Open source and provenance links in a new tab (Hub refuses framing)

harvard-botany-headers/index.html

@@ -55,7 +55,7 @@ details{margin-top:24px}img{max-width:100%;max-height:440px}pre{white-space:pre-
 <h1>GLAM extraction benchmark</h1><p>Harvard Botany Libraries: printed taxon headings and handwritten replacements</p>
 <label for="dataset">Dataset </label><select id="dataset"><option data-config="overall" value="../overall.html">Overall</option><option data-config="nls-index-cards" value="../index.html">nls-index-cards</option><option data-config="harvard-botany-headers" value="../harvard-botany-headers/index.html" selected>harvard-botany-headers</option></select>
 <p class="sub">30 documents · 10 models · Harvard University Botany Libraries</p>
-<p>Images and checked labels: <a href="https://huggingface.co/datasets/biglam/index-cards-harvard-botany-metropolitan-flora/tree/986938d535c7710d33f73219a918ecddf16b3483">biglam/index-cards-harvard-botany-metropolitan-flora</a>
+<p>Images and checked labels: <a target="_blank" rel="noopener noreferrer" href="https://huggingface.co/datasets/biglam/index-cards-harvard-botany-metropolitan-flora/tree/986938d535c7710d33f73219a918ecddf16b3483">biglam/index-cards-harvard-botany-metropolitan-flora</a>
 (public-domain). Visual assistant transcription, not source OCR. Printed fields: Human reviewed. Corrections: Replacement strings human reviewed; absence visually audited.</p>
 <p class="notice">Predictions were generated for this benchmark config and revision.</p>
 <p><b>Risk</b> is a wrong identifier or an invented field. <b>Workload</b> is a blank field that
@@ -91,11 +91,11 @@ Small score differences may be noise.</p>
 <details><summary>Inspect an example document and its checked output</summary>
 <img src="example.jpg" alt="Example document from the selected evaluation dataset"><pre id="gold"></pre></details>
 <details><summary>Reproducibility and downloads</summary><p>Config: <code>harvard-botany-headers</code>;
-split: <code>test</code>; <a href="https://huggingface.co/datasets/small-models-for-glam/glam-extraction-benchmark/tree/48e92ea1112d3fc2a32467e875d64e10129c4d46">dataset revision 48e92ea1112d</a>.
+split: <code>test</code>; <a target="_blank" rel="noopener noreferrer" href="https://huggingface.co/datasets/small-models-for-glam/glam-extraction-benchmark/tree/48e92ea1112d3fc2a32467e875d64e10129c4d46">dataset revision 48e92ea1112d</a>.
 Harness 0.0.1; scorer 2026-08-26a. Raw prediction files retain their original inference provenance.</p>
-<p><a href="scores.json">Scores and run provenance</a> · <a href="manifest.json">Dataset manifest</a> ·
-<a href="https://github.com/davanstrien/glam-extraction-benchmark">Code</a> ·
-<a href="https://huggingface.co/buckets/small-models-for-glam/glam-extraction-results">Raw predictions</a></p></details>
+<p><a target="_blank" rel="noopener noreferrer" href="scores.json">Scores and run provenance</a> · <a target="_blank" rel="noopener noreferrer" href="manifest.json">Dataset manifest</a> ·
+<a target="_blank" rel="noopener noreferrer" href="https://github.com/davanstrien/glam-extraction-benchmark">Code</a> ·
+<a target="_blank" rel="noopener noreferrer" href="https://huggingface.co/buckets/small-models-for-glam/glam-extraction-results">Raw predictions</a></p></details>
 </main><script>function drawSizePlot(rows) {
   const svg = document.getElementById('size-plot');
   const escapeText = value => String(value).replace(/[&<>"']/g, c => ({'&':'&amp;','<':'&lt;','>':'&gt;','"':'&quot;',"'":'&#39;'}[c]));

index.html

@@ -55,7 +55,7 @@ details{margin-top:24px}img{max-width:100%;max-height:440px}pre{white-space:pre-
 <h1>GLAM extraction benchmark</h1><p>National Library of Scotland manuscript index cards</p>
 <label for="dataset">Dataset </label><select id="dataset"><option data-config="overall" value="overall.html">Overall</option><option data-config="nls-index-cards" value="index.html" selected>nls-index-cards</option><option data-config="harvard-botany-headers" value="harvard-botany-headers/index.html">harvard-botany-headers</option></select>
 <p class="sub">98 documents · 10 models · National Library of Scotland</p>
-<p>Images and checked labels: <a href="https://huggingface.co/datasets/NationalLibraryOfScotland/index-cards-eval/tree/2a81070549d8493c2c538744a9dbbc1dc72cb146">NationalLibraryOfScotland/index-cards-eval</a>
+<p>Images and checked labels: <a target="_blank" rel="noopener noreferrer" href="https://huggingface.co/datasets/NationalLibraryOfScotland/index-cards-eval/tree/2a81070549d8493c2c538744a9dbbc1dc72cb146">NationalLibraryOfScotland/index-cards-eval</a>
 (cc0-1.0). Labels drafted by Qwen3.6-35B-A3B-GGUF-Q8_0 and reviewed by NLS cataloguers.</p>
 <p class="notice">Predictions were generated for this benchmark config and revision.</p>
 <p><b>Risk</b> is a wrong identifier or an invented field. <b>Workload</b> is a blank field that
@@ -91,11 +91,11 @@ Small score differences may be noise.</p>
 <details><summary>Inspect an example document and its checked output</summary>
 <img src="example.jpg" alt="Example document from the selected evaluation dataset"><pre id="gold"></pre></details>
 <details><summary>Reproducibility and downloads</summary><p>Config: <code>nls-index-cards</code>;
-split: <code>test</code>; <a href="https://huggingface.co/datasets/small-models-for-glam/glam-extraction-benchmark/tree/ecc9c02582f933ca89b73e6751e4ed76888cb24d">dataset revision ecc9c02582f9</a>.
+split: <code>test</code>; <a target="_blank" rel="noopener noreferrer" href="https://huggingface.co/datasets/small-models-for-glam/glam-extraction-benchmark/tree/ecc9c02582f933ca89b73e6751e4ed76888cb24d">dataset revision ecc9c02582f9</a>.
 Harness 0.0.1; scorer 2026-08-26a. Raw prediction files retain their original inference provenance.</p>
-<p><a href="scores.json">Scores and run provenance</a> · <a href="manifest.json">Dataset manifest</a> ·
-<a href="https://github.com/davanstrien/glam-extraction-benchmark">Code</a> ·
-<a href="https://huggingface.co/buckets/small-models-for-glam/glam-extraction-results">Raw predictions</a></p></details>
+<p><a target="_blank" rel="noopener noreferrer" href="scores.json">Scores and run provenance</a> · <a target="_blank" rel="noopener noreferrer" href="manifest.json">Dataset manifest</a> ·
+<a target="_blank" rel="noopener noreferrer" href="https://github.com/davanstrien/glam-extraction-benchmark">Code</a> ·
+<a target="_blank" rel="noopener noreferrer" href="https://huggingface.co/buckets/small-models-for-glam/glam-extraction-results">Raw predictions</a></p></details>
 </main><script>function drawSizePlot(rows) {
   const svg = document.getElementById('size-plot');
   const escapeText = value => String(value).replace(/[&<>"']/g, c => ({'&':'&amp;','<':'&lt;','>':'&gt;','"':'&quot;',"'":'&#39;'}[c]));

overall.html

@@ -79,16 +79,16 @@ Click a column heading to sort; click again to reverse.</p>
 <tr data-params="2.274"><td data-value="Qwen3.5-2B"><a href="https://huggingface.co/Qwen/Qwen3.5-2B" target="_blank" rel="noopener noreferrer" title="Qwen/Qwen3.5-2B">Qwen3.5-2B</a> <small>2.274B</small></td><td class="r" data-value="0.6328972448979592">63.3</td></tr></tbody></table></div>
 <p id="empty-results" hidden>No models match this size limit.</p>
 <p class="foot">Scores are shown on a 0–100 scale. Dataset weights, component scores and pinned
-revisions are recorded in <a href="overall-scores.json">the score download</a>.</p>
+revisions are recorded in <a target="_blank" rel="noopener noreferrer" href="overall-scores.json">the score download</a>.</p>
 <section class="plot-section" aria-labelledby="plot-title"><h2 id="plot-title">Size vs. overall extraction quality</h2>
 <p>Upper-left is better: smaller models, higher overall F1. Teal points show the observed Pareto frontier.</p>
 <svg id="size-plot" data-f1-column="1" viewBox="0 0 760 360" role="group" aria-label="Model size versus overall extraction F1"></svg>
 <p id="plot-note"></p></section>
 <details><summary>Sources and reproducibility</summary>
-<p><a href="overall-scores.json">Overall scores and component provenance</a> ·
-<a href="configs.json">Dataset configurations and revisions</a> ·
-<a href="https://github.com/davanstrien/glam-extraction-benchmark">Code</a> ·
-<a href="https://huggingface.co/buckets/small-models-for-glam/glam-extraction-results">Raw predictions</a></p>
+<p><a target="_blank" rel="noopener noreferrer" href="overall-scores.json">Overall scores and component provenance</a> ·
+<a target="_blank" rel="noopener noreferrer" href="configs.json">Dataset configurations and revisions</a> ·
+<a target="_blank" rel="noopener noreferrer" href="https://github.com/davanstrien/glam-extraction-benchmark">Code</a> ·
+<a target="_blank" rel="noopener noreferrer" href="https://huggingface.co/buckets/small-models-for-glam/glam-extraction-results">Raw predictions</a></p>
 <p>The overall view combines the selected dataset results; it does not create a new dataset or change individual scores.</p>
 </details></main><script>function drawSizePlot(rows) {
   const svg = document.getElementById('size-plot');