Upload 3 files
Browse files- index.html +58 -0
- script.js +15 -0
- style.css +127 -0
index.html
CHANGED
|
@@ -30,6 +30,64 @@
|
|
| 30 |
<!-- populated by script.js -->
|
| 31 |
</section>
|
| 32 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 33 |
<div class="controls">
|
| 34 |
<div class="search-pill">
|
| 35 |
<svg width="15" height="15" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round"><circle cx="11" cy="11" r="6"/><line x1="20" y1="20" x2="15.5" y2="15.5"/></svg>
|
|
|
|
| 30 |
<!-- populated by script.js -->
|
| 31 |
</section>
|
| 32 |
|
| 33 |
+
<section class="submit-panel" aria-label="Submission rules">
|
| 34 |
+
<button class="submit-toggle" id="submit-toggle" aria-expanded="false" aria-controls="submit-body">
|
| 35 |
+
<span class="submit-toggle-left">
|
| 36 |
+
<svg width="18" height="18" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round"><path d="M12 5v14M5 12h14"/></svg>
|
| 37 |
+
<span>Submit a model</span>
|
| 38 |
+
</span>
|
| 39 |
+
<span class="submit-toggle-arrow">▼</span>
|
| 40 |
+
</button>
|
| 41 |
+
|
| 42 |
+
<div class="submit-body" id="submit-body" hidden>
|
| 43 |
+
<p class="submit-intro">
|
| 44 |
+
There's no submission form — the leaderboard <em>is</em> <code>models.json</code>.
|
| 45 |
+
Add your entry and open a PR against this Space.
|
| 46 |
+
</p>
|
| 47 |
+
|
| 48 |
+
<h3>Rules</h3>
|
| 49 |
+
<ul class="rules-list">
|
| 50 |
+
<li>Model must be public on the Hugging Face Hub, with a link in <code>url</code>.</li>
|
| 51 |
+
<li>Evaluated on at least one Bench Labs benchmark using the method described in that dataset's README (lm-eval loglikelihood for <code>bench-mid-6-2026</code>, normalized exact-match for <code>bench-effortless-6-2026</code> / <code>bench-easy-6-2026</code>, blind panel rank-order for <code>bench-AGI</code>).</li>
|
| 52 |
+
<li>Include <code>n</code> (sample count) for every benchmark you report a score for.</li>
|
| 53 |
+
<li>If reporting a <code>bench-mid-6-2026</code> result, include the full <code>categories</code> breakdown, not just the headline score — this is what powers the detail panel.</li>
|
| 54 |
+
<li>Link to your eval run or write-up in <code>eval_source</code> (a blog post, a script, a notebook — anything reproducible).</li>
|
| 55 |
+
<li>No self-graded <code>bench-AGI</code> scores: the grader panel must not include the model being evaluated.</li>
|
| 56 |
+
<li>One model per entry. If you have multiple checkpoints, submit them separately with distinct <code>id</code>s.</li>
|
| 57 |
+
</ul>
|
| 58 |
+
|
| 59 |
+
<h3>How to submit</h3>
|
| 60 |
+
<ol class="steps-list">
|
| 61 |
+
<li>Fork this Space's repo.</li>
|
| 62 |
+
<li>Add an object to the <code>models</code> array in <code>models.json</code>, following the existing entries' shape (see <code>schema</code> below).</li>
|
| 63 |
+
<li>Leave <code>score: null</code> for any benchmark tier you haven't run yet — don't omit the key.</li>
|
| 64 |
+
<li>Open a PR back to <code>bench-labs/leaderboard</code> with your eval source linked in the description.</li>
|
| 65 |
+
<li>We verify reproducibility before merging. Once merged, it shows up here automatically — nothing else to deploy.</li>
|
| 66 |
+
</ol>
|
| 67 |
+
|
| 68 |
+
<h3>Entry schema</h3>
|
| 69 |
+
<pre><code id="schema-snippet">{
|
| 70 |
+
"id": "your-model-slug",
|
| 71 |
+
"name": "org/ModelName",
|
| 72 |
+
"org": "org",
|
| 73 |
+
"params_b": 1.5,
|
| 74 |
+
"url": "https://huggingface.co/org/ModelName",
|
| 75 |
+
"eval_source": "https://…",
|
| 76 |
+
"runs": {
|
| 77 |
+
"bench-effortless-6-2026": { "score": 0.0, "n": 240, "notes": "" },
|
| 78 |
+
"bench-easy-6-2026": { "score": 0.0, "n": 300, "notes": "" },
|
| 79 |
+
"bench-mid-6-2026": { "score": null, "n": null, "notes": "" },
|
| 80 |
+
"bench-AGI": { "score": null, "n": null, "notes": "" }
|
| 81 |
+
}
|
| 82 |
+
}</code></pre>
|
| 83 |
+
|
| 84 |
+
<a class="pr-link" href="https://huggingface.co/spaces/bench-labs/leaderboard/discussions" target="_blank" rel="noopener">
|
| 85 |
+
Open a PR on the Space
|
| 86 |
+
<svg width="14" height="14" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2.2" stroke-linecap="round" stroke-linejoin="round"><path d="M7 17 17 7M9 7h8v8"/></svg>
|
| 87 |
+
</a>
|
| 88 |
+
</div>
|
| 89 |
+
</section>
|
| 90 |
+
|
| 91 |
<div class="controls">
|
| 92 |
<div class="search-pill">
|
| 93 |
<svg width="15" height="15" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round"><circle cx="11" cy="11" r="6"/><line x1="20" y1="20" x2="15.5" y2="15.5"/></svg>
|
script.js
CHANGED
|
@@ -25,6 +25,10 @@
|
|
| 25 |
els.detailSubtitle = document.getElementById("detail-subtitle");
|
| 26 |
els.detailCats = document.getElementById("detail-cats");
|
| 27 |
els.detailClose = document.getElementById("detail-close");
|
|
|
|
|
|
|
|
|
|
|
|
|
| 28 |
|
| 29 |
try {
|
| 30 |
const res = await fetch("models.json", { cache: "no-store" });
|
|
@@ -52,6 +56,17 @@
|
|
| 52 |
"</td></tr>";
|
| 53 |
}
|
| 54 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 55 |
// -------------------- TIER STRIP --------------------
|
| 56 |
|
| 57 |
function buildTierStrip() {
|
|
|
|
| 25 |
els.detailSubtitle = document.getElementById("detail-subtitle");
|
| 26 |
els.detailCats = document.getElementById("detail-cats");
|
| 27 |
els.detailClose = document.getElementById("detail-close");
|
| 28 |
+
els.submitToggle = document.getElementById("submit-toggle");
|
| 29 |
+
els.submitBody = document.getElementById("submit-body");
|
| 30 |
+
|
| 31 |
+
bindSubmitPanel();
|
| 32 |
|
| 33 |
try {
|
| 34 |
const res = await fetch("models.json", { cache: "no-store" });
|
|
|
|
| 56 |
"</td></tr>";
|
| 57 |
}
|
| 58 |
|
| 59 |
+
// -------------------- SUBMIT PANEL --------------------
|
| 60 |
+
|
| 61 |
+
function bindSubmitPanel() {
|
| 62 |
+
if (!els.submitToggle) return;
|
| 63 |
+
els.submitToggle.addEventListener("click", () => {
|
| 64 |
+
const open = els.submitToggle.getAttribute("aria-expanded") === "true";
|
| 65 |
+
els.submitToggle.setAttribute("aria-expanded", String(!open));
|
| 66 |
+
els.submitBody.hidden = open;
|
| 67 |
+
});
|
| 68 |
+
}
|
| 69 |
+
|
| 70 |
// -------------------- TIER STRIP --------------------
|
| 71 |
|
| 72 |
function buildTierStrip() {
|
style.css
CHANGED
|
@@ -236,6 +236,133 @@ a:focus-visible { outline: 2px solid var(--accent); outline-offset: 4px; border-
|
|
| 236 |
border: 1px solid var(--border-soft);
|
| 237 |
}
|
| 238 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 239 |
/* -------------------- CONTROLS -------------------- */
|
| 240 |
|
| 241 |
.controls {
|
|
|
|
| 236 |
border: 1px solid var(--border-soft);
|
| 237 |
}
|
| 238 |
|
| 239 |
+
/* -------------------- SUBMIT PANEL -------------------- */
|
| 240 |
+
|
| 241 |
+
.submit-panel {
|
| 242 |
+
margin-bottom: 26px;
|
| 243 |
+
border-radius: var(--radius);
|
| 244 |
+
background: var(--surface);
|
| 245 |
+
border: 1px solid var(--border);
|
| 246 |
+
backdrop-filter: blur(20px) saturate(1.2);
|
| 247 |
+
-webkit-backdrop-filter: blur(20px) saturate(1.2);
|
| 248 |
+
box-shadow: var(--shadow-soft);
|
| 249 |
+
overflow: hidden;
|
| 250 |
+
}
|
| 251 |
+
|
| 252 |
+
.submit-toggle {
|
| 253 |
+
display: flex;
|
| 254 |
+
align-items: center;
|
| 255 |
+
justify-content: space-between;
|
| 256 |
+
width: 100%;
|
| 257 |
+
padding: 16px 22px;
|
| 258 |
+
background: transparent;
|
| 259 |
+
border: none;
|
| 260 |
+
cursor: pointer;
|
| 261 |
+
font-family: inherit;
|
| 262 |
+
font-size: 0.95rem;
|
| 263 |
+
font-weight: 600;
|
| 264 |
+
color: var(--heading);
|
| 265 |
+
transition: background var(--transition);
|
| 266 |
+
}
|
| 267 |
+
|
| 268 |
+
.submit-toggle:hover { background: rgba(255,255,255,0.35); }
|
| 269 |
+
.submit-toggle:focus-visible { outline: 2px solid var(--accent); outline-offset: -2px; }
|
| 270 |
+
|
| 271 |
+
.submit-toggle-left {
|
| 272 |
+
display: flex;
|
| 273 |
+
align-items: center;
|
| 274 |
+
gap: 10px;
|
| 275 |
+
color: var(--accent);
|
| 276 |
+
}
|
| 277 |
+
|
| 278 |
+
.submit-toggle-left span { color: var(--heading); }
|
| 279 |
+
|
| 280 |
+
.submit-toggle-arrow {
|
| 281 |
+
font-size: 0.7rem;
|
| 282 |
+
color: var(--accent);
|
| 283 |
+
transition: transform var(--transition);
|
| 284 |
+
}
|
| 285 |
+
|
| 286 |
+
.submit-toggle[aria-expanded="true"] .submit-toggle-arrow { transform: rotate(180deg); }
|
| 287 |
+
|
| 288 |
+
.submit-body {
|
| 289 |
+
padding: 4px 26px 26px;
|
| 290 |
+
border-top: 1px solid var(--border-soft);
|
| 291 |
+
}
|
| 292 |
+
|
| 293 |
+
.submit-intro {
|
| 294 |
+
margin: 18px 0 20px;
|
| 295 |
+
color: var(--text);
|
| 296 |
+
font-size: 0.92rem;
|
| 297 |
+
line-height: 1.6;
|
| 298 |
+
}
|
| 299 |
+
|
| 300 |
+
.submit-body h3 {
|
| 301 |
+
font-size: 0.95rem;
|
| 302 |
+
margin: 22px 0 10px;
|
| 303 |
+
}
|
| 304 |
+
|
| 305 |
+
.submit-body h3:first-of-type { margin-top: 4px; }
|
| 306 |
+
|
| 307 |
+
.rules-list, .steps-list {
|
| 308 |
+
padding-left: 22px;
|
| 309 |
+
color: var(--text);
|
| 310 |
+
font-size: 0.88rem;
|
| 311 |
+
line-height: 1.7;
|
| 312 |
+
}
|
| 313 |
+
|
| 314 |
+
.rules-list li, .steps-list li { margin-bottom: 6px; }
|
| 315 |
+
|
| 316 |
+
.submit-body code {
|
| 317 |
+
background: rgba(255,255,255,0.7);
|
| 318 |
+
padding: 1px 6px;
|
| 319 |
+
border-radius: 6px;
|
| 320 |
+
border: 1px solid var(--border-soft);
|
| 321 |
+
font-size: 0.85em;
|
| 322 |
+
color: var(--heading);
|
| 323 |
+
}
|
| 324 |
+
|
| 325 |
+
.submit-body pre {
|
| 326 |
+
background: rgba(255,255,255,0.75);
|
| 327 |
+
border: 1px solid var(--border-soft);
|
| 328 |
+
border-radius: 14px;
|
| 329 |
+
padding: 18px 20px;
|
| 330 |
+
overflow-x: auto;
|
| 331 |
+
margin: 10px 0 18px;
|
| 332 |
+
box-shadow: var(--shadow-soft);
|
| 333 |
+
}
|
| 334 |
+
|
| 335 |
+
.submit-body pre code {
|
| 336 |
+
background: none;
|
| 337 |
+
border: none;
|
| 338 |
+
padding: 0;
|
| 339 |
+
font-family: ui-monospace, SFMono-Regular, Menlo, Monaco, Consolas, monospace;
|
| 340 |
+
font-size: 0.82rem;
|
| 341 |
+
line-height: 1.7;
|
| 342 |
+
color: var(--text);
|
| 343 |
+
}
|
| 344 |
+
|
| 345 |
+
.pr-link {
|
| 346 |
+
display: inline-flex;
|
| 347 |
+
align-items: center;
|
| 348 |
+
gap: 8px;
|
| 349 |
+
margin-top: 4px;
|
| 350 |
+
padding: 10px 18px;
|
| 351 |
+
border-radius: 999px;
|
| 352 |
+
background: var(--accent);
|
| 353 |
+
color: #fff;
|
| 354 |
+
font-size: 0.85rem;
|
| 355 |
+
font-weight: 600;
|
| 356 |
+
box-shadow: var(--shadow-soft);
|
| 357 |
+
transition: transform var(--transition), box-shadow var(--transition), opacity var(--transition);
|
| 358 |
+
}
|
| 359 |
+
|
| 360 |
+
.pr-link:hover {
|
| 361 |
+
opacity: 1;
|
| 362 |
+
transform: translateY(-2px);
|
| 363 |
+
box-shadow: var(--shadow-hover);
|
| 364 |
+
}
|
| 365 |
+
|
| 366 |
/* -------------------- CONTROLS -------------------- */
|
| 367 |
|
| 368 |
.controls {
|