index.html

<!DOCTYPE html>
<html lang="en">
<head>
<meta charset="utf-8" />
<meta name="viewport" content="width=device-width, initial-scale=1.0" />
<meta name="description" content="Optitransfer: open models built to place at the top of their class on public leaderboards, from the work the open-source community has already published. One open model per class, every result measured against its base and published with its evidence." />
<meta name="keywords" content="open-weights LLM, model merging, capability consolidation, Qwen2.5, 7B, reproducible evaluation, benchmarks, leaderboards, crdt-merge" />
<meta name="author" content="Ryan Gillespie" />
<meta property="og:title" content="Optitransfer" />
<meta property="og:description" content="Open models built to place at the top of their class on public leaderboards, from the work the open-source community has already published." />
<meta property="og:type" content="website" />
<meta property="og:url" content="https://huggingface.co/OptiTransferData" />
<title>Optitransfer</title>
<style>
  .container, .container * { margin: 0; padding: 0; box-sizing: border-box; }
  /* Hugging Face scopes this stylesheet to the card's wrapper, so :root and body never match: every variable and
     base style lives on .container. Colours inherit from the page so the card reads in the light and dark themes. */
  .container {
    --surface: rgba(127, 127, 127, 0.08);
    --border: rgba(127, 127, 127, 0.28);
    --mono: 'SF Mono', 'Cascadia Code', 'Fira Code', Consolas, monospace;
    --sans: -apple-system, BlinkMacSystemFont, 'Segoe UI', system-ui, sans-serif;
    font-family: var(--sans);
    line-height: 1.6;
    -webkit-font-smoothing: antialiased;
    max-width: 760px;
    margin: 0 auto;
    padding: 3rem 1.5rem;
  }
  .header {
    text-align: center;
    margin-bottom: 4rem;
  }
  .header h1 {
    font-size: 1.75rem;
    font-weight: 500;
    letter-spacing: 0.08em;
    text-transform: uppercase;
    color: inherit;
    margin-bottom: 1rem;
  }
  .header .tagline {
    font-size: 1.05rem;
    color: inherit;
    opacity: 0.68;
    font-weight: 400;
    max-width: 480px;
    margin: 0 auto;
  }
  .header .install {
    margin-top: 2rem;
    display: inline-block;
    font-family: var(--mono);
    font-size: 0.875rem;
    color: inherit;
    background: var(--surface);
    border: 1px solid var(--border);
    padding: 0.6rem 1.5rem;
    border-radius: 6px;
  }
  .badges {
    display: flex;
    justify-content: center;
    gap: 0.5rem;
    flex-wrap: wrap;
    margin-top: 1.5rem;
  }
  .badges a { opacity: 0.85; transition: opacity 0.2s; }
  .badges a:hover { opacity: 1; }
  .section { margin-bottom: 3rem; }
  .section h2 {
    font-size: 0.75rem;
    font-weight: 600;
    letter-spacing: 0.12em;
    text-transform: uppercase;
    color: inherit;
    opacity: 0.68;
    margin-bottom: 1.25rem;
    padding-bottom: 0.5rem;
    border-bottom: 1px solid var(--border);
  }
  .stats {
    display: grid;
    grid-template-columns: repeat(4, 1fr);
    gap: 1rem;
    text-align: center;
  }
  .stat { background: var(--surface); border: 1px solid var(--border); border-radius: 8px; padding: 1rem 0.5rem; }
  .stat .number {
    font-family: var(--mono);
    font-size: 1.5rem;
    font-weight: 600;
    color: inherit;
  }
  .stat .label {
    font-size: 0.7rem;
    font-weight: 500;
    letter-spacing: 0.08em;
    text-transform: uppercase;
    color: inherit;
    opacity: 0.68;
    margin-top: 0.25rem;
  }
  .links {
    display: grid;
    grid-template-columns: repeat(auto-fit, minmax(200px, 1fr));
    gap: 1rem;
  }
  .link-card {
    background: var(--surface);
    border: 1px solid var(--border);
    border-radius: 8px;
    padding: 1.25rem;
    text-decoration: none;
    color: inherit;
    transition: border-color 0.2s, transform 0.15s;
    display: block;
  }
  .link-card:hover {
    border-color: inherit;
    transform: translateY(-1px);
  }
  .link-card .label {
    font-size: 0.7rem;
    font-weight: 600;
    letter-spacing: 0.1em;
    text-transform: uppercase;
    color: inherit;
    opacity: 0.68;
    margin-bottom: 0.35rem;
  }
  .link-card .value { font-size: 0.95rem; }
  .spaces { display: grid; gap: 0.75rem; }
  .space {
    display: flex;
    justify-content: space-between;
    align-items: center;
    background: var(--surface);
    border: 1px solid var(--border);
    border-radius: 8px;
    padding: 1rem 1.25rem;
    text-decoration: none;
    color: inherit;
    transition: border-color 0.2s;
  }
  a.space:hover { border-color: inherit; }
  .space .desc { max-width: 560px; }
  .space .name { font-weight: 500; font-size: 0.95rem; }
  .space .desc { font-size: 0.8rem; color: inherit;
    opacity: 0.68; }
  .space .arrow { color: inherit;
    opacity: 0.68; font-size: 1.1rem; }
  .research-list { list-style: none; padding: 0; }
  .research-list li {
    color: inherit;
    opacity: 0.68;
    font-size: 0.9rem;
    padding: 0.4rem 0;
    border-bottom: 1px solid var(--border);
  }
  .research-list li:last-child { border-bottom: none; }
  .research-list li::before {
    content: '';
    display: inline-block;
    width: 6px;
    height: 6px;
    border-radius: 50%;
    background: currentColor;
    margin-right: 0.75rem;
    vertical-align: middle;
  }
  .footer {
    text-align: center;
    color: inherit;
    opacity: 0.68;
    font-size: 0.8rem;
    margin-top: 2rem;
    padding-top: 2rem;
    border-top: 1px solid var(--border);
  }
  .footer a { color: inherit; text-decoration: underline; }
  .footer a:hover { text-decoration: underline; }
  @media (max-width: 600px) {
    .container { padding: 2rem 1rem; }
    .stats { grid-template-columns: repeat(2, 1fr); }
    .links { grid-template-columns: 1fr; }
  }
  .section p { font-size: 0.98rem; margin-bottom: 0.75rem; opacity: 0.9; }
  .section p strong { font-weight: 600; }
  .stats.three { grid-template-columns: repeat(3, 1fr); }
  .research-list li strong { font-weight: 600; }
</style>
</head>
<body>
<div class="container">

  <header class="header">
    <h1>Optitransfer</h1>
    <p class="tagline">
      Open models built to place at the top of their class on public leaderboards, from the work the open-source community has already published.
    </p>
    <div class="badges">
      <a href="https://huggingface.co/Optitransfer"><img src="https://img.shields.io/badge/models-converge%20v1%20%C2%B7%20v2-blue" alt="Models" /></a>
      <a href="https://arxiv.org/abs/2605.19373"><img src="https://img.shields.io/badge/arXiv-2605.19373-b31b1b" alt="Paper" /></a>
      <a href="https://arxiv.org/abs/2607.10305"><img src="https://img.shields.io/badge/arXiv-2607.10305-b31b1b" alt="Paper" /></a>
      <a href="https://pypi.org/project/crdt-merge/"><img src="https://img.shields.io/pypi/v/crdt-merge?label=crdt-merge&color=blue" alt="crdt-merge" /></a>
    </div>
  </header>

  <div class="section">
    <h2>At a Glance</h2>
    <div class="stats">
      <div class="stat"><div class="number">7B</div><div class="label">Class in progress</div></div>
      <div class="stat"><div class="number">2</div><div class="label">Public releases</div></div>
      <div class="stat"><div class="number">12</div><div class="label">Board benchmarks</div></div>
      <div class="stat"><div class="number">2</div><div class="label">Papers</div></div>
    </div>
  </div>

  <div class="section">
    <h2>The Thesis</h2>
    <p>Thousands of fine-tuned models sit on public repositories. Each adds a skill to a shared base model. Their training is already paid for.</p>
    <p><strong>Hypothesis:</strong> that capability can be consolidated into one model per class without pre-training, with trade-offs between skills controlled and capability then maximised across the board through iteration. Converge is the programme that tests it, one release at a time, in public.</p>
  </div>

  <div class="section">
    <h2>Where We Are</h2>
    <ul class="research-list">
      <li><strong>v2 against its base.</strong> 12 benchmarks: 2 improved, 2 regressed, 8 with no measurable difference.</li>
      <li><strong>Improved.</strong> GSM8K under HELM's protocol: 87.0 against 83.3 (+3.7, calibrated). MMLU under HELM's protocol: +0.47 (calibrated).</li>
      <li><strong>Regressed.</strong> MATH Level 5: -5.74. MMLU-Pro: -1.89.</li>
      <li><strong>Defect found.</strong> In the merge that produced v1, inherited by v2. Corrected. The corrected v1 measures at its base model's level on MMLU-Pro (TIGER-Lab).</li>
      <li><strong>Not yet shown.</strong> The end state: a release that improves on its base on every axis. Each iteration is built to recover the regressions and raise the rest.</li>
    </ul>
  </div>

  <div class="section">
    <h2>Why It Matters</h2>
    <ul class="research-list">
      <li><strong>Science.</strong> Whether trade-offs can be controlled, and then removed through iteration, is an open, testable question.</li>
      <li><strong>Economics.</strong> If the hypothesis holds, the marginal cost of a stronger open model is evaluation compute, not pre-training.</li>
      <li><strong>Adoption.</strong> Capability without dependence on one vendor, with a record of every model change.</li>
    </ul>
  </div>

  <div class="section">
    <h2>The Goal</h2>
    <p><strong>Top placements in class, across the public leaderboards.</strong></p>
    <ul class="research-list">
      <li><strong>Scope.</strong> 7B first. Then each larger size class and other model families.</li>
      <li><strong>Method.</strong> Each release controls the trade-offs between skills. Iteration then raises every axis: reasoning, mathematics, code, instruction following and knowledge.</li>
      <li><strong>End state.</strong> Improvement on every axis, with no statistically significant regression against the base model.</li>
      <li><strong>Done.</strong> A class is complete when no public fine-tune improves any axis further.</li>
    </ul>
  </div>

  <div class="section">
    <h2>The Process</h2>
    <ul class="research-list">
      <li><strong>Start</strong> from a strong open base model in the class</li>
      <li><strong>Discover</strong> the compatible fine-tunes the community has published for it</li>
      <li><strong>Assimilate</strong> them into one model</li>
      <li><strong>Measure</strong> every axis against the base model, on identical items and public protocols</li>
      <li><strong>Release</strong> with every gain and every regression published, against the base model and the previous release</li>
      <li><strong>Iterate</strong> to recover the regressions and raise the weaker axes, until nothing more improves. Then the next class</li>
    </ul>
    <p style="margin-top:1rem">The construction method is proprietary. The evaluation is public.</p>
  </div>

  <div class="section">
    <h2>Releases</h2>
    <p>Published under this organisation from converge v3.</p>
    <div class="spaces">
      <div class="space">
        <div>
          <div class="name">converge v3</div>
          <div class="desc">Next iteration. Built to recover v2's regressions and raise the remaining axes</div>
        </div>
      </div>
      <div class="space">
        <div>
          <div class="name">converge v4</div>
          <div class="desc">Planned. The iteration after v3</div>
        </div>
      </div>
      <div class="space">
        <div>
          <div class="name">Larger classes</div>
          <div class="desc">Planned. The same process, at each larger size class and other model families</div>
        </div>
      </div>
    </div>
  </div>

  <div class="section">
    <h2>Research</h2>
    <p>The research behind the releases is on the founder's page, <a href="https://huggingface.co/Optitransfer" style="text-decoration:underline">@Optitransfer</a>.</p>
    <div class="spaces">
      <a href="https://huggingface.co/Optitransfer/Qwen2.5-7B-Instruct-converge-collective-v1" class="space">
        <div>
          <div class="name">converge v1</div>
          <div class="desc">Public. First 7B experiment, on Qwen2.5-7B-Instruct. Carries the defect; see the notice on its card</div>
        </div>
        <span class="arrow">&rarr;</span>
      </a>
      <a href="https://huggingface.co/Optitransfer/Qwen2.5-7B-Instruct-converge-collective-v2" class="space">
        <div>
          <div class="name">converge v2</div>
          <div class="desc">Public. Built on v1, with the 12-benchmark board above. Inherits the defect</div>
        </div>
        <span class="arrow">&rarr;</span>
      </a>
      <div class="space">
        <div>
          <div class="name">converge v1, corrected</div>
          <div class="desc">Built. Full board in progress</div>
        </div>
      </div>
      <div class="space">
        <div>
          <div class="name">converge v2, corrected</div>
          <div class="desc">Built. Full board in progress</div>
        </div>
      </div>
    </div>
    <p style="margin-top:1rem">The corrected models' boards are published when complete, whichever way they fall.</p>
  </div>

  <div class="section">
    <h2>Evidence Standard</h2>
    <ul class="research-list">
      <li><strong>Paired.</strong> Every score is the model minus its base on the same items, with a 95% confidence interval and an exact significance test</li>
      <li><strong>Regressions reported.</strong> A card lists what went down as prominently as what went up</li>
      <li><strong>Calibration labelled.</strong> Calibrated means the harness first reproduced the base model's published score. Every result is marked either way</li>
      <li><strong>Public protocols.</strong> HELM-protocol GSM8K and MMLU, MMLU-Pro (TIGER-Lab), ZeroEval, EvalPlus HumanEval+ and MBPP+, MATH, AIME, ARC, IFEval</li>
      <li><strong>Corrected in public.</strong> When a defect is found, the affected cards say so first</li>
    </ul>
  </div>

  <div class="section">
    <h2>Open Questions</h2>
    <ul class="research-list">
      <li>Can trade-offs be controlled, and every axis then raised through iteration, across a full board?</li>
      <li>Does it beat ensembling, routing and best-of-n sampling at matched compute?</li>
      <li>Does it carry to larger classes and other model families?</li>
      <li>Does it hold on calibration, hallucination and instruction following?</li>
    </ul>
    <p style="margin-top:1rem">Each answer is published as it is measured.</p>
  </div>

  <div class="section">
    <h2>Who It Is For</h2>
    <ul class="research-list">
      <li><strong>Open community.</strong> A stronger open model per class, with the evidence to check it</li>
      <li><strong>Regulated industries.</strong> Healthcare and finance teams whose fine-tunes cannot leave their boundary, and who must show how each model change was made</li>
      <li><strong>National programmes.</strong> Capability that grows by contribution, without dependence on a few vendors</li>
    </ul>
    <p style="margin-top:1rem"><strong>Open model, paid guarantees.</strong> The public model stays open. The planned commercial offering is private consolidation of an organisation's own fine-tunes, inside its boundary, with the same measurement and evidence.</p>
  </div>

  <div class="section">
    <h2>Demonstrations</h2>
    <div class="spaces">
      <a href="https://huggingface.co/spaces/Optitransfer/crdt-merge" class="space">
        <div>
          <div class="name">crdt-merge</div>
          <div class="desc">Merge real models with full provenance</div>
        </div>
        <span class="arrow">&rarr;</span>
      </a>
      <a href="https://huggingface.co/spaces/Optitransfer/convergence-lab" class="space">
        <div>
          <div class="name">Convergence Lab</div>
          <div class="desc">26 strategies, convergence visualisation, experiments</div>
        </div>
        <span class="arrow">&rarr;</span>
      </a>
      <a href="https://huggingface.co/spaces/Optitransfer/crdt-merge-data" class="space">
        <div>
          <div class="name">Data Playground</div>
          <div class="desc">CRDT merging on tabular data</div>
        </div>
        <span class="arrow">&rarr;</span>
      </a>
      <a href="https://huggingface.co/spaces/Optitransfer/crdt-merge-federation" class="space">
        <div>
          <div class="name">Federation</div>
          <div class="desc">Multi-node gossip convergence</div>
        </div>
        <span class="arrow">&rarr;</span>
      </a>
    </div>
  </div>

  <div class="section">
    <h2>Links</h2>
    <div class="links">
      <a href="https://huggingface.co/Optitransfer" class="link-card">
        <div class="label">Models</div>
        <div class="value">huggingface.co/Optitransfer</div>
      </a>
      <a href="https://github.com/mgillr/crdt-merge" class="link-card">
        <div class="label">Source</div>
        <div class="value">github.com/mgillr/crdt-merge</div>
      </a>
      <a href="https://github.com/mgillr/acfa-rs" class="link-card">
        <div class="label">Source</div>
        <div class="value">github.com/mgillr/acfa-rs</div>
      </a>
      <a href="https://arxiv.org/abs/2605.19373" class="link-card">
        <div class="label">Paper</div>
        <div class="value">arXiv:2605.19373</div>
      </a>
      <a href="https://arxiv.org/abs/2607.10305" class="link-card">
        <div class="label">Paper</div>
        <div class="value">arXiv:2607.10305</div>
      </a>
      <a href="https://pypi.org/project/crdt-merge/" class="link-card">
        <div class="label">Package</div>
        <div class="value">pip install crdt-merge</div>
      </a>
    </div>
  </div>

  <footer class="footer">
    Switzerland &middot; <a href="mailto:data@optitransfer.ch">data@optitransfer.ch</a>
  </footer>

</div>
</body>
</html>