Tether AI Research taraf\u0131ndan geli\u015ftirilen Genesis III, bilim, teknoloji, m\u00fchendislik ve matematik alanlar\u0131nda daha yetkin k\u00fc\u00e7\u00fck yapay zek\u00e2 modelleri geli\u015ftirmek \u00fczere tasarlanan 191,43 milyar tokenl\u0131k sentetik veri setini sunuyor. Genesis III ile e\u011fitilen modeller, bilim ve STEM testlerinde a\u00e7\u0131k kaynakl\u0131 referanslar\u0131 geride b\u0131rak\u0131rken, yapay zek\u00e2n\u0131n g\u00fcnl\u00fck cihazlarda yerel olarak \u00e7al\u0131\u015ft\u0131r\u0131lmas\u0131na y\u00f6nelik yeni olanaklar sa\u011fl\u0131yor.<\/p>\n
Genesis III, daha k\u00fc\u00e7\u00fck yapay zek\u00e2 modellerinin performans\u0131n\u0131 art\u0131rarak yapay zek\u00e2n\u0131n do\u011frudan telefon, bilgisayar ve yerel sunucularda \u00e7al\u0131\u015ft\u0131r\u0131lmas\u0131na y\u00f6nelik olanaklar\u0131 geni\u015fletiyor. QVAC Genesis I ve II\u2019nin devam\u0131 olan veri seti, 159,6 milyon dok\u00fcmanda 191,43 milyar token i\u00e7eriyor ve QVAC Genesis III ara\u015ft\u0131rmas\u0131, COLM 2026\u2019da sunulmak \u00fczere kabul edildi.<\/p>\n
Genesis III, STEM e\u011fitiminde yeni olanaklar sunuyor<\/p>\n
Genesis III ile e\u011fitilen modeller, bilim ve STEM alanlar\u0131ndaki ak\u0131l y\u00fcr\u00fctme testlerinde a\u00e7\u0131k kaynakl\u0131 referanslar\u0131 geride b\u0131rakt\u0131. 1,7 milyar parametreli model, benchmark yan\u0131tlar\u0131n\u0131n y\u00fczde 99,45\u2019inde ge\u00e7erli sonu\u00e7 \u00fcretirken; veri seti, 19 STEM alan\u0131n\u0131 kapsayarak e\u011fitimde fizik problemlerini \u00e7\u00f6zebilen, hatalar\u0131 a\u00e7\u0131klayabilen ve ak\u0131l y\u00fcr\u00fctme s\u00fcre\u00e7lerini analiz edebilen yapay zek\u00e2 \u00f6\u011fretmenlerinin geli\u015ftirilmesine olanak sa\u011fl\u0131yor. Daha k\u00fc\u00e7\u00fck modellerin yerel olarak \u00e7al\u0131\u015fabilmesi, bu teknolojilerin s\u0131n\u0131rl\u0131 internet eri\u015fimine sahip ortamlarda da kullan\u0131lmas\u0131n\u0131 m\u00fcmk\u00fcn k\u0131l\u0131yor.<\/p>\n
Ak\u0131l y\u00fcr\u00fctmeye destek<\/p>\n
Ara\u015ft\u0131rmac\u0131lar ve profesyoneller i\u00e7in Genesis III, hassas bilgi ve sorular\u0131 s\u00fcrekli olarak harici sunuculara g\u00f6ndermeden yerel ortamda \u00e7al\u0131\u015fabilen uzmanla\u015fm\u0131\u015f teknik asistanlar\u0131n geli\u015ftirilmesini destekliyor. 159,6 milyon dok\u00fcmandan olu\u015fan veri seti, modellerin yaln\u0131zca do\u011fru cevab\u0131 bulmas\u0131na odaklanmak yerine, nas\u0131l d\u00fc\u015f\u00fcnd\u00fc\u011f\u00fcn\u00fc ve hatalar\u0131n\u0131 nas\u0131l d\u00fczeltece\u011fini \u00f6\u011frenmesini hedefliyor. Bu kapsamda Failure Analysis y\u00f6ntemi, modelin yapt\u0131\u011f\u0131 hatalar\u0131 analiz ederek yanl\u0131\u015f ak\u0131l y\u00fcr\u00fctmenin nedenini ve do\u011fru \u00e7\u00f6z\u00fcm yolunu \u00f6\u011fretirken; Option-Level Reasoning y\u00f6ntemi, do\u011fru cevab\u0131n neden do\u011fru oldu\u011funu ve di\u011fer se\u00e7eneklerin neden yanl\u0131\u015f oldu\u011funu a\u00e7\u0131kl\u0131yor. B\u00f6ylece Genesis III, daha k\u00fc\u00e7\u00fck yapay zek\u00e2 modellerinin daha az i\u015flem g\u00fcc\u00fcyle daha do\u011fru, g\u00fcvenilir ve a\u00e7\u0131klanabilir sonu\u00e7lar \u00fcretmesine katk\u0131 sa\u011flamay\u0131 ama\u00e7l\u0131yor.<\/p>\n
Performans kar\u015f\u0131la\u015ft\u0131rmas\u0131<\/p>\n
Ara\u015ft\u0131rmac\u0131lar, Genesis III\u2019\u00fcn Option-Level veri setiyle e\u011fitilen modellerin Qwen, Llama, SmolLM ve Gemma gibi farkl\u0131 mimarilerde Cosmopedia-v2\u2019ye k\u0131yasla daha y\u00fcksek performans g\u00f6sterdi\u011fini belirledi. Ayn\u0131 token say\u0131s\u0131yla kar\u015f\u0131la\u015ft\u0131r\u0131ld\u0131\u011f\u0131nda Genesis III, ARC-Easy\u2019de 28,57, ARC-Challenge\u2019da 21,35 ve MMLU STEM\u2019de 15,03 puan art\u0131\u015f sa\u011flad\u0131. Genesis III ile \u00f6n e\u011fitimden ge\u00e7irilen model, daha b\u00fcy\u00fck ve ek matematik ile kod verileriyle e\u011fitilmi\u015f Cosmo-1B\u2019yi de de\u011ferlendirilen testlerde geride b\u0131rakt\u0131.<\/p>\n
\u201cAmac\u0131m\u0131z, faydal\u0131 yapay zek\u00e2y\u0131 insanlar\u0131n sahip oldu\u011fu cihazlara ta\u015f\u0131mak\u201d<\/p>\n
Tether CEO\u2019su Paolo Ardoino, \u201cYapay zek\u00e2 sekt\u00f6r\u00fcn\u00fcn b\u00fcy\u00fck b\u00f6l\u00fcm\u00fc modelleri b\u00fcy\u00fctmeye ve onlara daha fazla i\u015flem g\u00fcc\u00fc sa\u011flamaya odakland\u0131. Genesis III ise daha k\u00fc\u00e7\u00fck modellerin \u00f6\u011frendi\u011fi veriyi iyile\u015ftirdi\u011fimizde neler olabilece\u011fini g\u00f6steriyor. STEM a\u00e7\u0131s\u0131ndan bu, bir modele yaln\u0131zca nihai cevab\u0131 \u00f6\u011fretmekten fazlas\u0131 anlam\u0131na geliyor. Bir \u015feyin neden do\u011fru oldu\u011funu, ak\u0131l y\u00fcr\u00fctmenin nerede hatal\u0131 oldu\u011funu ve hatan\u0131n nas\u0131l d\u00fczeltilece\u011fini \u00f6\u011fretmek anlam\u0131na geliyor. Amac\u0131m\u0131z, faydal\u0131 yapay zek\u00e2y\u0131 b\u00fcy\u00fck bulut altyap\u0131lar\u0131ndan insanlar\u0131n hali haz\u0131rda sahip oldu\u011fu cihazlara ta\u015f\u0131mak\u201d dedi.<\/p>\n
\u00a0<\/p>\n
QVAC Hakk\u0131nda<\/p>\n
QVAC, a\u00e7\u0131k, merkeziyetsiz ve uyarlanabilir yapay zek\u00e2 sistemleri geli\u015ftirmeye odaklanan Tether\u2019in ileri yapay zek\u00e2 ara\u015ft\u0131rma giri\u015fimidir. Misyonu Local AI ve Infinite Intelligence olarak tan\u0131mlan\u0131yor. QVAC; yapay zek\u00e2n\u0131n herhangi bir cihazda \u00e7al\u0131\u015ft\u0131\u011f\u0131 ve \u00f6\u011frendi\u011fi, g\u00fcc\u00fcn kurumsal veri merkezlerinde yo\u011funla\u015fmas\u0131 yerine bireyleri ve topluluklar\u0131 g\u00fc\u00e7lendiren bir d\u00fcnya vizyonuyla hareket ediyor.<\/p>\n
\u00a0<\/p>\n
Tether AI Research Hakk\u0131nda<\/p>\n
Tether AI Research, Tether\u2019in teknoloji arac\u0131l\u0131\u011f\u0131yla \u00f6zg\u00fcrl\u00fc\u011f\u00fc, \u015feffafl\u0131\u011f\u0131 ve inovasyonu geli\u015ftirme vizyonunun bir par\u00e7as\u0131d\u0131r. Misyonu, gereksiz arac\u0131lar olmadan insanlar\u0131n ve kurulu\u015flar\u0131n do\u011frudan ba\u011flant\u0131 kurmas\u0131n\u0131 ve bilgi payla\u015fmas\u0131n\u0131 sa\u011flamakt\u0131r. G\u00fcvenli, e\u015fler aras\u0131 sistemler geli\u015ftirerek Tether AI Research, kullan\u0131c\u0131lara verileri, ileti\u015fimleri ve dijital etkile\u015fimleri \u00fczerinde daha fazla kontrol sunmay\u0131 ama\u00e7l\u0131yor. Tether AI Research, merkezi modellerin yerini gizlilik, verimlilik ve dayan\u0131kl\u0131l\u0131k i\u00e7in tasarlanm\u0131\u015f merkeziyetsiz altyap\u0131n\u0131n ald\u0131\u011f\u0131 yeni bir bilgi ak\u0131\u015f\u0131 modeli geli\u015ftirmeyi hedefliyor.<\/p>\n
<\/strong> Tether AI Research, bilim, teknoloji, m\u00fchendislik ve matematik alanlar\u0131nda daha yetkin k\u00fc\u00e7\u00fck yapay zek\u00e2 modelleri geli\u015ftirmek amac\u0131yla tasarlanan 191,43 milyar tokenl\u0131k sentetik veri seti QVAC Genesis III\u2019\u00fc yay\u0131mlad\u0131.<\/p>\n","protected":false},"author":1,"featured_media":158168,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"googlesitekit_rrm_CAow07ekDA:productID":"","footnotes":""},"categories":[4,1,9],"tags":[],"class_list":["post-158167","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-dunya","category-ekonomi","category-manset"],"_links":{"self":[{"href":"https:\/\/ekonomigundemi.com.tr\/index.php\/wp-json\/wp\/v2\/posts\/158167","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/ekonomigundemi.com.tr\/index.php\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/ekonomigundemi.com.tr\/index.php\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/ekonomigundemi.com.tr\/index.php\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/ekonomigundemi.com.tr\/index.php\/wp-json\/wp\/v2\/comments?post=158167"}],"version-history":[{"count":0,"href":"https:\/\/ekonomigundemi.com.tr\/index.php\/wp-json\/wp\/v2\/posts\/158167\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/ekonomigundemi.com.tr\/index.php\/wp-json\/wp\/v2\/media\/158168"}],"wp:attachment":[{"href":"https:\/\/ekonomigundemi.com.tr\/index.php\/wp-json\/wp\/v2\/media?parent=158167"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/ekonomigundemi.com.tr\/index.php\/wp-json\/wp\/v2\/categories?post=158167"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/ekonomigundemi.com.tr\/index.php\/wp-json\/wp\/v2\/tags?post=158167"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}
\nEKONOM\u0130 GAZETES\u0130 <\/p>\n","protected":false},"excerpt":{"rendered":"