| <html lang="en"><head><meta charset="utf-8"><meta name="viewport" content="width=device-width,initial-scale=1"><style> | |
| body{font-family:Inter,system-ui,sans-serif;margin:0;background:#f7f5ff;color:#211a3a}main{max-width:1080px;margin:auto;padding:28px}h1{font-size:2.05rem;margin:.2rem 0}.lead{color:#5b4a84}.grid{display:grid;grid-template-columns:repeat(auto-fit,minmax(205px,1fr));gap:14px;margin:22px 0}.card{background:white;border:1px solid #dcd4f3;border-radius:14px;padding:16px;box-shadow:0 3px 12px #33236418}.n{font-size:1.7rem;font-weight:750;color:#6d28d9}.scope{border-left:5px solid #d97706;background:#fff8e8;padding:14px 18px;border-radius:8px}img{width:100%;background:white;border-radius:14px;border:1px solid #dcd4f3}code{background:#eee9fb;padding:.1rem .3rem;border-radius:4px}</style></head><body><main> | |
| <p class="lead">Independent reproduction · arXiv 2605.01082v1 · OpenReview mrtg4NmvAe</p><h1>Networked Information Aggregation for Binary Classification</h1> | |
| <p>Exact BCE identities, sequential logit-passing, and depth-bottleneck evidence with primary sources pinned.</p> | |
| <section class="grid"><div class="card"><div class="n">22/22</div><div>scientific gates</div></div><div class="card"><div class="n">48</div><div>network panels</div></div><div class="card"><div class="n">56</div><div>lower-bound cells</div></div><div class="card"><div class="n">48/48</div><div>probability control worse</div></div></section> | |
| <img src="outputs/networked_information_audit.png" alt="BCE identity, protocol, and lower-bound diagnostics"> | |
| <p class="scope"><strong>Scope:</strong> The upper result requires M-coverage and bounded features/coefficients. The lower construction has D=kp and p≤k−1; “matching” means the depth bottleneck, not identical exponents. The protocol passes logits, not probabilities.</p> | |
| </main></body></html> | |