intro.html

<div class="amb-head">
  <p class="eyebrow">System one decision model for GUI grounding</p>
  <p class="title">GUI - <span class="fast">Decisions</span> , decoding free GUI grounding</p>
  <p class="lede">Give it a screenshot and an instruction. Instead of <em>writing</em> <code>{"action": "click", "point": [412, 288]}</code>
  token by token, the model fills a fixed template of slots, and every answer is read from one forward pass: the action, and the
  point as a 256-bin distribution over x and over y. Nothing is decoded; the heatmap is that distribution.</p>
  <p class="links">
    <a href="https://huggingface.co/blog/infinitylogesh/gui-decisions-stop-decoding-coordinates">blog post</a> ·
    <a href="https://github.com/infinitylogesh/gui-decisions">code</a> ·
    <a href="https://huggingface.co/infinitylogesh/GUI-Decisions-31B">GUI-Decisions-31B</a> ·
    <a href="https://huggingface.co/infinitylogesh/GUI-Decisions-2B">GUI-Decisions-2B</a>
  </p>
</div>