https://datafranca.org/wiki/index.php?action=history&feed=atom&title=Meta_CLIP 2026-04-09T01:29:23Z Historique des versions pour cette page sur le wiki MediaWiki 1.39.5 https://datafranca.org/wiki/index.php?title=Meta_CLIP&diff=116731&oldid=prev 2025-08-25T12:25:05Z

<p></p> <table style="background-color: #fff; color: #202122;" data-mw="interface"> <col class="diff-marker" /> <col class="diff-content" /> <col class="diff-marker" /> <col class="diff-content" /> <tr class="diff-title" lang="fr"> <td colspan="2" style="background-color: #fff; color: #202122; text-align: center;">← Version précédente</td> <td colspan="2" style="background-color: #fff; color: #202122; text-align: center;">Version du 25 août 2025 à 08:25</td> </tr><tr><td colspan="2" class="diff-lineno" id="mw-diff-left-l11">Ligne 11 :</td> <td colspan="2" class="diff-lineno">Ligne 11 :</td></tr> <tr><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><br/></td><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><br/></td></tr> <tr><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>  The first recipe for training CLIP models from scratch on worldwide web-scale image-text pairs spanning 300+ languages. The work addresses the challenge of scaling CLIP beyond English-only data while avoiding the &quot;curse of multilinguality&quot; - where multilingual models perform worse on English tasks than their English-only counterparts. </div></td><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>  The first recipe for training CLIP models from scratch on worldwide web-scale image-text pairs spanning 300+ languages. The work addresses the challenge of scaling CLIP beyond English-only data while avoiding the &quot;curse of multilinguality&quot; - where multilingual models perform worse on English tasks than their English-only counterparts. </div></td></tr> <tr><td class="diff-marker" data-marker="−"></td><td style="color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #ffe49c; vertical-align: top; white-space: pre-wrap;"><div>Meta CLIP 2 presents a recipe for training CLIP models on worldwide multilingual data from scratch. The work demonstrates that the curse of multilinguality can be overcome through careful scaling of metadata construction, data curation algorithms, and training frameworks.</div></td><td class="diff-marker" data-marker="+"></td><td style="color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #a3d3ff; vertical-align: top; white-space: pre-wrap;"><div><ins style="font-weight: bold; text-decoration: none;"> </ins>Meta CLIP 2 presents a recipe for training CLIP models on worldwide multilingual data from scratch. The work demonstrates that the curse of multilinguality can be overcome through careful scaling of metadata construction, data curation algorithms, and training frameworks.</div></td></tr> <tr><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>   </div></td><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>   </div></td></tr> <tr><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>  MetaCLIP 2, trained on worldwide web-scale image-text pairs, improves zero-shot classification and multilingual benchmarks without system-level confounding factors.</div></td><td class="diff-marker"></td><td style="background-color: #f8f9fa; color: #202122; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #eaecf0; vertical-align: top; white-space: pre-wrap;"><div>  MetaCLIP 2, trained on worldwide web-scale image-text pairs, improves zero-shot classification and multilingual benchmarks without system-level confounding factors.</div></td></tr> </table>

Pitpitt https://datafranca.org/wiki/index.php?title=Meta_CLIP&diff=116730&oldid=prev 2025-08-25T12:24:28Z

<p>Page créée avec « ==en construction== == Définition == XXXXXXXXX == Français == &#039;&#039;&#039;  Meta CLIP 2&#039;&#039;&#039; == Anglais == &#039;&#039;&#039; Meta CLIP 2&#039;&#039;&#039; The first recipe for training CLIP models from scratch on worldwide web-scale image-text pairs spanning 300+ languages. The work addresses the challenge of scaling CLIP beyond English-only data while avoiding the &quot;curse of multilinguality&quot; - where multilingual models perform worse on English tasks than their English-only counterparts.  Meta... »</p> <p><b>Nouvelle page</b></p><div>==en construction==<br /> <br /> == Définition ==<br /> XXXXXXXXX<br /> <br /> == Français ==<br /> &#039;&#039;&#039;  Meta CLIP 2&#039;&#039;&#039;<br /> <br /> == Anglais ==<br /> &#039;&#039;&#039; Meta CLIP 2&#039;&#039;&#039;<br /> <br /> The first recipe for training CLIP models from scratch on worldwide web-scale image-text pairs spanning 300+ languages. The work addresses the challenge of scaling CLIP beyond English-only data while avoiding the &quot;curse of multilinguality&quot; - where multilingual models perform worse on English tasks than their English-only counterparts. <br /> Meta CLIP 2 presents a recipe for training CLIP models on worldwide multilingual data from scratch. The work demonstrates that the curse of multilinguality can be overcome through careful scaling of metadata construction, data curation algorithms, and training frameworks.<br /> <br /> MetaCLIP 2, trained on worldwide web-scale image-text pairs, improves zero-shot classification and multilingual benchmarks without system-level confounding factors.<br /> <br /> == Source ==<br /> <br /> [https://huggingface.co/papers/2507.22062 Source : huggingface]<br /> <br /> <br /> [[Catégorie:vocabulary]]</div>

Pitpitt