 34px 0; } | |
| .card { background: var(--card); border: 1px solid #d8d3c8; border-radius: 14px; padding: 22px; box-shadow: 0 8px 24px #17202a10; } | |
| .metric { font-size: 2.35rem; font-weight: 800; color: var(--green); margin: 6px 0; } | |
| .wide { grid-column: span 2; } | |
| .caveat { border-top: 5px solid var(--gold); } | |
| code { background: #eef2f5; padding: 2px 5px; border-radius: 4px; } | |
| footer { color: var(--muted); border-top: 1px solid #cbc5b8; padding-top: 16px; } | |
| @media (max-width: 760px) { main { padding: 24px; } .grid { grid-template-columns: 1fr; } .wide { grid-column: auto; } } | |
| </style> | |
| </head> | |
| <body> | |
| <main> | |
| <header> | |
| <h1>The Optimal Sample Complexity of Linear Contracts</h1> | |
| <p class="sub">Clean-room CPU reproduction 路 OpenReview ry5HitnXzc 路 exact finite evidence at the paper's stated constants</p> | |
| </header> | |
| <section class="grid"> | |
| <article class="card"><h2>Upper bound</h2><div class="metric">0 / 163,840</div><p>Epsilon-regret violations across four finite-model families and five epsilon/delta settings.</p></article> | |
| <article class="card"><h2>Failure bound</h2><div class="metric">0.0003657</div><p>Largest one-sided 95% upper bound, below the smallest tested <code>delta=0.01</code>.</p></article> | |
| <article class="card"><h2>Verification</h2><div class="metric">15 / 15</div><p>Deterministic tests passing on local CPU.</p></article> | |
| <article class="card wide"><h2>Matching lower-bound mechanism</h2><p>The cited two-type/two-action construction was rebuilt directly. Both wrong-region gaps equal <code>2 epsilon</code>. Across 16 exact Bayes tests, <code>N epsilon^2 / ln(1/delta)</code> stays between <strong>0.00834</strong> and <strong>0.02153</strong>, exercising the claimed rate.</p></article> | |
| <article class="card"><h2>Negative control</h2><p>Shared latent types violate i.i.d. sampling and fail, showing the positive certificate is not an always-success check.</p></article> | |
| <article class="card wide caveat"><h2>Scope boundary</h2><p>Finite executable certificates support the theorem mechanisms, constants, and scaling; they cannot prove a universal distribution-free theorem. The paper states constant 6912 but later uses 13824 in its proof, and the cited lower-bound probabilities require epsilon below 1/8.</p></article> | |
| <article class="card"><h2>Reproduction bundle</h2><p>Source, configurations, tests, deterministic CSV/JSON results, requirements, and source audit are packaged together.</p></article> | |
| </section> | |
| <footer>Artifact project: repro-the-optimal-sample-complexity-of-linear-contracts</footer> | |
| </main> | |
| </body> | |
| </html> | |