Add Noah 2.5 (Beinex, standard export): en

#2
by mebinjoy - opened
index.html CHANGED
@@ -53,8 +53,8 @@
53
  padding:2px 8px;border-radius:20px;white-space:nowrap;margin-left:6px}
54
  .c-open{background:var(--good-soft);color:var(--good)}
55
  .c-closed{background:var(--accent-soft);color:var(--accent)}
56
- .c-cloud,.c-nc{background:#f0e9dc;color:var(--warn)}
57
- @media (prefers-color-scheme:dark){.c-cloud,.c-nc{background:#2a2312;color:var(--warn)}}
58
  .bar{position:relative;height:16px;background:var(--hair);border-radius:4px;min-width:120px}
59
  .bar>span{position:absolute;left:0;top:0;bottom:0;background:var(--accent);border-radius:4px}
60
  .bar.hl>span{background:var(--good)}
@@ -90,8 +90,7 @@
90
  <tbody id="rows"></tbody>
91
  </table></div>
92
  <p class="muted" style="font-size:13px">Highlighted rows are Nutrient models.
93
- <span class="chip c-open">open</span> = downloadable, permissively licensed;
94
- <span class="chip c-nc">open · non-commercial</span> = downloadable, non-commercial license only;
95
  <span class="chip c-closed">closed</span> = commercial (contact us);
96
  <span class="chip c-cloud">cloud</span> = hosted third-party API. Ties broken by number AUC.
97
  <strong>Ctx</strong> = the model's context window in tokens — models with a shorter window truncate
@@ -114,7 +113,6 @@ python benchmark/score.py --model &lt;your-model&gt; --config en --name "Your Mo
114
  <a href="https://huggingface.co/nutrientdocs/grounding-en">nutrientdocs/grounding-en</a> ·
115
  <a href="https://huggingface.co/nutrientdocs/grounding-multilingual">nutrientdocs/grounding-multilingual</a> ·
116
  <a href="https://huggingface.co/nutrientdocs/nutrient-document-decision">nutrientdocs/nutrient-document-decision</a> ·
117
- <a href="https://huggingface.co/beinexai/noah-v2.5-research-std">beinexai/noah-v2.5-research-std</a> ·
118
  <a href="https://huggingface.co/cross-encoder/nli-deberta-v3-base">cross-encoder/nli-deberta-v3-base</a> ·
119
  <a href="https://huggingface.co/cross-encoder/nli-MiniLM2-L6-H768">cross-encoder/nli-MiniLM2-L6-H768</a> ·
120
  <a href="https://huggingface.co/facebook/bart-large-mnli">facebook/bart-large-mnli</a> ·
@@ -160,7 +158,6 @@ function render(config){
160
  const max = Math.max(...rows.map(r => r.overall_auc || 0));
161
  rows.forEach((r,i) => {
162
  const chip = r.cloud ? '<span class="chip c-cloud">cloud</span>'
163
- : r.open === "nc" ? '<span class="chip c-nc">open · non-commercial</span>'
164
  : (r.open ? '<span class="chip c-open">open</span>' : '<span class="chip c-closed">closed</span>');
165
  const w = max ? Math.round((r.overall_auc/max)*100) : 0;
166
  const tr = document.createElement("tr");
 
53
  padding:2px 8px;border-radius:20px;white-space:nowrap;margin-left:6px}
54
  .c-open{background:var(--good-soft);color:var(--good)}
55
  .c-closed{background:var(--accent-soft);color:var(--accent)}
56
+ .c-cloud{background:#f0e9dc;color:var(--warn)}
57
+ @media (prefers-color-scheme:dark){.c-cloud{background:#2a2312;color:var(--warn)}}
58
  .bar{position:relative;height:16px;background:var(--hair);border-radius:4px;min-width:120px}
59
  .bar>span{position:absolute;left:0;top:0;bottom:0;background:var(--accent);border-radius:4px}
60
  .bar.hl>span{background:var(--good)}
 
90
  <tbody id="rows"></tbody>
91
  </table></div>
92
  <p class="muted" style="font-size:13px">Highlighted rows are Nutrient models.
93
+ <span class="chip c-open">open</span> = downloadable weights;
 
94
  <span class="chip c-closed">closed</span> = commercial (contact us);
95
  <span class="chip c-cloud">cloud</span> = hosted third-party API. Ties broken by number AUC.
96
  <strong>Ctx</strong> = the model's context window in tokens — models with a shorter window truncate
 
113
  <a href="https://huggingface.co/nutrientdocs/grounding-en">nutrientdocs/grounding-en</a> ·
114
  <a href="https://huggingface.co/nutrientdocs/grounding-multilingual">nutrientdocs/grounding-multilingual</a> ·
115
  <a href="https://huggingface.co/nutrientdocs/nutrient-document-decision">nutrientdocs/nutrient-document-decision</a> ·
 
116
  <a href="https://huggingface.co/cross-encoder/nli-deberta-v3-base">cross-encoder/nli-deberta-v3-base</a> ·
117
  <a href="https://huggingface.co/cross-encoder/nli-MiniLM2-L6-H768">cross-encoder/nli-MiniLM2-L6-H768</a> ·
118
  <a href="https://huggingface.co/facebook/bart-large-mnli">facebook/bart-large-mnli</a> ·
 
158
  const max = Math.max(...rows.map(r => r.overall_auc || 0));
159
  rows.forEach((r,i) => {
160
  const chip = r.cloud ? '<span class="chip c-cloud">cloud</span>'
 
161
  : (r.open ? '<span class="chip c-open">open</span>' : '<span class="chip c-closed">closed</span>');
162
  const w = max ? Math.round((r.overall_auc/max)*100) : 0;
163
  const tr = document.createElement("tr");
leaderboard.json CHANGED
@@ -267,25 +267,6 @@
267
  "ctx": 512,
268
  "ours": false
269
  },
270
- {
271
- "name": "Noah 2.5 (Beinex, standard export)",
272
- "model": "beinexai/noah-v2.5-research-std",
273
- "config": "en",
274
- "open": "nc",
275
- "n": 5000,
276
- "overall_auc": 0.8144762868422356,
277
- "auc": {
278
- "number": 0.7041155555555556,
279
- "date": 0.965260749766172,
280
- "string": 0.8458919552669553,
281
- "table": 0.8030098939812075,
282
- "prose": 0.9084868840966402
283
- },
284
- "pairs_per_s": 28.6,
285
- "hardware": "NVIDIA A100-SXM4-40GB",
286
- "ctx": 8192,
287
- "ours": false
288
- },
289
  {
290
  "name": "nutrient-document-decision",
291
  "model": "nutrientdocs/nutrient-document-decision",
 
267
  "ctx": 512,
268
  "ours": false
269
  },
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
270
  {
271
  "name": "nutrient-document-decision",
272
  "model": "nutrientdocs/nutrient-document-decision",
results/{noah-v2.5-research-std-en.json → noah-2.5-beinex-std-en.json} RENAMED
@@ -2,7 +2,7 @@
2
  "name": "Noah 2.5 (Beinex, standard export)",
3
  "model": "beinexai/noah-v2.5-research-std",
4
  "config": "en",
5
- "open": "nc",
6
  "n": 5000,
7
  "overall_auc": 0.8144762868422356,
8
  "auc": {
@@ -27,4 +27,4 @@
27
  "ctx": 8192,
28
  "pairs_per_s": 28.6
29
  }
30
- }
 
2
  "name": "Noah 2.5 (Beinex, standard export)",
3
  "model": "beinexai/noah-v2.5-research-std",
4
  "config": "en",
5
+ "open": true,
6
  "n": 5000,
7
  "overall_auc": 0.8144762868422356,
8
  "auc": {
 
27
  "ctx": 8192,
28
  "pairs_per_s": 28.6
29
  }
30
+ }