Rewrite the explorer text

app.js

@@ -31,4 +31,4 @@ Promise.all(['benchmark.json','catalogue.json','summary.json'].map(async url=>{c
  records=r;catalogue=c;
  for(const id of ['model-a','model-b'])for(const [name,metric]of Object.entries(s.metrics)){const o=document.createElement('option');o.value=name;o.textContent=`${name} · ${metric.protocol}`;$(id).append(o);}
  $('model-a').value='RoboJev-27B';$('model-b').value='GPT-6 Astra';render();
-}).catch(()=>{$('count').textContent='The data could not be loaded.';$('table').innerHTML='<div class="empty error">Please reload the page, or use the download links above.</div>';});
+}).catch(()=>{$('count').textContent='Unable to load the data.';$('table').innerHTML='<div class="empty error">Please reload the page, or use the download links above.</div>';});

index.html

@@ -1 +1 @@
-<!doctype html><html lang="en"><head><meta charset="utf-8"><meta name="viewport" content="width=device-width,initial-scale=1"><meta name="description" content="Explore RoboJev's recorded robot success benchmarks and sampled LeRobot catalogue."><title>RoboJev · Explore the data</title><link rel="stylesheet" href="style.css"></head><body><div class="wrap"><header class="nav"><a class="brand" href="https://huggingface.co/spaces/josefchen/RoboJev">RoboJev<span> ◉</span></a><nav class="navlinks" aria-label="Resources"><a href="https://huggingface.co/josefchen/RoboJev-27B">Model</a><a href="https://huggingface.co/spaces/josefchen/RoboJev-demo">Try it</a><a href="https://huggingface.co/spaces/josefchen/RoboJev">Writeup</a></nav></header><main><section class="hero"><p class="eyebrow">Open evaluation data · 23 October 2026</p><h1>See every answer.</h1><p class="intro">Compare recorded predictions with the human labels, inspect disagreements, or find a dataset in the LeRobot catalogue.</p><div class="actions"><a class="pill" href="https://huggingface.co/datasets/josefchen/RoboJev-evals">Download the data <span class="arrow" aria-hidden="true">↗</span></a><a class="pill secondary" href="benchmark.csv" download>Benchmark CSV</a></div></section><div class="bezel"><div class="glass stats"><div class="stat"><strong>927</strong><span>labelled evaluation episodes</span></div><div class="stat"><strong>29</strong><span>model / protocol entries</span></div><div class="stat"><strong>7,641</strong><span>datasets in the catalogue</span></div></div><section class="section"><div class="tabs" aria-label="Choose a collection"><button id="benchmark-tab" type="button" aria-pressed="true">Benchmark</button><button id="catalogue-tab" type="button" aria-pressed="false">Catalogue</button></div><div id="benchmark-intro"><h2>Compare the models.</h2><p class="muted">862 episodes have a clear label. The 65 ambiguous community clips are included for inspection and excluded from scored metrics.</p><details><summary>Inputs and comparison methods</summary><p>RoboJev's released model reads ten evenly spaced front-camera frames plus final frames from other cameras when available. The frontier LLMs answered zero-shot from four front-camera frames. <code>clip-v1</code> means four frames, <code>dense-v1</code> means ten plus other views, and <code>video</code> means whole-video input.</p><p>SO-101: 268 held-out runs; RoboJev trained on other ArmnetBench sessions. Unseen labs: 459 runs from eight labs outside training. Community: 200 clips labelled by one person, 135 with a clear outcome. Missing results are shown as a dash. Earlier model variants remain available in the selector.</p><p><a href="summary.json">Metrics, paired bootstrap intervals and protocol details (JSON)</a></p></details></div><div id="catalogue-intro" hidden><h2>A sample of robot evaluations.</h2><p class="muted">68,555 sampled episodes, totalling 501 hours. These are catalogue records, not a complete census of Hugging Face. Predictions from the earlier four-frame model are omitted; the current release has not regraded this catalogue.</p><a href="catalogue.jsonl" download>Download catalogue JSONL</a></div><div class="filters"><label class="field grow">Search task, dataset or robot<input id="search" type="search" placeholder="Try ring, folding or SO-101" autocomplete="off"></label><label class="field bench-only">Evaluation set<select id="suite"><option value="all">All evaluations</option><option value="so101">SO-101 · 268</option><option value="unseen">Unseen labs · 459</option><option value="community">Community · 200</option></select></label><label class="field bench-only">Show<select id="filter"><option value="all">All outcomes</option><option value="disagree">Models disagree</option><option value="wrong">RoboJev misses</option><option value="unclear">Unclear human label</option></select></label></div><div class="filters bench-only"><label class="field">First model<select id="model-a" aria-label="First model"></select></label><label class="field">Compare with<select id="model-b" aria-label="Compare with"></select></label></div><p id="count" role="status" aria-live="polite" class="muted">Loading recorded results…</p><div id="table" class="table-scroll" tabindex="0" aria-label="Scrollable results"><div class="loading">Loading data…</div></div><div class="pagination"><button id="previous" type="button" disabled>← Previous</button><span id="page" class="muted"></span><button id="next" type="button" disabled>Next →</button></div></section></main><footer class="footer">RoboJev-27B · Released model and measured results. Source dataset links appear on every row. <a href="https://huggingface.co/josefchen/RoboJev-27B#video-credits">Credits</a></footer></div><script src="app.js"></script></body></html>
+<!doctype html><html lang="en"><head><meta charset="utf-8"><meta name="viewport" content="width=device-width,initial-scale=1"><meta name="description" content="Explore RoboJev's recorded robot success benchmarks and sampled LeRobot catalogue."><title>RoboJev · Explore the data</title><link rel="stylesheet" href="style.css"></head><body><div class="wrap"><header class="nav"><a class="brand" href="https://huggingface.co/spaces/josefchen/RoboJev">RoboJev<span> ◉</span></a><nav class="navlinks" aria-label="Resources"><a href="https://huggingface.co/josefchen/RoboJev-27B">Model</a><a href="https://huggingface.co/spaces/josefchen/RoboJev-demo">Try it</a><a href="https://huggingface.co/spaces/josefchen/RoboJev">Writeup</a></nav></header><main><section class="hero"><p class="eyebrow">Open evaluation data · 23 October 2026</p><h1>See all answers.</h1><p class="intro">You can compare recorded predictions against human labels, investigate disagreements, or browse datasets in the LeRobot catalogue.</p><div class="actions"><a class="pill" href="https://huggingface.co/datasets/josefchen/RoboJev-evals">Download the data <span class="arrow" aria-hidden="true">↗</span></a><a class="pill secondary" href="benchmark.csv" download>Benchmark CSV</a></div></section><div class="bezel"><div class="glass stats"><div class="stat"><strong>927</strong><span>labelled evaluation episodes</span></div><div class="stat"><strong>29</strong><span>model / protocol entries</span></div><div class="stat"><strong>7,641</strong><span>datasets in the catalogue</span></div></div><section class="section"><div class="tabs" aria-label="Choose a collection"><button id="benchmark-tab" type="button" aria-pressed="true">Benchmark</button><button id="catalogue-tab" type="button" aria-pressed="false">Catalogue</button></div><div id="benchmark-intro"><h2>Comparison of the models</h2><p class="muted">Out of 927 labelled episodes, 862 have a definitive label. The remaining 65 ambiguous community clips are presented here for your examination but are not included in the metrics scoring.</p><details><summary>Inputs and comparison methods</summary><p>The released RoboJev reads 10 evenly-spaced front-camera frames, as well as end-frames from other cameras if available. Frontier LLMs answered in a zero-shot manner using just four front-camera frames. <code>clip-v1</code> means 4 frames, <code>dense-v1</code> means 10 frames + other views, <code>video</code> means whole-video input.</p><p>SO-101 has 268 held-out runs, and RoboJev was trained on other ArmnetBench sessions. The unseen labs contain 459 runs from eight labs that we didn't use for training. The community contains 200 clips that are only labelled by one person, and 135 of them have a clear outcome. We show missing results with a dash and all previous model variants can be selected.</p><p><a href="summary.json">Metrics, paired bootstrap intervals and protocol details (JSON)</a></p></details></div><div id="catalogue-intro" hidden><h2>A sample of robots being evaluated on the Hub.</h2><p class="muted">68,555 episodes sampled, 501 hours total. These are not a full census of Hugging Face; they are just catalogue entries. The predictions from the previous four-frame model are not included, and we have not regraded this catalogue.</p><a href="catalogue.jsonl" download>Download catalogue JSONL</a></div><div class="filters"><label class="field grow">Search task, dataset or robot<input id="search" type="search" placeholder="Try ring, folding or SO-101" autocomplete="off"></label><label class="field bench-only">Evaluation set<select id="suite"><option value="all">All evaluations</option><option value="so101">SO-101 · 268</option><option value="unseen">Unseen labs · 459</option><option value="community">Community · 200</option></select></label><label class="field bench-only">Show<select id="filter"><option value="all">All outcomes</option><option value="disagree">Models disagree</option><option value="wrong">RoboJev misses</option><option value="unclear">Unclear human label</option></select></label></div><div class="filters bench-only"><label class="field">First model<select id="model-a" aria-label="First model"></select></label><label class="field">Compare with<select id="model-b" aria-label="Compare with"></select></label></div><p id="count" role="status" aria-live="polite" class="muted">Loading recorded results…</p><div id="table" class="table-scroll" tabindex="0" aria-label="Scrollable results"><div class="loading">Loading data…</div></div><div class="pagination"><button id="previous" type="button" disabled>← Previous</button><span id="page" class="muted"></span><button id="next" type="button" disabled>Next →</button></div></section></main><footer class="footer">RoboJev-27B · Here is the released model, and here are its measured results. Every row contains a link to the source dataset. <a href="https://huggingface.co/josefchen/RoboJev-27B#video-credits">Credits</a></footer></div><script src="app.js"></script></body></html>