{"id":86975,"date":"2026-07-27T22:50:56","date_gmt":"2026-07-27T19:50:56","guid":{"rendered":"https:\/\/twodots.gr\/?p=86975"},"modified":"2026-07-27T22:50:57","modified_gmt":"2026-07-27T19:50:57","slug":"craft-apo-ta-scores-sti-stochevmeni-veltiosi-ton-montelon-ai","status":"publish","type":"post","link":"https:\/\/twodots.gr\/en\/craft-apo-ta-scores-sti-stochevmeni-veltiosi-ton-montelon-ai\/","title":{"rendered":"CRAFT: From Scores to Targeted Improvement of AI Models"},"content":{"rendered":"<div class=\"td-article-lede\">\n<p><strong>\u03a4\u03bf CRAFT \u03bc\u03b5\u03c4\u03b1\u03c4\u03c1\u03ad\u03c0\u03b5\u03b9 \u03c4\u03b7\u03bd \u03b1\u03be\u03b9\u03bf\u03bb\u03cc\u03b3\u03b7\u03c3\u03b7 \u03b5\u03bd\u03cc\u03c2 \u03bc\u03bf\u03bd\u03c4\u03ad\u03bb\u03bf\u03c5 AI \u03b1\u03c0\u03cc score \u03c3\u03b5 \u03c3\u03c7\u03ad\u03b4\u03b9\u03bf \u03c3\u03c4\u03bf\u03c7\u03b5\u03c5\u03bc\u03ad\u03bd\u03b7\u03c2 \u03b2\u03b5\u03bb\u03c4\u03af\u03c9\u03c3\u03b7\u03c2.<\/strong> \u0391\u03bd\u03c4\u03af \u03bd\u03b1 \u03bb\u03ad\u03b5\u03b9 \u03bc\u03cc\u03bd\u03bf \u03c0\u03bf\u03cd \u03b1\u03c0\u03ad\u03c4\u03c5\u03c7\u03b5 \u03c4\u03bf \u03bc\u03bf\u03bd\u03c4\u03ad\u03bb\u03bf, \u03b1\u03bd\u03c4\u03b9\u03bc\u03b5\u03c4\u03c9\u03c0\u03af\u03b6\u03b5\u03b9 \u03ba\u03ac\u03b8\u03b5 \u03ba\u03c1\u03b9\u03c4\u03ae\u03c1\u03b9\u03bf \u03bc\u03b9\u03b1\u03c2 rubric \u03c9\u03c2 \u03b1\u03bd\u03b9\u03c7\u03bd\u03b5\u03c5\u03c4\u03ae \u03b9\u03ba\u03b1\u03bd\u03cc\u03c4\u03b7\u03c4\u03b1\u03c2, \u03bf\u03c1\u03b3\u03b1\u03bd\u03ce\u03bd\u03b5\u03b9 \u03c4\u03b1 \u03ba\u03c1\u03b9\u03c4\u03ae\u03c1\u03b9\u03b1 \u03c3\u03b5 \u03b9\u03b5\u03c1\u03b1\u03c1\u03c7\u03af\u03b1 \u03ba\u03b1\u03b9 \u03b5\u03bd\u03c4\u03bf\u03c0\u03af\u03b6\u03b5\u03b9 \u03c4\u03bf\u03c5\u03c2 \u03b1\u03b4\u03cd\u03bd\u03b1\u03bc\u03bf\u03c5\u03c2 \u03ba\u03cc\u03bc\u03b2\u03bf\u03c5\u03c2 \u03c3\u03c4\u03b7 \u03c3\u03c9\u03c3\u03c4\u03ae \u03ba\u03bb\u03af\u03bc\u03b1\u03ba\u03b1.<\/p>\n<p>\u0397 \u03bc\u03ad\u03b8\u03bf\u03b4\u03bf\u03c2 \u03c4\u03c9\u03bd Vipul Gupta, Zihao Wang, Razvan-Gabriel Dumitru, MohammadHossein Rezaei, Aakash Sabharwal \u03ba\u03b1\u03b9 Yunzhong He \u03c3\u03c5\u03bd\u03b4\u03ad\u03b5\u03b9 \u03c4\u03b7 \u03b4\u03b9\u03ac\u03b3\u03bd\u03c9\u03c3\u03b7 \u03bc\u03b5 \u03c4\u03bf post-training: \u03bf\u03b9 \u03bc\u03b5\u03c4\u03c1\u03b7\u03bc\u03ad\u03bd\u03b5\u03c2 \u03b1\u03b4\u03c5\u03bd\u03b1\u03bc\u03af\u03b5\u03c2 \u03b3\u03af\u03bd\u03bf\u03bd\u03c4\u03b1\u03b9 brief \u03b3\u03b9\u03b1 \u03c4\u03b7 \u03b4\u03b7\u03bc\u03b9\u03bf\u03c5\u03c1\u03b3\u03af\u03b1 \u03c3\u03c4\u03bf\u03c7\u03b5\u03c5\u03bc\u03ad\u03bd\u03c9\u03bd supervised fine-tuning \u03c0\u03b1\u03c1\u03b1\u03b4\u03b5\u03b9\u03b3\u03bc\u03ac\u03c4\u03c9\u03bd. \u0393\u03b9\u03b1 \u03bc\u03b9\u03b1 \u03b5\u03c0\u03b9\u03c7\u03b5\u03af\u03c1\u03b7\u03c3\u03b7, \u03b7 \u03b1\u03be\u03af\u03b1 \u03b2\u03c1\u03af\u03c3\u03ba\u03b5\u03c4\u03b1\u03b9 \u03c3\u03c4\u03b7 \u03bc\u03b5\u03c4\u03ac\u03b2\u03b1\u03c3\u03b7 \u03b1\u03c0\u03cc \u03c4\u03bf \u03b3\u03b5\u03bd\u03b9\u03ba\u03cc \u00ab\u03c7\u03c1\u03b5\u03b9\u03ac\u03b6\u03b5\u03c4\u03b1\u03b9 \u03ba\u03b1\u03bb\u03cd\u03c4\u03b5\u03c1\u03bf \u03bc\u03bf\u03bd\u03c4\u03ad\u03bb\u03bf\u00bb \u03c3\u03c4\u03bf \u03c3\u03c5\u03b3\u03ba\u03b5\u03ba\u03c1\u03b9\u03bc\u03ad\u03bd\u03bf \u00ab\u03c0\u03bf\u03b9\u03b1 \u03b9\u03ba\u03b1\u03bd\u03cc\u03c4\u03b7\u03c4\u03b1 \u03bb\u03b5\u03af\u03c0\u03b5\u03b9 \u03ba\u03b1\u03b9 \u03c0\u03bf\u03b9\u03b1 \u03b4\u03b5\u03b4\u03bf\u03bc\u03ad\u03bd\u03b1 \u03c0\u03c1\u03ad\u03c0\u03b5\u03b9 \u03bd\u03b1 \u03c3\u03c5\u03bb\u03bb\u03b5\u03c7\u03b8\u03bf\u03cd\u03bd\u00bb.<\/p>\n<\/div>\n<div class=\"td-article-note\">\n<p><strong>\u038c\u03c1\u03b9\u03bf \u03b3\u03b5\u03bd\u03af\u03ba\u03b5\u03c5\u03c3\u03b7\u03c2:<\/strong> \u03b7 \u03b5\u03c1\u03b3\u03b1\u03c3\u03af\u03b1 \u03b1\u03be\u03b9\u03bf\u03bb\u03cc\u03b3\u03b7\u03c3\u03b5 \u03c4\u03ad\u03c3\u03c3\u03b5\u03c1\u03b1 open-source \u03bc\u03bf\u03bd\u03c4\u03ad\u03bb\u03b1 \u03bc\u03cc\u03bd\u03bf \u03c3\u03b5 \u03bd\u03bf\u03bc\u03b9\u03ba\u03ac \u03ba\u03b1\u03b9 \u03c7\u03c1\u03b7\u03bc\u03b1\u03c4\u03bf\u03bf\u03b9\u03ba\u03bf\u03bd\u03bf\u03bc\u03b9\u03ba\u03ac \u03b4\u03b5\u03b4\u03bf\u03bc\u03ad\u03bd\u03b1. \u0397 \u03c7\u03c1\u03ae\u03c3\u03b7 \u03c4\u03b7\u03c2 \u03bb\u03bf\u03b3\u03b9\u03ba\u03ae\u03c2 \u03c4\u03bf\u03c5 CRAFT \u03c3\u03b5 e-commerce, customer support \u03ae marketing \u03b5\u03af\u03bd\u03b1\u03b9 \u03c4\u03b5\u03ba\u03bc\u03b7\u03c1\u03b9\u03c9\u03bc\u03ad\u03bd\u03b7 \u03b5\u03c0\u03b9\u03c7\u03b5\u03b9\u03c1\u03b7\u03c3\u03b9\u03b1\u03ba\u03ae \u03c0\u03c1\u03bf\u03c3\u03b1\u03c1\u03bc\u03bf\u03b3\u03ae \u03c4\u03bf\u03c5 design pattern, \u03cc\u03c7\u03b9 \u03b1\u03c0\u03bf\u03c4\u03ad\u03bb\u03b5\u03c3\u03bc\u03b1 \u03c0\u03bf\u03c5 \u03bc\u03ad\u03c4\u03c1\u03b7\u03c3\u03b1\u03bd \u03bf\u03b9 \u03c3\u03c5\u03b3\u03b3\u03c1\u03b1\u03c6\u03b5\u03af\u03c2.<\/p>\n<\/div>\n<div class=\"td-article-toc\">\n<div class=\"td-toc-title\">Contents<\/div>\n<ul>\n<li><a href=\"#keno-pou-giati-apotychias\">\u03a4\u03bf \u03ba\u03b5\u03bd\u03cc \u03b1\u03bd\u03ac\u03bc\u03b5\u03c3\u03b1 \u03c3\u03c4\u03bf \u00ab\u03c0\u03bf\u03cd\u00bb \u03ba\u03b1\u03b9 \u03c3\u03c4\u03bf \u00ab\u03b3\u03b9\u03b1\u03c4\u03af\u00bb \u03c4\u03b7\u03c2 \u03b1\u03c0\u03bf\u03c4\u03c5\u03c7\u03af\u03b1\u03c2<\/a><\/li>\n<li><a href=\"#criterion-perigrafi-ikanotitas\">\u03a0\u03ce\u03c2 \u03ad\u03bd\u03b1 \u03ba\u03c1\u03b9\u03c4\u03ae\u03c1\u03b9\u03bf \u03b3\u03af\u03bd\u03b5\u03c4\u03b1\u03b9 \u03c0\u03b5\u03c1\u03b9\u03b3\u03c1\u03b1\u03c6\u03ae \u03b9\u03ba\u03b1\u03bd\u03cc\u03c4\u03b7\u03c4\u03b1\u03c2<\/a><\/li>\n<li><a href=\"#ierarchia-ikanotiton-polla-epipeda\">\u0397 \u03b9\u03b5\u03c1\u03b1\u03c1\u03c7\u03af\u03b1 \u03b9\u03ba\u03b1\u03bd\u03bf\u03c4\u03ae\u03c4\u03c9\u03bd \u03ba\u03b1\u03b9 \u03b3\u03b9\u03b1\u03c4\u03af \u03b4\u03b5\u03bd \u03b1\u03c1\u03ba\u03b5\u03af \u03ad\u03bd\u03b1 \u03b5\u03c0\u03af\u03c0\u03b5\u03b4\u03bf<\/a><\/li>\n<li><a href=\"#capability-tree-adynamoi-komvoi\">\u0391\u03c0\u03cc \u03c4\u03bf capability tree \u03c3\u03c4\u03bf\u03c5\u03c2 \u03b1\u03b4\u03cd\u03bd\u03b1\u03bc\u03bf\u03c5\u03c2 \u03ba\u03cc\u03bc\u03b2\u03bf\u03c5\u03c2<\/a><\/li>\n<li><a href=\"#axiologisi-brief-training-data\">\u0397 \u03b1\u03be\u03b9\u03bf\u03bb\u03cc\u03b3\u03b7\u03c3\u03b7 \u03b3\u03af\u03bd\u03b5\u03c4\u03b1\u03b9 brief \u03b3\u03b9\u03b1 \u03c4\u03b7 \u03b4\u03b7\u03bc\u03b9\u03bf\u03c5\u03c1\u03b3\u03af\u03b1 training data<\/a><\/li>\n<li><a href=\"#baselines-evaltree-random\">\u039f\u03b9 \u03b4\u03cd\u03bf baselines \u03ba\u03b1\u03b9 \u03c4\u03b9 \u03b1\u03ba\u03c1\u03b9\u03b2\u03ce\u03c2 \u03b5\u03bb\u03ad\u03b3\u03c7\u03bf\u03c5\u03bd<\/a><\/li>\n<li><a href=\"#montela-domains-benchmarks\">\u03a4\u03bf \u03c0\u03b5\u03b9\u03c1\u03b1\u03bc\u03b1\u03c4\u03b9\u03ba\u03cc \u03c0\u03b5\u03b4\u03af\u03bf: \u03c4\u03ad\u03c3\u03c3\u03b5\u03c1\u03b1 \u03bc\u03bf\u03bd\u03c4\u03ad\u03bb\u03b1 \u03ba\u03b1\u03b9 \u03b4\u03cd\u03bf \u03b1\u03c0\u03b1\u03b9\u03c4\u03b7\u03c4\u03b9\u03ba\u03ac domains<\/a><\/li>\n<li><a href=\"#apotelesmata-nomika-benchmarks\">\u03a4\u03b9 \u03ad\u03b4\u03b5\u03b9\u03be\u03b1\u03bd \u03c4\u03b1 \u03b1\u03c0\u03bf\u03c4\u03b5\u03bb\u03ad\u03c3\u03bc\u03b1\u03c4\u03b1 \u03c3\u03c4\u03b1 \u03bd\u03bf\u03bc\u03b9\u03ba\u03ac benchmarks<\/a><\/li>\n<li><a href=\"#apotelesmata-chrimatooikonomika\">\u0397 \u03c0\u03b9\u03bf \u03ba\u03b1\u03b8\u03b1\u03c1\u03ae \u03b5\u03b9\u03ba\u03cc\u03bd\u03b1 \u03c3\u03c4\u03b1 \u03c7\u03c1\u03b7\u03bc\u03b1\u03c4\u03bf\u03bf\u03b9\u03ba\u03bf\u03bd\u03bf\u03bc\u03b9\u03ba\u03ac<\/a><\/li>\n<li><a href=\"#dynamiki-epilogi-vathous\">\u0393\u03b9\u03b1\u03c4\u03af \u03b7 \u03b4\u03c5\u03bd\u03b1\u03bc\u03b9\u03ba\u03ae \u03b5\u03c0\u03b9\u03bb\u03bf\u03b3\u03ae \u03b2\u03ac\u03b8\u03bf\u03c5\u03c2 \u03ad\u03c7\u03b5\u03b9 \u03c0\u03c1\u03b1\u03ba\u03c4\u03b9\u03ba\u03ae \u03b1\u03be\u03af\u03b1<\/a><\/li>\n<li><a href=\"#craft-epicheirisiako-ai-workflow\">\u03a0\u03ce\u03c2 \u03bc\u03b5\u03c4\u03b1\u03c6\u03c1\u03ac\u03b6\u03b5\u03c4\u03b1\u03b9 \u03b7 \u03ad\u03c1\u03b5\u03c5\u03bd\u03b1 \u03c3\u03b5 \u03b5\u03c0\u03b9\u03c7\u03b5\u03b9\u03c1\u03b7\u03c3\u03b9\u03b1\u03ba\u03cc AI workflow<\/a><\/li>\n<li><a href=\"#proypotheseis-oria-craft\">\u039f\u03b9 \u03c0\u03c1\u03bf\u03cb\u03c0\u03bf\u03b8\u03ad\u03c3\u03b5\u03b9\u03c2 \u03ba\u03b1\u03b9 \u03c4\u03b1 \u03cc\u03c1\u03b9\u03b1 \u03c0\u03bf\u03c5 \u03b4\u03b5\u03bd \u03c0\u03c1\u03ad\u03c0\u03b5\u03b9 \u03bd\u03b1 \u03b1\u03b3\u03bd\u03bf\u03b7\u03b8\u03bf\u03cd\u03bd<\/a><\/li>\n<li><a href=\"#dashboard-systima-synechous-mathisis\">\u0391\u03c0\u03cc dashboard \u03b5\u03c0\u03af\u03b4\u03bf\u03c3\u03b7\u03c2 \u03c3\u03b5 \u03c3\u03cd\u03c3\u03c4\u03b7\u03bc\u03b1 \u03c3\u03c5\u03bd\u03b5\u03c7\u03bf\u03cd\u03c2 \u03bc\u03ac\u03b8\u03b7\u03c3\u03b7\u03c2<\/a><\/li>\n<\/ul>\n<\/div>\n<h2 id=\"keno-pou-giati-apotychias\">\u03a4\u03bf \u03ba\u03b5\u03bd\u03cc \u03b1\u03bd\u03ac\u03bc\u03b5\u03c3\u03b1 \u03c3\u03c4\u03bf \u00ab\u03c0\u03bf\u03cd\u00bb \u03ba\u03b1\u03b9 \u03c3\u03c4\u03bf \u00ab\u03b3\u03b9\u03b1\u03c4\u03af\u00bb \u03c4\u03b7\u03c2 \u03b1\u03c0\u03bf\u03c4\u03c5\u03c7\u03af\u03b1\u03c2<\/h2>\n<p>\u039f\u03b9 \u03c0\u03b5\u03c1\u03b9\u03c3\u03c3\u03cc\u03c4\u03b5\u03c1\u03b5\u03c2 \u03b1\u03be\u03b9\u03bf\u03bb\u03bf\u03b3\u03ae\u03c3\u03b5\u03b9\u03c2 \u03bf\u03c1\u03b3\u03b1\u03bd\u03ce\u03bd\u03bf\u03c5\u03bd \u03c4\u03b1 \u03b1\u03c0\u03bf\u03c4\u03b5\u03bb\u03ad\u03c3\u03bc\u03b1\u03c4\u03b1 \u03b1\u03bd\u03ac task, \u03b8\u03ad\u03bc\u03b1 \u03ae \u03ba\u03b1\u03c4\u03b7\u03b3\u03bf\u03c1\u03af\u03b1. \u0388\u03c4\u03c3\u03b9, \u03bc\u03b9\u03b1 \u03bf\u03bc\u03ac\u03b4\u03b1 \u03bc\u03c0\u03bf\u03c1\u03b5\u03af \u03bd\u03b1 \u03bc\u03ac\u03b8\u03b5\u03b9 \u03cc\u03c4\u03b9 \u03c4\u03bf \u03bc\u03bf\u03bd\u03c4\u03ad\u03bb\u03bf \u03c4\u03b7\u03c2 \u03b5\u03af\u03bd\u03b1\u03b9 \u03b1\u03b4\u03cd\u03bd\u03b1\u03bc\u03bf \u03c3\u03c4\u03b1 \u03c7\u03c1\u03b7\u03bc\u03b1\u03c4\u03bf\u03bf\u03b9\u03ba\u03bf\u03bd\u03bf\u03bc\u03b9\u03ba\u03ac \u03ae \u03c3\u03c4\u03b1 \u03bc\u03b1\u03b8\u03b7\u03bc\u03b1\u03c4\u03b9\u03ba\u03ac. \u0391\u03c5\u03c4\u03ae \u03b7 \u03b4\u03b9\u03ac\u03b3\u03bd\u03c9\u03c3\u03b7, \u03cc\u03bc\u03c9\u03c2, \u03c0\u03b1\u03c1\u03b1\u03bc\u03ad\u03bd\u03b5\u03b9 \u03c5\u03c0\u03b5\u03c1\u03b2\u03bf\u03bb\u03b9\u03ba\u03ac \u03b3\u03b5\u03bd\u03b9\u03ba\u03ae \u03b3\u03b9\u03b1 \u03bd\u03b1 \u03ba\u03b1\u03b8\u03bf\u03b4\u03b7\u03b3\u03ae\u03c3\u03b5\u03b9 \u03c4\u03b7 \u03c3\u03c5\u03bb\u03bb\u03bf\u03b3\u03ae \u03b4\u03b5\u03b4\u03bf\u03bc\u03ad\u03bd\u03c9\u03bd. \u03a3\u03c4\u03bf \u03af\u03b4\u03b9\u03bf \u03bc\u03b1\u03b8\u03b7\u03bc\u03b1\u03c4\u03b9\u03ba\u03cc \u03c0\u03c1\u03cc\u03b2\u03bb\u03b7\u03bc\u03b1, \u03b3\u03b9\u03b1 \u03c0\u03b1\u03c1\u03ac\u03b4\u03b5\u03b9\u03b3\u03bc\u03b1, \u03c4\u03bf \u03bc\u03bf\u03bd\u03c4\u03ad\u03bb\u03bf \u03bc\u03c0\u03bf\u03c1\u03b5\u03af \u03bd\u03b1 \u03b5\u03c0\u03b9\u03bb\u03ad\u03be\u03b5\u03b9 \u03bb\u03ac\u03b8\u03bf\u03c2 \u03b5\u03be\u03af\u03c3\u03c9\u03c3\u03b7, \u03bd\u03b1 \u03ba\u03ac\u03bd\u03b5\u03b9 \u03b1\u03c1\u03b9\u03b8\u03bc\u03b7\u03c4\u03b9\u03ba\u03cc \u03bb\u03ac\u03b8\u03bf\u03c2, \u03bd\u03b1 \u03c0\u03b1\u03c1\u03b1\u03bb\u03b5\u03af\u03c8\u03b5\u03b9 \u03ad\u03bd\u03b1 \u03b1\u03bd\u03b1\u03b3\u03ba\u03b1\u03af\u03bf \u03b5\u03bd\u03b4\u03b9\u03ac\u03bc\u03b5\u03c3\u03bf \u03b2\u03ae\u03bc\u03b1 \u03ae \u03bd\u03b1 \u03bc\u03b7 \u03b3\u03c1\u03ac\u03c8\u03b5\u03b9 \u03c4\u03b9\u03c2 \u03b1\u03c0\u03b1\u03b9\u03c4\u03bf\u03cd\u03bc\u03b5\u03bd\u03b5\u03c2 \u03bc\u03bf\u03bd\u03ac\u03b4\u03b5\u03c2. \u039a\u03ac\u03b8\u03b5 \u03b1\u03c0\u03bf\u03c4\u03c5\u03c7\u03af\u03b1 \u03b1\u03c0\u03b1\u03b9\u03c4\u03b5\u03af \u03b4\u03b9\u03b1\u03c6\u03bf\u03c1\u03b5\u03c4\u03b9\u03ba\u03ae \u03c0\u03b1\u03c1\u03ad\u03bc\u03b2\u03b1\u03c3\u03b7.<\/p>\n<p>\u039f\u03b9 rubrics \u03c0\u03c1\u03bf\u03c3\u03c6\u03ad\u03c1\u03bf\u03c5\u03bd \u03c0\u03b9\u03bf \u03bb\u03b5\u03c0\u03c4\u03bf\u03bc\u03b5\u03c1\u03ae \u03c0\u03b1\u03c1\u03b1\u03c4\u03ae\u03c1\u03b7\u03c3\u03b7, \u03b5\u03c0\u03b5\u03b9\u03b4\u03ae \u03c7\u03c9\u03c1\u03af\u03b6\u03bf\u03c5\u03bd \u03c4\u03b7\u03bd \u03c0\u03bf\u03b9\u03cc\u03c4\u03b7\u03c4\u03b1 \u03bc\u03b9\u03b1\u03c2 \u03b1\u03c0\u03ac\u03bd\u03c4\u03b7\u03c3\u03b7\u03c2 \u03c3\u03b5 \u03c1\u03b7\u03c4\u03ad\u03c2 \u03b1\u03c0\u03b1\u03b9\u03c4\u03ae\u03c3\u03b5\u03b9\u03c2. \u039c\u03b9\u03b1 \u03b1\u03c0\u03ac\u03bd\u03c4\u03b7\u03c3\u03b7 \u03bc\u03c0\u03bf\u03c1\u03b5\u03af \u03bd\u03b1 \u03c0\u03b5\u03c1\u03bd\u03ac \u03c4\u03bf \u03ba\u03c1\u03b9\u03c4\u03ae\u03c1\u03b9\u03bf \u03c4\u03b7\u03c2 \u03c0\u03c1\u03b1\u03b3\u03bc\u03b1\u03c4\u03bf\u03bb\u03bf\u03b3\u03b9\u03ba\u03ae\u03c2 \u03c4\u03b5\u03ba\u03bc\u03b7\u03c1\u03af\u03c9\u03c3\u03b7\u03c2, \u03b1\u03bb\u03bb\u03ac \u03bd\u03b1 \u03b1\u03c0\u03bf\u03c4\u03c5\u03b3\u03c7\u03ac\u03bd\u03b5\u03b9 \u03c3\u03c4\u03b7\u03bd \u03c0\u03bb\u03b7\u03c1\u03cc\u03c4\u03b7\u03c4\u03b1. \u039c\u03c0\u03bf\u03c1\u03b5\u03af \u03bd\u03b1 \u03b5\u03ba\u03c4\u03b5\u03bb\u03b5\u03af \u03c3\u03c9\u03c3\u03c4\u03ac \u03c4\u03bf\u03bd \u03c3\u03c5\u03bb\u03bb\u03bf\u03b3\u03b9\u03c3\u03bc\u03cc, \u03b1\u03bb\u03bb\u03ac \u03bd\u03b1 \u03b4\u03af\u03bd\u03b5\u03b9 \u03c4\u03bf \u03c4\u03b5\u03bb\u03b9\u03ba\u03cc \u03b1\u03c0\u03bf\u03c4\u03ad\u03bb\u03b5\u03c3\u03bc\u03b1 \u03c3\u03b5 \u03bb\u03ac\u03b8\u03bf\u03c2 \u03bc\u03bf\u03c1\u03c6\u03ae. \u03a4\u03bf CRAFT \u03b1\u03be\u03b9\u03bf\u03c0\u03bf\u03b9\u03b5\u03af \u03b1\u03c5\u03c4\u03ac \u03c4\u03b1 \u03ba\u03c1\u03b9\u03c4\u03ae\u03c1\u03b9\u03b1 \u03cc\u03c7\u03b9 \u03bc\u03cc\u03bd\u03bf \u03c9\u03c2 \u03bc\u03b7\u03c7\u03b1\u03bd\u03b9\u03c3\u03bc\u03cc \u03b2\u03b1\u03b8\u03bc\u03bf\u03bb\u03cc\u03b3\u03b7\u03c3\u03b7\u03c2, \u03b1\u03bb\u03bb\u03ac \u03c9\u03c2 \u03bc\u03b9\u03ba\u03c1\u03ac, \u03b5\u03c0\u03b1\u03bd\u03b1\u03bb\u03b1\u03bc\u03b2\u03b1\u03bd\u03cc\u03bc\u03b5\u03bd\u03b1 probes \u03c0\u03bf\u03c5 \u03b1\u03c0\u03bf\u03ba\u03b1\u03bb\u03cd\u03c0\u03c4\u03bf\u03c5\u03bd \u03c4\u03b9\u03c2 \u03b9\u03ba\u03b1\u03bd\u03cc\u03c4\u03b7\u03c4\u03b5\u03c2 \u03c4\u03b9\u03c2 \u03bf\u03c0\u03bf\u03af\u03b5\u03c2 \u03b1\u03c0\u03b1\u03b9\u03c4\u03b5\u03af \u03ba\u03ac\u03b8\u03b5 \u03ba\u03b1\u03bb\u03ae \u03b1\u03c0\u03ac\u03bd\u03c4\u03b7\u03c3\u03b7.<\/p>\n<div class=\"td-comparison\">\n<p class=\"td-comparison-title\">\u0391\u03c0\u03cc \u03c4\u03b7\u03bd \u03ad\u03bd\u03b4\u03b5\u03b9\u03be\u03b7 \u03b1\u03c0\u03bf\u03c4\u03c5\u03c7\u03af\u03b1\u03c2 \u03c3\u03c4\u03b7 \u03b4\u03b9\u03bf\u03c1\u03b8\u03c9\u03c4\u03b9\u03ba\u03ae \u03b5\u03bd\u03ad\u03c1\u03b3\u03b5\u03b9\u03b1<\/p>\n<div class=\"td-comparison-cards td-comparison-cards--horizontal\">\n<div class=\"td-comparison-grid td-comparison-grid--two\">\n<div class=\"td-platform-card\">\n<h3>Aggregate score<\/h3>\n<p>\u0394\u03b5\u03af\u03c7\u03bd\u03b5\u03b9 \u03b1\u03bd \u03b7 \u03c3\u03c5\u03bd\u03bf\u03bb\u03b9\u03ba\u03ae \u03b5\u03c0\u03af\u03b4\u03bf\u03c3\u03b7 \u03b1\u03bd\u03ad\u03b2\u03b7\u03ba\u03b5 \u03ae \u03ad\u03c0\u03b5\u03c3\u03b5, \u03b1\u03bb\u03bb\u03ac \u03c3\u03c5\u03b3\u03c7\u03c9\u03bd\u03b5\u03cd\u03b5\u03b9 \u03b4\u03b9\u03b1\u03c6\u03bf\u03c1\u03b5\u03c4\u03b9\u03ba\u03ad\u03c2 \u03b1\u03b9\u03c4\u03af\u03b5\u03c2: \u03bb\u03ac\u03b8\u03bf\u03c2 \u03b3\u03bd\u03ce\u03c3\u03b7, \u03b5\u03bb\u03bb\u03b9\u03c0\u03ae \u03c3\u03c5\u03bb\u03bb\u03bf\u03b3\u03b9\u03c3\u03bc\u03cc, \u03c0\u03b1\u03c1\u03ac\u03bb\u03b5\u03b9\u03c8\u03b7 \u03b1\u03c0\u03b1\u03af\u03c4\u03b7\u03c3\u03b7\u03c2 \u03ae \u03bb\u03ac\u03b8\u03bf\u03c2 \u03bc\u03bf\u03c1\u03c6\u03ae \u03b1\u03c0\u03ac\u03bd\u03c4\u03b7\u03c3\u03b7\u03c2.<\/p>\n<div class=\"td-badge-row\"><span class=\"td-badge\">Measurement<\/span><span class=\"td-badge\">\u03a3\u03cd\u03b3\u03ba\u03c1\u03b9\u03c3\u03b7<\/span><span class=\"td-badge\">\u0393\u03b5\u03bd\u03b9\u03ba\u03cc \u03c3\u03ae\u03bc\u03b1<\/span><\/div>\n<\/div>\n<div class=\"td-platform-card td-platform-card--navy\">\n<h3>CRAFT capability node<\/h3>\n<p>\u03a3\u03c5\u03bd\u03b4\u03ad\u03b5\u03b9 \u03c3\u03c5\u03b3\u03ba\u03b5\u03ba\u03c1\u03b9\u03bc\u03ad\u03bd\u03b1 rubric failures \u03bc\u03b5 \u03bc\u03b9\u03b1 \u03b5\u03c0\u03b1\u03bd\u03b1\u03bb\u03b1\u03bc\u03b2\u03b1\u03bd\u03cc\u03bc\u03b5\u03bd\u03b7 \u03b9\u03ba\u03b1\u03bd\u03cc\u03c4\u03b7\u03c4\u03b1, \u03c4\u03bf pass rate \u03ba\u03b1\u03b9 \u03c4\u03b1 \u03c0\u03b1\u03c1\u03b1\u03b4\u03b5\u03af\u03b3\u03bc\u03b1\u03c4\u03b1 \u03c0\u03bf\u03c5 \u03c4\u03b7 \u03c3\u03c4\u03b7\u03c1\u03af\u03b6\u03bf\u03c5\u03bd, \u03ce\u03c3\u03c4\u03b5 \u03b7 \u03b5\u03c0\u03cc\u03bc\u03b5\u03bd\u03b7 \u03c3\u03c5\u03bb\u03bb\u03bf\u03b3\u03ae training data \u03bd\u03b1 \u03ad\u03c7\u03b5\u03b9 \u03c3\u03b1\u03c6\u03ae \u03c3\u03c4\u03cc\u03c7\u03bf.<\/p>\n<div class=\"td-badge-row\"><span class=\"td-badge\">\u0394\u03b9\u03ac\u03b3\u03bd\u03c9\u03c3\u03b7<\/span><span class=\"td-badge\">Granularity<\/span><span class=\"td-badge\">Training target<\/span><\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n<p>\u0397 \u03ba\u03c1\u03af\u03c3\u03b9\u03bc\u03b7 \u03bc\u03b5\u03c4\u03b1\u03c4\u03cc\u03c0\u03b9\u03c3\u03b7 \u03b5\u03af\u03bd\u03b1\u03b9 \u03cc\u03c4\u03b9 \u03b7 \u03bc\u03ad\u03b8\u03bf\u03b4\u03bf\u03c2 \u03b4\u03b5\u03bd \u03ba\u03bf\u03b9\u03c4\u03ac\u03b6\u03b5\u03b9 \u03bc\u03b5\u03bc\u03bf\u03bd\u03c9\u03bc\u03ad\u03bd\u03b1 \u03ba\u03ac\u03b8\u03b5 rubric. \u03a3\u03c5\u03b3\u03ba\u03b5\u03bd\u03c4\u03c1\u03ce\u03bd\u03b5\u03b9 \u03c7\u03b9\u03bb\u03b9\u03ac\u03b4\u03b5\u03c2 \u03ba\u03c1\u03b9\u03c4\u03ae\u03c1\u03b9\u03b1 \u03b1\u03c0\u03cc \u03b4\u03b9\u03b1\u03c6\u03bf\u03c1\u03b5\u03c4\u03b9\u03ba\u03ac prompts \u03ba\u03b1\u03b9 \u03b1\u03bd\u03b1\u03b6\u03b7\u03c4\u03ac \u03b5\u03c0\u03b1\u03bd\u03b1\u03bb\u03b1\u03bc\u03b2\u03b1\u03bd\u03cc\u03bc\u03b5\u03bd\u03b1 \u03bc\u03bf\u03c4\u03af\u03b2\u03b1 \u03b9\u03ba\u03b1\u03bd\u03cc\u03c4\u03b7\u03c4\u03b1\u03c2. \u039c\u03b5 \u03b1\u03c5\u03c4\u03cc\u03bd \u03c4\u03bf\u03bd \u03c4\u03c1\u03cc\u03c0\u03bf, \u03bc\u03b9\u03b1 \u03b1\u03c0\u03bf\u03c4\u03c5\u03c7\u03af\u03b1 \u03c0\u03bf\u03c5 \u03b8\u03b1 \u03ad\u03bc\u03bf\u03b9\u03b1\u03b6\u03b5 \u03b1\u03c0\u03bf\u03bc\u03bf\u03bd\u03c9\u03bc\u03ad\u03bd\u03b7 \u03bc\u03c0\u03bf\u03c1\u03b5\u03af \u03bd\u03b1 \u03b5\u03bd\u03c4\u03b1\u03c7\u03b8\u03b5\u03af \u03c3\u03b5 \u03bc\u03b9\u03b1 \u03b5\u03c5\u03c1\u03cd\u03c4\u03b5\u03c1\u03b7, \u03bc\u03b5\u03c4\u03c1\u03ae\u03c3\u03b9\u03bc\u03b7 \u03b1\u03b4\u03c5\u03bd\u03b1\u03bc\u03af\u03b1 \u03c4\u03bf\u03c5 \u03c3\u03c5\u03b3\u03ba\u03b5\u03ba\u03c1\u03b9\u03bc\u03ad\u03bd\u03bf\u03c5 \u03bc\u03bf\u03bd\u03c4\u03ad\u03bb\u03bf\u03c5.<\/p>\n<h2 id=\"criterion-perigrafi-ikanotitas\">\u03a0\u03ce\u03c2 \u03ad\u03bd\u03b1 \u03ba\u03c1\u03b9\u03c4\u03ae\u03c1\u03b9\u03bf \u03b3\u03af\u03bd\u03b5\u03c4\u03b1\u03b9 \u03c0\u03b5\u03c1\u03b9\u03b3\u03c1\u03b1\u03c6\u03ae \u03b9\u03ba\u03b1\u03bd\u03cc\u03c4\u03b7\u03c4\u03b1\u03c2<\/h2>\n<p>\u03a4\u03bf \u03c0\u03c1\u03ce\u03c4\u03bf \u03b2\u03ae\u03bc\u03b1 \u03c4\u03bf\u03c5 CRAFT \u03b5\u03af\u03bd\u03b1\u03b9 \u03bd\u03b1 \u00ab\u03b9\u03c3\u03b9\u03ce\u03c3\u03b5\u03b9\u00bb \u03c4\u03bf dataset \u03c3\u03b5 \u03b6\u03b5\u03cd\u03b3\u03b7 prompt\u2013\u03ba\u03c1\u03b9\u03c4\u03b7\u03c1\u03af\u03bf\u03c5. \u0391\u03bd \u03ad\u03bd\u03b1 prompt \u03ad\u03c7\u03b5\u03b9 \u03c4\u03ad\u03c3\u03c3\u03b5\u03c1\u03b1 \u03b4\u03b9\u03b1\u03c6\u03bf\u03c1\u03b5\u03c4\u03b9\u03ba\u03ac \u03ba\u03c1\u03b9\u03c4\u03ae\u03c1\u03b9\u03b1, \u03c0\u03b1\u03c1\u03ac\u03b3\u03b5\u03b9 \u03c4\u03ad\u03c3\u03c3\u03b5\u03c1\u03b9\u03c2 \u03be\u03b5\u03c7\u03c9\u03c1\u03b9\u03c3\u03c4\u03ad\u03c2 \u03bc\u03bf\u03bd\u03ac\u03b4\u03b5\u03c2 \u03b1\u03bd\u03ac\u03bb\u03c5\u03c3\u03b7\u03c2. \u0388\u03bd\u03b1 \u03c0\u03c1\u03cc\u03c3\u03b8\u03b5\u03c4\u03bf \u03b3\u03bb\u03c9\u03c3\u03c3\u03b9\u03ba\u03cc \u03bc\u03bf\u03bd\u03c4\u03ad\u03bb\u03bf \u03b4\u03b9\u03b1\u03b2\u03ac\u03b6\u03b5\u03b9 \u03ba\u03ac\u03b8\u03b5 \u03b6\u03b5\u03cd\u03b3\u03bf\u03c2 \u03ba\u03b1\u03b9 \u03b4\u03b7\u03bc\u03b9\u03bf\u03c5\u03c1\u03b3\u03b5\u03af \u03bc\u03b9\u03b1 \u03c3\u03cd\u03bd\u03c4\u03bf\u03bc\u03b7, \u03b1\u03c0\u03bf\u03c3\u03c5\u03bd\u03b4\u03b5\u03b4\u03b5\u03bc\u03ad\u03bd\u03b7 \u03b1\u03c0\u03cc \u03c4\u03bf \u03c3\u03c5\u03b3\u03ba\u03b5\u03ba\u03c1\u03b9\u03bc\u03ad\u03bd\u03bf \u03c0\u03b1\u03c1\u03ac\u03b4\u03b5\u03b9\u03b3\u03bc\u03b1 \u03c0\u03b5\u03c1\u03b9\u03b3\u03c1\u03b1\u03c6\u03ae \u03c4\u03b7\u03c2 \u03b9\u03ba\u03b1\u03bd\u03cc\u03c4\u03b7\u03c4\u03b1\u03c2 \u03c0\u03bf\u03c5 \u03c7\u03c1\u03b5\u03b9\u03ac\u03b6\u03b5\u03c4\u03b1\u03b9 \u03b3\u03b9\u03b1 \u03bd\u03b1 \u03b9\u03ba\u03b1\u03bd\u03bf\u03c0\u03bf\u03b9\u03b7\u03b8\u03b5\u03af \u03c4\u03bf \u03ba\u03c1\u03b9\u03c4\u03ae\u03c1\u03b9\u03bf.<\/p>\n<p>\u0397 \u03b1\u03c0\u03bf\u03c3\u03cd\u03bd\u03b4\u03b5\u03c3\u03b7 \u03b1\u03c0\u03cc \u03c4\u03bf \u03b1\u03c1\u03c7\u03b9\u03ba\u03cc wording \u03ad\u03c7\u03b5\u03b9 \u03c0\u03c1\u03b1\u03ba\u03c4\u03b9\u03ba\u03ae \u03c3\u03b7\u03bc\u03b1\u03c3\u03af\u03b1. \u0394\u03cd\u03bf \u03ba\u03c1\u03b9\u03c4\u03ae\u03c1\u03b9\u03b1 \u03bc\u03c0\u03bf\u03c1\u03b5\u03af \u03bd\u03b1 \u03b5\u03af\u03bd\u03b1\u03b9 \u03b3\u03c1\u03b1\u03bc\u03bc\u03ad\u03bd\u03b1 \u03c0\u03bf\u03bb\u03cd \u03b4\u03b9\u03b1\u03c6\u03bf\u03c1\u03b5\u03c4\u03b9\u03ba\u03ac \u03ba\u03b1\u03b9 \u03bd\u03b1 \u03c0\u03c1\u03bf\u03ad\u03c1\u03c7\u03bf\u03bd\u03c4\u03b1\u03b9 \u03b1\u03c0\u03cc \u03b4\u03b9\u03b1\u03c6\u03bf\u03c1\u03b5\u03c4\u03b9\u03ba\u03ad\u03c2 \u03b5\u03c1\u03c9\u03c4\u03ae\u03c3\u03b5\u03b9\u03c2, \u03b1\u03bb\u03bb\u03ac \u03bd\u03b1 \u03b5\u03bb\u03ad\u03b3\u03c7\u03bf\u03c5\u03bd \u03c4\u03b7\u03bd \u03af\u03b4\u03b9\u03b1 \u03c5\u03c0\u03bf\u03ba\u03b5\u03af\u03bc\u03b5\u03bd\u03b7 \u03b4\u03b5\u03be\u03b9\u03cc\u03c4\u03b7\u03c4\u03b1. \u03a3\u03c4\u03bf \u03c7\u03c1\u03b7\u03bc\u03b1\u03c4\u03bf\u03bf\u03b9\u03ba\u03bf\u03bd\u03bf\u03bc\u03b9\u03ba\u03cc \u03c0\u03b1\u03c1\u03ac\u03b4\u03b5\u03b9\u03b3\u03bc\u03b1 \u03c4\u03b7\u03c2 \u03b5\u03c1\u03b3\u03b1\u03c3\u03af\u03b1\u03c2, \u03b4\u03b9\u03b1\u03c6\u03bf\u03c1\u03b5\u03c4\u03b9\u03ba\u03ac \u03ba\u03c1\u03b9\u03c4\u03ae\u03c1\u03b9\u03b1 \u03bc\u03c0\u03bf\u03c1\u03bf\u03cd\u03bd \u03bd\u03b1 \u03b1\u03bd\u03c4\u03b9\u03c3\u03c4\u03bf\u03b9\u03c7\u03bf\u03cd\u03bd \u03c3\u03b5 \u03b9\u03ba\u03b1\u03bd\u03cc\u03c4\u03b7\u03c4\u03b5\u03c2 \u03cc\u03c0\u03c9\u03c2 \u03b7 \u03b1\u03bd\u03ac\u03bb\u03c5\u03c3\u03b7 \u03c4\u03b7\u03c2 \u03bc\u03b5\u03c4\u03ac\u03b4\u03bf\u03c3\u03b7\u03c2 \u03bc\u03b9\u03b1\u03c2 \u03b1\u03bb\u03bb\u03b1\u03b3\u03ae\u03c2 \u03b5\u03c0\u03b9\u03c4\u03bf\u03ba\u03af\u03c9\u03bd, \u03b7 \u03c3\u03c4\u03c1\u03b1\u03c4\u03b7\u03b3\u03b9\u03ba\u03ae \u03ba\u03b1\u03c4\u03b1\u03bd\u03bf\u03bc\u03ae \u03ba\u03b5\u03c6\u03b1\u03bb\u03b1\u03b9\u03bf\u03c5\u03c7\u03b9\u03ba\u03ce\u03bd \u03b4\u03b1\u03c0\u03b1\u03bd\u03ce\u03bd \u03ae \u03b7 \u03b4\u03b9\u03b1\u03bc\u03cc\u03c1\u03c6\u03c9\u03c3\u03b7 \u03c0\u03bf\u03bb\u03b9\u03c4\u03b9\u03ba\u03ae\u03c2 \u03c0\u03bb\u03b7\u03c1\u03c9\u03bc\u03ce\u03bd. \u0397 \u03c0\u03b5\u03c1\u03b9\u03b3\u03c1\u03b1\u03c6\u03ae \u03b9\u03ba\u03b1\u03bd\u03cc\u03c4\u03b7\u03c4\u03b1\u03c2 \u03b5\u03c0\u03b9\u03c4\u03c1\u03ad\u03c0\u03b5\u03b9 \u03c3\u03b5 \u03c3\u03c5\u03bd\u03b1\u03c6\u03ae \u03ba\u03c1\u03b9\u03c4\u03ae\u03c1\u03b9\u03b1 \u03bd\u03b1 \u03b2\u03c1\u03b5\u03b8\u03bf\u03cd\u03bd \u03c3\u03c4\u03bf\u03bd \u03af\u03b4\u03b9\u03bf \u03c7\u03ce\u03c1\u03bf, \u03b1\u03ba\u03cc\u03bc\u03b7 \u03ba\u03b9 \u03cc\u03c4\u03b1\u03bd \u03b4\u03b5\u03bd \u03bc\u03bf\u03b9\u03c1\u03ac\u03b6\u03bf\u03bd\u03c4\u03b1\u03b9 \u03c4\u03b9\u03c2 \u03af\u03b4\u03b9\u03b5\u03c2 \u03bb\u03ad\u03be\u03b5\u03b9\u03c2.<\/p>\n<p>\u03a4\u03bf \u03b1\u03c0\u03bf\u03c4\u03ad\u03bb\u03b5\u03c3\u03bc\u03b1 \u03b5\u03af\u03bd\u03b1\u03b9 \u03bc\u03b9\u03b1 \u03c3\u03c5\u03bb\u03bb\u03bf\u03b3\u03ae \u03b1\u03c0\u03cc tuples \u03c0\u03bf\u03c5 \u03b4\u03b9\u03b1\u03c4\u03b7\u03c1\u03bf\u03cd\u03bd \u03c4\u03bf \u03b1\u03c1\u03c7\u03b9\u03ba\u03cc prompt, \u03c4\u03bf \u03ba\u03c1\u03b9\u03c4\u03ae\u03c1\u03b9\u03bf \u03ba\u03b1\u03b9 \u03c4\u03b7\u03bd \u03b5\u03be\u03b1\u03b3\u03cc\u03bc\u03b5\u03bd\u03b7 \u03b9\u03ba\u03b1\u03bd\u03cc\u03c4\u03b7\u03c4\u03b1. \u0388\u03c4\u03c3\u03b9, \u03ba\u03ac\u03b8\u03b5 \u03bc\u03b5\u03bb\u03bb\u03bf\u03bd\u03c4\u03b9\u03ba\u03cc\u03c2 \u03ba\u03cc\u03bc\u03b2\u03bf\u03c2 \u03c4\u03b7\u03c2 \u03b9\u03b5\u03c1\u03b1\u03c1\u03c7\u03af\u03b1\u03c2 \u03c0\u03b1\u03c1\u03b1\u03bc\u03ad\u03bd\u03b5\u03b9 \u03c3\u03c5\u03bd\u03b4\u03b5\u03b4\u03b5\u03bc\u03ad\u03bd\u03bf\u03c2 \u03bc\u03b5 \u03c4\u03b1 \u03c0\u03c1\u03b1\u03b3\u03bc\u03b1\u03c4\u03b9\u03ba\u03ac \u03c0\u03b1\u03c1\u03b1\u03b4\u03b5\u03af\u03b3\u03bc\u03b1\u03c4\u03b1 \u03ba\u03b1\u03b9 \u03c4\u03b9\u03c2 \u03c0\u03c1\u03b1\u03b3\u03bc\u03b1\u03c4\u03b9\u03ba\u03ad\u03c2 \u03b1\u03c0\u03bf\u03c6\u03ac\u03c3\u03b5\u03b9\u03c2 \u03b2\u03b1\u03b8\u03bc\u03bf\u03bb\u03cc\u03b3\u03b7\u03c3\u03b7\u03c2 \u03c0\u03bf\u03c5 \u03c4\u03bf\u03bd \u03c3\u03c4\u03b7\u03c1\u03af\u03b6\u03bf\u03c5\u03bd.<\/p>\n<h2 id=\"ierarchia-ikanotiton-polla-epipeda\">\u0397 \u03b9\u03b5\u03c1\u03b1\u03c1\u03c7\u03af\u03b1 \u03b9\u03ba\u03b1\u03bd\u03bf\u03c4\u03ae\u03c4\u03c9\u03bd \u03ba\u03b1\u03b9 \u03b3\u03b9\u03b1\u03c4\u03af \u03b4\u03b5\u03bd \u03b1\u03c1\u03ba\u03b5\u03af \u03ad\u03bd\u03b1 \u03b5\u03c0\u03af\u03c0\u03b5\u03b4\u03bf<\/h2>\n<p>\u03a3\u03c4\u03b7 \u03c3\u03c5\u03bd\u03ad\u03c7\u03b5\u03b9\u03b1, \u03c4\u03bf CRAFT \u03bf\u03c1\u03b3\u03b1\u03bd\u03ce\u03bd\u03b5\u03b9 \u03c4\u03b9\u03c2 \u03c0\u03b5\u03c1\u03b9\u03b3\u03c1\u03b1\u03c6\u03ad\u03c2 \u03b9\u03ba\u03b1\u03bd\u03bf\u03c4\u03ae\u03c4\u03c9\u03bd \u03c3\u03b5 \u03ad\u03bd\u03b1 \u03b9\u03b5\u03c1\u03b1\u03c1\u03c7\u03b9\u03ba\u03cc \u03b4\u03ad\u03bd\u03c4\u03c1\u03bf. \u03a4\u03b1 \u03c6\u03cd\u03bb\u03bb\u03b1 \u03b1\u03bd\u03c4\u03b9\u03c3\u03c4\u03bf\u03b9\u03c7\u03bf\u03cd\u03bd \u03c3\u03c4\u03b1 \u03bc\u03b5\u03bc\u03bf\u03bd\u03c9\u03bc\u03ad\u03bd\u03b1 \u03b6\u03b5\u03cd\u03b3\u03b7 prompt\u2013\u03ba\u03c1\u03b9\u03c4\u03b7\u03c1\u03af\u03bf\u03c5, \u03b5\u03bd\u03ce \u03bf\u03b9 \u03b5\u03c3\u03c9\u03c4\u03b5\u03c1\u03b9\u03ba\u03bf\u03af \u03ba\u03cc\u03bc\u03b2\u03bf\u03b9 \u03c0\u03b5\u03c1\u03b9\u03b3\u03c1\u03ac\u03c6\u03bf\u03c5\u03bd \u03cc\u03bb\u03bf \u03ba\u03b1\u03b9 \u03b3\u03b5\u03bd\u03b9\u03ba\u03cc\u03c4\u03b5\u03c1\u03b5\u03c2 \u03b9\u03ba\u03b1\u03bd\u03cc\u03c4\u03b7\u03c4\u03b5\u03c2. \u0397 \u03ba\u03b1\u03c4\u03b1\u03c3\u03ba\u03b5\u03c5\u03ae \u03b3\u03af\u03bd\u03b5\u03c4\u03b1\u03b9 \u03b1\u03c0\u03cc \u03ba\u03ac\u03c4\u03c9 \u03c0\u03c1\u03bf\u03c2 \u03c4\u03b1 \u03c0\u03ac\u03bd\u03c9, \u03b5\u03c0\u03af\u03c0\u03b5\u03b4\u03bf \u03c0\u03c1\u03bf\u03c2 \u03b5\u03c0\u03af\u03c0\u03b5\u03b4\u03bf, \u03bc\u03b5 \u03b4\u03b9\u03b1\u03b4\u03b9\u03ba\u03b1\u03c3\u03af\u03b1 \u03c0\u03c1\u03bf\u03c3\u03b1\u03c1\u03bc\u03bf\u03c3\u03bc\u03ad\u03bd\u03b7 \u03b1\u03c0\u03cc \u03c4\u03bf\u03bd hierarchizer \u03c4\u03bf\u03c5 Clio.<\/p>\n<p>\u038c\u03c4\u03b1\u03bd \u03c4\u03bf \u03b1\u03c1\u03c7\u03b9\u03ba\u03cc dataset \u03c0\u03b5\u03c1\u03b9\u03bb\u03b1\u03bc\u03b2\u03ac\u03bd\u03b5\u03b9 \u03ae\u03b4\u03b7 \u03ba\u03b1\u03c4\u03b7\u03b3\u03bf\u03c1\u03b9\u03ba\u03ac metadata, \u03b1\u03c5\u03c4\u03ac \u03b4\u03b9\u03b1\u03c4\u03b7\u03c1\u03bf\u03cd\u03bd\u03c4\u03b1\u03b9 \u03c3\u03c4\u03b1 \u03b1\u03bd\u03ce\u03c4\u03b5\u03c1\u03b1 \u03b5\u03c0\u03af\u03c0\u03b5\u03b4\u03b1. \u03a3\u03c4\u03b1 \u03c0\u03b5\u03b9\u03c1\u03ac\u03bc\u03b1\u03c4\u03b1 \u03c4\u03b7\u03c2 \u03b5\u03c1\u03b3\u03b1\u03c3\u03af\u03b1\u03c2, \u03c4\u03bf domain \u03ba\u03b1\u03b9 \u03b7 \u03c5\u03c0\u03bf\u03ba\u03b1\u03c4\u03b7\u03b3\u03bf\u03c1\u03af\u03b1 \u03c4\u03b7\u03c2 rubric \u03b1\u03c0\u03bf\u03c4\u03ad\u03bb\u03b5\u03c3\u03b1\u03bd \u03c4\u03b1 \u03b4\u03cd\u03bf \u03c3\u03c4\u03b1\u03b8\u03b5\u03c1\u03ac \u03b1\u03bd\u03ce\u03c4\u03b5\u03c1\u03b1 \u03b5\u03c0\u03af\u03c0\u03b5\u03b4\u03b1. \u039a\u03ac\u03c4\u03c9 \u03b1\u03c0\u03cc \u03b1\u03c5\u03c4\u03ac, \u03bf\u03b9 \u03b5\u03c1\u03b5\u03c5\u03bd\u03b7\u03c4\u03ad\u03c2 \u03b4\u03b7\u03bc\u03b9\u03bf\u03cd\u03c1\u03b3\u03b7\u03c3\u03b1\u03bd learned subtrees. \u0388\u03c4\u03c3\u03b9, \u03bf\u03b9 \u03b1\u03c5\u03c4\u03cc\u03bc\u03b1\u03c4\u03b1 \u03c0\u03b1\u03c1\u03b1\u03b3\u03cc\u03bc\u03b5\u03bd\u03b5\u03c2 \u03bf\u03bc\u03ac\u03b4\u03b5\u03c2 \u03b4\u03b5\u03bd \u03b1\u03bd\u03b1\u03bc\u03b5\u03b9\u03b3\u03bd\u03cd\u03bf\u03c5\u03bd \u03ba\u03b1\u03c4\u03b7\u03b3\u03bf\u03c1\u03af\u03b5\u03c2 \u03c0\u03bf\u03c5 \u03bf\u03b9 \u03b4\u03b7\u03bc\u03b9\u03bf\u03c5\u03c1\u03b3\u03bf\u03af \u03c4\u03bf\u03c5 dataset \u03b5\u03af\u03c7\u03b1\u03bd \u03c3\u03ba\u03cc\u03c0\u03b9\u03bc\u03b1 \u03b4\u03b9\u03b1\u03c7\u03c9\u03c1\u03af\u03c3\u03b5\u03b9.<\/p>\n<p>\u03a3\u03c4\u03bf \u03bb\u03b5\u03c0\u03c4\u03cc\u03c4\u03b5\u03c1\u03bf learned \u03b5\u03c0\u03af\u03c0\u03b5\u03b4\u03bf, \u03bf\u03b9 \u03c0\u03b5\u03c1\u03b9\u03b3\u03c1\u03b1\u03c6\u03ad\u03c2 \u03bc\u03b5\u03c4\u03b1\u03c4\u03c1\u03ad\u03c0\u03bf\u03bd\u03c4\u03b1\u03b9 \u03c3\u03b5 embeddings \u03ba\u03b1\u03b9 \u03c7\u03c9\u03c1\u03af\u03b6\u03bf\u03bd\u03c4\u03b1\u03b9 \u03bc\u03b5 k-means. \u0388\u03c0\u03b5\u03b9\u03c4\u03b1, \u03ad\u03bd\u03b1 LLM \u03bf\u03bd\u03bf\u03bc\u03ac\u03b6\u03b5\u03b9 \u03ba\u03ac\u03b8\u03b5 cluster \u03b1\u03c0\u03cc \u03b4\u03b5\u03af\u03b3\u03bc\u03b1 \u03c4\u03c9\u03bd \u03bc\u03b5\u03bb\u03ce\u03bd \u03c4\u03bf\u03c5. \u0393\u03b9\u03b1 \u03c4\u03b7\u03bd \u03ba\u03b1\u03c4\u03b1\u03c3\u03ba\u03b5\u03c5\u03ae \u03c4\u03c9\u03bd \u03b1\u03bd\u03ce\u03c4\u03b5\u03c1\u03c9\u03bd \u03b5\u03c0\u03b9\u03c0\u03ad\u03b4\u03c9\u03bd, \u03bf\u03b9 child clusters \u03bf\u03bc\u03b1\u03b4\u03bf\u03c0\u03bf\u03b9\u03bf\u03cd\u03bd\u03c4\u03b1\u03b9 \u03c3\u03b5 \u03b3\u03b5\u03b9\u03c4\u03bf\u03bd\u03b9\u03ad\u03c2, \u03b4\u03b7\u03bc\u03b9\u03bf\u03c5\u03c1\u03b3\u03bf\u03cd\u03bd\u03c4\u03b1\u03b9 \u03c5\u03c0\u03bf\u03c8\u03ae\u03c6\u03b9\u03b5\u03c2 parent capabilities, \u03c3\u03c5\u03b3\u03c7\u03c9\u03bd\u03b5\u03cd\u03bf\u03bd\u03c4\u03b1\u03b9 \u03b5\u03c0\u03b9\u03ba\u03b1\u03bb\u03cd\u03c8\u03b5\u03b9\u03c2, \u03ba\u03ac\u03b8\u03b5 \u03c0\u03b1\u03b9\u03b4\u03af \u03b4\u03c1\u03bf\u03bc\u03bf\u03bb\u03bf\u03b3\u03b5\u03af\u03c4\u03b1\u03b9 \u03c3\u03c4\u03bf\u03bd \u03ba\u03b1\u03c4\u03b1\u03bb\u03bb\u03b7\u03bb\u03cc\u03c4\u03b5\u03c1\u03bf \u03b3\u03bf\u03bd\u03ad\u03b1 \u03ba\u03b1\u03b9 \u03b7 \u03c4\u03b5\u03bb\u03b9\u03ba\u03ae \u03c0\u03b5\u03c1\u03b9\u03b3\u03c1\u03b1\u03c6\u03ae \u03c4\u03bf\u03c5 \u03b3\u03bf\u03bd\u03ad\u03b1 \u03b5\u03c0\u03b1\u03bd\u03b1\u03b3\u03c1\u03ac\u03c6\u03b5\u03c4\u03b1\u03b9 \u03bc\u03b5 \u03b2\u03ac\u03c3\u03b7 \u03c4\u03b1 \u03c0\u03b1\u03b9\u03b4\u03b9\u03ac \u03c0\u03bf\u03c5 \u03c0\u03c1\u03ac\u03b3\u03bc\u03b1\u03c4\u03b9 \u03ad\u03bb\u03b1\u03b2\u03b5.<\/p>\n<p>\u0397 \u03b5\u03c1\u03b3\u03b1\u03c3\u03af\u03b1 \u03c0\u03b5\u03c1\u03b9\u03b3\u03c1\u03ac\u03c6\u03b5\u03b9 \u03ba\u03b1\u03b9 \u03bc\u03b7\u03c7\u03b1\u03bd\u03b9\u03c3\u03bc\u03cc \u03b5\u03bb\u03ad\u03b3\u03c7\u03bf\u03c5 \u03c4\u03b7\u03c2 \u03c3\u03c4\u03b1\u03b8\u03b5\u03c1\u03cc\u03c4\u03b7\u03c4\u03b1\u03c2 \u03c4\u03b7\u03c2 \u03b4\u03c1\u03bf\u03bc\u03bf\u03bb\u03cc\u03b3\u03b7\u03c3\u03b7\u03c2. \u0388\u03bd\u03b1 \u03b1\u03bd\u03b5\u03be\u03ac\u03c1\u03c4\u03b7\u03c4\u03bf \u03bc\u03bf\u03bd\u03c4\u03ad\u03bb\u03bf \u03b5\u03c0\u03b9\u03c7\u03b5\u03b9\u03c1\u03b5\u03af \u03b5\u03c0\u03b1\u03bd\u03b5\u03b9\u03bb\u03b7\u03bc\u03bc\u03ad\u03bd\u03b1 \u03bd\u03b1 \u03c4\u03bf\u03c0\u03bf\u03b8\u03b5\u03c4\u03ae\u03c3\u03b5\u03b9 \u03b4\u03b5\u03af\u03b3\u03bc\u03b1\u03c4\u03b1 child clusters \u03c3\u03c4\u03bf\u03c5\u03c2 \u03c5\u03c0\u03bf\u03c8\u03ae\u03c6\u03b9\u03bf\u03c5\u03c2 \u03b3\u03bf\u03bd\u03b5\u03af\u03c2. \u038c\u03c4\u03b1\u03bd \u03ad\u03bd\u03b1 \u03c0\u03b1\u03b9\u03b4\u03af \u03b1\u03bb\u03bb\u03ac\u03b6\u03b5\u03b9 \u03c3\u03c5\u03c7\u03bd\u03ac \u03b3\u03bf\u03bd\u03ad\u03b1, \u03b1\u03c5\u03c4\u03cc \u03b1\u03c0\u03bf\u03c4\u03b5\u03bb\u03b5\u03af \u03ad\u03bd\u03b4\u03b5\u03b9\u03be\u03b7 \u03cc\u03c4\u03b9 \u03bf\u03b9 parent descriptions \u03b5\u03c0\u03b9\u03ba\u03b1\u03bb\u03cd\u03c0\u03c4\u03bf\u03bd\u03c4\u03b1\u03b9 \u03ba\u03b1\u03b9 \u03c0\u03c1\u03ad\u03c0\u03b5\u03b9 \u03bd\u03b1 \u03b1\u03bd\u03b1\u03b8\u03b5\u03c9\u03c1\u03b7\u03b8\u03bf\u03cd\u03bd. \u0395\u03c0\u03b9\u03c0\u03bb\u03ad\u03bf\u03bd, \u03b7 \u03c3\u03b5\u03b9\u03c1\u03ac \u03c4\u03c9\u03bd \u03b3\u03bf\u03bd\u03ad\u03c9\u03bd \u03b1\u03bd\u03b1\u03ba\u03b1\u03c4\u03b5\u03cd\u03b5\u03c4\u03b1\u03b9 \u03b3\u03b9\u03b1 \u03bd\u03b1 \u03bc\u03b5\u03b9\u03c9\u03b8\u03b5\u03af \u03c4\u03bf position bias.<\/p>\n<p>\u03a4\u03bf \u03b2\u03ac\u03b8\u03bf\u03c2 \u03c4\u03bf\u03c5 \u03b4\u03ad\u03bd\u03c4\u03c1\u03bf\u03c5 \u03b4\u03b5\u03bd \u03bf\u03c1\u03af\u03b6\u03b5\u03c4\u03b1\u03b9 \u03b1\u03c5\u03b8\u03b1\u03af\u03c1\u03b5\u03c4\u03b1 \u03b5\u03ba \u03c4\u03c9\u03bd \u03c0\u03c1\u03bf\u03c4\u03ad\u03c1\u03c9\u03bd. \u03a0\u03c1\u03bf\u03c3\u03b1\u03c1\u03bc\u03cc\u03b6\u03b5\u03c4\u03b1\u03b9 \u03c3\u03c4\u03bf \u03bc\u03ad\u03b3\u03b5\u03b8\u03bf\u03c2 \u03c4\u03b7\u03c2 \u03bf\u03bc\u03ac\u03b4\u03b1\u03c2, \u03ce\u03c3\u03c4\u03b5 \u03bc\u03b5\u03b3\u03b1\u03bb\u03cd\u03c4\u03b5\u03c1\u03b5\u03c2 \u03c3\u03c5\u03bb\u03bb\u03bf\u03b3\u03ad\u03c2 \u03bd\u03b1 \u03b1\u03c0\u03bf\u03ba\u03c4\u03bf\u03cd\u03bd \u03bb\u03b5\u03c0\u03c4\u03cc\u03c4\u03b5\u03c1\u03b5\u03c2 \u03ba\u03b1\u03b9 \u03b2\u03b1\u03b8\u03cd\u03c4\u03b5\u03c1\u03b5\u03c2 \u03b9\u03b5\u03c1\u03b1\u03c1\u03c7\u03af\u03b5\u03c2, \u03b5\u03bd\u03ce \u03bc\u03b9\u03ba\u03c1\u03cc\u03c4\u03b5\u03c1\u03b5\u03c2 \u03bd\u03b1 \u03bc\u03ad\u03bd\u03bf\u03c5\u03bd \u03c0\u03b9\u03bf \u03c1\u03b7\u03c7\u03ad\u03c2. \u03a3\u03c4\u03b1 \u03c0\u03b5\u03b9\u03c1\u03ac\u03bc\u03b1\u03c4\u03b1, \u03b7 \u03c4\u03b5\u03bb\u03b9\u03ba\u03ae \u03b4\u03bf\u03bc\u03ae \u03b5\u03af\u03c7\u03b5 \u03ad\u03be\u03b9 \u03b5\u03c0\u03af\u03c0\u03b5\u03b4\u03b1 \u03ba\u03b1\u03b9 \u03ba\u03ac\u03bb\u03c5\u03c8\u03b5 20.443 \u03b6\u03b5\u03cd\u03b3\u03b7 prompt\u2013rubric. \u0393\u03b9\u03b1 \u03c4\u03b9\u03c2 \u03b4\u03b9\u03b1\u03b4\u03b9\u03ba\u03b1\u03c3\u03af\u03b5\u03c2 \u03bf\u03bd\u03bf\u03bc\u03b1\u03c3\u03af\u03b1\u03c2, \u03b5\u03bd\u03bf\u03c0\u03bf\u03af\u03b7\u03c3\u03b7\u03c2, routing \u03ba\u03b1\u03b9 refinement \u03c7\u03c1\u03b7\u03c3\u03b9\u03bc\u03bf\u03c0\u03bf\u03b9\u03ae\u03b8\u03b7\u03ba\u03b5 Gemini 2.5 Pro, \u03b3\u03b9\u03b1 \u03c4\u03bf\u03c5\u03c2 \u03b5\u03bb\u03ad\u03b3\u03c7\u03bf\u03c5\u03c2 \u03c3\u03c4\u03b1\u03b8\u03b5\u03c1\u03cc\u03c4\u03b7\u03c4\u03b1\u03c2 Gemini 2.5 Flash \u03ba\u03b1\u03b9 \u03b3\u03b9\u03b1 \u03c4\u03b1 embeddings \u03c4\u03bf text-embedding-3-small.<\/p>\n<h2 id=\"capability-tree-adynamoi-komvoi\">\u0391\u03c0\u03cc \u03c4\u03bf capability tree \u03c3\u03c4\u03bf\u03c5\u03c2 \u03b1\u03b4\u03cd\u03bd\u03b1\u03bc\u03bf\u03c5\u03c2 \u03ba\u03cc\u03bc\u03b2\u03bf\u03c5\u03c2<\/h2>\n<p>\u03a4\u03bf \u03b4\u03ad\u03bd\u03c4\u03c1\u03bf \u03b1\u03c0\u03cc \u03bc\u03cc\u03bd\u03bf \u03c4\u03bf\u03c5 \u03c0\u03b5\u03c1\u03b9\u03b3\u03c1\u03ac\u03c6\u03b5\u03b9 \u03c4\u03b9\u03c2 \u03b9\u03ba\u03b1\u03bd\u03cc\u03c4\u03b7\u03c4\u03b5\u03c2 \u03c4\u03bf\u03c5 dataset, \u03cc\u03c7\u03b9 \u03c4\u03b9\u03c2 \u03b1\u03b4\u03c5\u03bd\u03b1\u03bc\u03af\u03b5\u03c2 \u03b5\u03bd\u03cc\u03c2 \u03c3\u03c5\u03b3\u03ba\u03b5\u03ba\u03c1\u03b9\u03bc\u03ad\u03bd\u03bf\u03c5 \u03bc\u03bf\u03bd\u03c4\u03ad\u03bb\u03bf\u03c5. \u0393\u03b9\u03b1 \u03bd\u03b1 \u03b3\u03af\u03bd\u03b5\u03b9 model-specific, \u03c4\u03bf target model \u03b1\u03c0\u03b1\u03bd\u03c4\u03ac \u03c3\u03b5 \u03ba\u03ac\u03b8\u03b5 prompt \u03ba\u03b1\u03b9 \u03ad\u03bd\u03b1 LLM-as-a-judge \u03b5\u03bb\u03ad\u03b3\u03c7\u03b5\u03b9 \u03ba\u03ac\u03b8\u03b5 \u03c3\u03c7\u03b5\u03c4\u03b9\u03ba\u03cc \u03ba\u03c1\u03b9\u03c4\u03ae\u03c1\u03b9\u03bf \u03c9\u03c2 pass \u03ae fail. \u039f\u03b9 \u03b1\u03c0\u03bf\u03c6\u03ac\u03c3\u03b5\u03b9\u03c2 \u03b1\u03c5\u03c4\u03ad\u03c2 \u03c3\u03c5\u03bd\u03b4\u03ad\u03bf\u03bd\u03c4\u03b1\u03b9 \u03bc\u03b5 \u03c4\u03b1 \u03c6\u03cd\u03bb\u03bb\u03b1 \u03ba\u03b1\u03b9 \u03c3\u03c4\u03b7 \u03c3\u03c5\u03bd\u03ad\u03c7\u03b5\u03b9\u03b1 \u03b1\u03b8\u03c1\u03bf\u03af\u03b6\u03bf\u03bd\u03c4\u03b1\u03b9 \u03c0\u03c1\u03bf\u03c2 \u03c4\u03b1 \u03c0\u03ac\u03bd\u03c9. \u039a\u03ac\u03b8\u03b5 \u03b5\u03c3\u03c9\u03c4\u03b5\u03c1\u03b9\u03ba\u03cc\u03c2 \u03ba\u03cc\u03bc\u03b2\u03bf\u03c2 \u03b1\u03c0\u03bf\u03ba\u03c4\u03ac \u03ad\u03bd\u03b1 pass rate, \u03b4\u03b7\u03bb\u03b1\u03b4\u03ae \u03c4\u03bf\u03bd \u03bc\u03ad\u03c3\u03bf \u03c1\u03c5\u03b8\u03bc\u03cc \u03b5\u03c0\u03b9\u03c4\u03c5\u03c7\u03af\u03b1\u03c2 \u03c4\u03c9\u03bd \u03ba\u03c1\u03b9\u03c4\u03b7\u03c1\u03af\u03c9\u03bd \u03c0\u03bf\u03c5 \u03c0\u03b5\u03c1\u03b9\u03ad\u03c7\u03b5\u03b9.<\/p>\n<p>\u0391\u03c5\u03c4\u03cc \u03c3\u03b7\u03bc\u03b1\u03af\u03bd\u03b5\u03b9 \u03cc\u03c4\u03b9 \u03ba\u03ac\u03b8\u03b5 \u03ba\u03cc\u03bc\u03b2\u03bf\u03c2 \u03ad\u03c7\u03b5\u03b9 \u03b4\u03cd\u03bf \u03cc\u03c8\u03b5\u03b9\u03c2: \u03bc\u03b9\u03b1 \u03c3\u03b7\u03bc\u03b1\u03c3\u03b9\u03bf\u03bb\u03bf\u03b3\u03b9\u03ba\u03ae \u03c0\u03b5\u03c1\u03b9\u03b3\u03c1\u03b1\u03c6\u03ae \u03ba\u03b1\u03b9 \u03bc\u03b9\u03b1 \u03b5\u03bc\u03c0\u03b5\u03b9\u03c1\u03b9\u03ba\u03ae \u03b5\u03ba\u03c4\u03af\u03bc\u03b7\u03c3\u03b7 \u03c4\u03bf\u03c5 \u03c0\u03cc\u03c3\u03bf \u03b1\u03be\u03b9\u03cc\u03c0\u03b9\u03c3\u03c4\u03b1 \u03c4\u03bf \u03bc\u03bf\u03bd\u03c4\u03ad\u03bb\u03bf \u03b9\u03ba\u03b1\u03bd\u03bf\u03c0\u03bf\u03b9\u03b5\u03af \u03c4\u03b7 \u03c3\u03c5\u03b3\u03ba\u03b5\u03ba\u03c1\u03b9\u03bc\u03ad\u03bd\u03b7 \u03b9\u03ba\u03b1\u03bd\u03cc\u03c4\u03b7\u03c4\u03b1. \u0397 \u03c0\u03bf\u03b9\u03cc\u03c4\u03b7\u03c4\u03b1 \u03c4\u03b7\u03c2 \u03b4\u03b9\u03ac\u03b3\u03bd\u03c9\u03c3\u03b7\u03c2 \u03b5\u03be\u03b1\u03c1\u03c4\u03ac\u03c4\u03b1\u03b9 \u03b2\u03ad\u03b2\u03b1\u03b9\u03b1 \u03b1\u03c0\u03cc \u03c4\u03b7 \u03c3\u03c5\u03bd\u03ad\u03c0\u03b5\u03b9\u03b1 \u03c4\u03bf\u03c5 judge. \u0397 \u03b1\u03bd\u03ac\u03bb\u03c5\u03c3\u03b7 \u03b1\u03bd\u03ac \u03b9\u03ba\u03b1\u03bd\u03cc\u03c4\u03b7\u03c4\u03b1 \u03b2\u03bf\u03b7\u03b8\u03ac \u03bd\u03b1 \u03be\u03b5\u03c7\u03c9\u03c1\u03af\u03c3\u03b5\u03b9 \u03b1\u03bd \u03b7 <a href=\"https:\/\/twodots.gr\/spark-techniti-noimosyni-xerei-apantisi-den-energopoiei-sosti-skepsi\/\">AI \u03b3\u03bd\u03c9\u03c1\u03af\u03b6\u03b5\u03b9 \u03c4\u03b7\u03bd \u03b1\u03c0\u03ac\u03bd\u03c4\u03b7\u03c3\u03b7 \u03b1\u03bb\u03bb\u03ac \u03b4\u03b5\u03bd \u03b5\u03bd\u03b5\u03c1\u03b3\u03bf\u03c0\u03bf\u03b9\u03b5\u03af \u03c4\u03b7 \u03c3\u03c9\u03c3\u03c4\u03ae \u03c3\u03c5\u03bb\u03bb\u03bf\u03b3\u03b9\u03c3\u03c4\u03b9\u03ba\u03ae<\/a> \u03b1\u03c0\u03cc \u03bc\u03b9\u03b1 \u03c0\u03c1\u03b1\u03b3\u03bc\u03b1\u03c4\u03b9\u03ba\u03ae \u03ad\u03bb\u03bb\u03b5\u03b9\u03c8\u03b7 \u03b3\u03bd\u03ce\u03c3\u03b7\u03c2. \u0391\u03bd \u03bf evaluator \u03b5\u03c6\u03b1\u03c1\u03bc\u03cc\u03b6\u03b5\u03b9 \u03c4\u03b1 \u03ba\u03c1\u03b9\u03c4\u03ae\u03c1\u03b9\u03b1 \u03b1\u03c3\u03c4\u03b1\u03b8\u03ce\u03c2, \u03bf\u03b9 \u03b2\u03b1\u03b8\u03bc\u03bf\u03bb\u03bf\u03b3\u03af\u03b5\u03c2 \u03cc\u03bb\u03c9\u03bd \u03c4\u03c9\u03bd \u03b1\u03bd\u03ce\u03c4\u03b5\u03c1\u03c9\u03bd \u03ba\u03cc\u03bc\u03b2\u03c9\u03bd \u03ba\u03bb\u03b7\u03c1\u03bf\u03bd\u03bf\u03bc\u03bf\u03cd\u03bd \u03b1\u03c5\u03c4\u03ae \u03c4\u03b7\u03bd \u03b1\u03c3\u03c4\u03ac\u03b8\u03b5\u03b9\u03b1.<\/p>\n<p>\u03a4\u03bf CRAFT \u03b4\u03b5\u03bd \u03b5\u03c0\u03b9\u03bb\u03ad\u03b3\u03b5\u03b9 \u03cc\u03bb\u03b5\u03c2 \u03c4\u03b9\u03c2 \u03b1\u03b4\u03c5\u03bd\u03b1\u03bc\u03af\u03b5\u03c2 \u03b1\u03c0\u03cc \u03ad\u03bd\u03b1 \u03c3\u03c4\u03b1\u03b8\u03b5\u03c1\u03cc \u03b5\u03c0\u03af\u03c0\u03b5\u03b4\u03bf. \u039e\u03b5\u03ba\u03b9\u03bd\u03ac \u03b1\u03c0\u03cc \u03c0\u03ac\u03bd\u03c9 \u03ba\u03b1\u03b9 \u03b1\u03bd\u03b1\u03b6\u03b7\u03c4\u03ac \u03c7\u03b1\u03bc\u03b7\u03bb\u03ad\u03c2 \u03b5\u03c0\u03b9\u03b4\u03cc\u03c3\u03b5\u03b9\u03c2 \u03c3\u03b5 \u03b4\u03b9\u03b1\u03c6\u03bf\u03c1\u03b5\u03c4\u03b9\u03ba\u03ac \u03b2\u03ac\u03b8\u03b7. \u0388\u03bd\u03b1 \u03b9\u03c3\u03c7\u03c5\u03c1\u03cc\u03c4\u03b5\u03c1\u03bf \u03bc\u03bf\u03bd\u03c4\u03ad\u03bb\u03bf \u03bc\u03c0\u03bf\u03c1\u03b5\u03af \u03bd\u03b1 \u03b1\u03c0\u03bf\u03c4\u03c5\u03b3\u03c7\u03ac\u03bd\u03b5\u03b9 \u03bc\u03cc\u03bd\u03bf \u03c3\u03b5 \u03bc\u03b9\u03b1 \u03c3\u03c4\u03b5\u03bd\u03ae, \u03b5\u03be\u03b5\u03b9\u03b4\u03b9\u03ba\u03b5\u03c5\u03bc\u03ad\u03bd\u03b7 \u03b9\u03ba\u03b1\u03bd\u03cc\u03c4\u03b7\u03c4\u03b1. \u0388\u03bd\u03b1 \u03b1\u03c3\u03b8\u03b5\u03bd\u03ad\u03c3\u03c4\u03b5\u03c1\u03bf \u03bc\u03bf\u03bd\u03c4\u03ad\u03bb\u03bf \u03bc\u03c0\u03bf\u03c1\u03b5\u03af \u03bd\u03b1 \u03ad\u03c7\u03b5\u03b9 \u03c0\u03c1\u03cc\u03b2\u03bb\u03b7\u03bc\u03b1 \u03c3\u03b5 \u03ad\u03bd\u03b1\u03bd \u03b5\u03c5\u03c1\u03cd\u03c4\u03b5\u03c1\u03bf \u03b3\u03bf\u03bd\u03b9\u03ba\u03cc \u03ba\u03cc\u03bc\u03b2\u03bf. \u0391\u03bd \u03b7 \u03b5\u03c0\u03b9\u03bb\u03bf\u03b3\u03ae \u03b3\u03b9\u03bd\u03cc\u03c4\u03b1\u03bd \u03c0\u03ac\u03bd\u03c4\u03b1 \u03c3\u03c4\u03bf \u03bb\u03b5\u03c0\u03c4\u03cc\u03c4\u03b5\u03c1\u03bf \u03b5\u03c0\u03af\u03c0\u03b5\u03b4\u03bf, \u03bc\u03b9\u03b1 \u03b3\u03b5\u03bd\u03b9\u03ba\u03ae \u03b1\u03b4\u03c5\u03bd\u03b1\u03bc\u03af\u03b1 \u03b8\u03b1 \u03ba\u03b1\u03c4\u03b1\u03ba\u03b5\u03c1\u03bc\u03b1\u03c4\u03b9\u03b6\u03cc\u03c4\u03b1\u03bd. \u0391\u03bd \u03b3\u03b9\u03bd\u03cc\u03c4\u03b1\u03bd \u03c0\u03ac\u03bd\u03c4\u03b1 \u03c3\u03b5 \u03c5\u03c8\u03b7\u03bb\u03cc\u03c4\u03b5\u03c1\u03bf \u03b5\u03c0\u03af\u03c0\u03b5\u03b4\u03bf, \u03bc\u03b9\u03b1 \u03b5\u03b9\u03b4\u03b9\u03ba\u03ae \u03b1\u03b4\u03c5\u03bd\u03b1\u03bc\u03af\u03b1 \u03b8\u03b1 \u03c7\u03b1\u03bd\u03cc\u03c4\u03b1\u03bd \u03bc\u03ad\u03c3\u03b1 \u03c3\u03b5 \u03ad\u03bd\u03b1\u03bd \u03c5\u03c0\u03b5\u03c1\u03b2\u03bf\u03bb\u03b9\u03ba\u03ac \u03b3\u03b5\u03bd\u03b9\u03ba\u03cc \u03bc\u03ad\u03c3\u03bf \u03cc\u03c1\u03bf.<\/p>\n<p>\u03a3\u03c4\u03b1 \u03ba\u03cd\u03c1\u03b9\u03b1 \u03c0\u03b5\u03b9\u03c1\u03ac\u03bc\u03b1\u03c4\u03b1, \u03ad\u03bd\u03b1\u03c2 \u03ba\u03cc\u03bc\u03b2\u03bf\u03c2 \u03ae\u03c4\u03b1\u03bd \u03b5\u03c0\u03b9\u03bb\u03ad\u03be\u03b9\u03bc\u03bf\u03c2 \u03bc\u03cc\u03bd\u03bf \u03cc\u03c4\u03b1\u03bd \u03c5\u03c0\u03bf\u03c3\u03c4\u03b7\u03c1\u03b9\u03b6\u03cc\u03c4\u03b1\u03bd \u03b1\u03c0\u03cc \u03c0\u03b5\u03c1\u03b9\u03c3\u03c3\u03cc\u03c4\u03b5\u03c1\u03b1 \u03b1\u03c0\u03cc 30 \u03ba\u03c1\u03b9\u03c4\u03ae\u03c1\u03b9\u03b1. \u0397 \u03b1\u03bd\u03b1\u03b6\u03ae\u03c4\u03b7\u03c3\u03b7 \u03be\u03b5\u03ba\u03b9\u03bd\u03bf\u03cd\u03c3\u03b5 \u03bc\u03b5 \u03b1\u03c5\u03c3\u03c4\u03b7\u03c1\u03cc threshold \u03c3\u03c4\u03bf pass rate \u03ba\u03b1\u03b9 \u03c4\u03bf \u03c7\u03b1\u03bb\u03ac\u03c1\u03c9\u03bd\u03b5 \u03c3\u03c4\u03b1\u03b4\u03b9\u03b1\u03ba\u03ac \u03bc\u03ad\u03c7\u03c1\u03b9 \u03bd\u03b1 \u03c3\u03c5\u03b3\u03ba\u03b5\u03bd\u03c4\u03c1\u03c9\u03b8\u03b5\u03af \u03bf \u03b5\u03c0\u03b9\u03b8\u03c5\u03bc\u03b7\u03c4\u03cc\u03c2 \u03b1\u03c1\u03b9\u03b8\u03bc\u03cc\u03c2 \u03b1\u03b4\u03cd\u03bd\u03b1\u03bc\u03c9\u03bd \u03ba\u03cc\u03bc\u03b2\u03c9\u03bd. \u039c\u03cc\u03bb\u03b9\u03c2 \u03b5\u03c0\u03b9\u03bb\u03b5\u03b3\u03cc\u03c4\u03b1\u03bd \u03ad\u03bd\u03b1\u03c2 \u03ba\u03cc\u03bc\u03b2\u03bf\u03c2, \u03b1\u03c0\u03bf\u03ba\u03bb\u03b5\u03af\u03bf\u03bd\u03c4\u03b1\u03bd \u03bf\u03b9 \u03c0\u03c1\u03cc\u03b3\u03bf\u03bd\u03bf\u03b9 \u03ba\u03b1\u03b9 \u03bf\u03b9 \u03b1\u03c0\u03cc\u03b3\u03bf\u03bd\u03bf\u03af \u03c4\u03bf\u03c5, \u03ce\u03c3\u03c4\u03b5 \u03bf\u03b9 \u03c4\u03b5\u03bb\u03b9\u03ba\u03ad\u03c2 \u03c0\u03b5\u03c1\u03b9\u03bf\u03c7\u03ad\u03c2 \u03b1\u03b4\u03c5\u03bd\u03b1\u03bc\u03af\u03b1\u03c2 \u03bd\u03b1 \u03bc\u03b7\u03bd \u03b5\u03c0\u03b9\u03ba\u03b1\u03bb\u03cd\u03c0\u03c4\u03bf\u03bd\u03c4\u03b1\u03b9.<\/p>\n<h2 id=\"axiologisi-brief-training-data\">\u0397 \u03b1\u03be\u03b9\u03bf\u03bb\u03cc\u03b3\u03b7\u03c3\u03b7 \u03b3\u03af\u03bd\u03b5\u03c4\u03b1\u03b9 brief \u03b3\u03b9\u03b1 \u03c4\u03b7 \u03b4\u03b7\u03bc\u03b9\u03bf\u03c5\u03c1\u03b3\u03af\u03b1 training data<\/h2>\n<p>\u039c\u03b5\u03c4\u03ac \u03c4\u03b7 \u03b4\u03b9\u03ac\u03b3\u03bd\u03c9\u03c3\u03b7, \u03ba\u03ac\u03b8\u03b5 \u03b5\u03c0\u03b9\u03bb\u03b5\u03b3\u03bc\u03ad\u03bd\u03bf\u03c2 \u03ba\u03cc\u03bc\u03b2\u03bf\u03c2 \u03bc\u03b5\u03c4\u03b1\u03c4\u03c1\u03ad\u03c0\u03b5\u03c4\u03b1\u03b9 \u03c3\u03b5 \u03c3\u03c4\u03cc\u03c7\u03bf \u03b4\u03b7\u03bc\u03b9\u03bf\u03c5\u03c1\u03b3\u03af\u03b1\u03c2 supervised fine-tuning \u03b4\u03b5\u03b4\u03bf\u03bc\u03ad\u03bd\u03c9\u03bd. \u0397 \u03bc\u03ad\u03b8\u03bf\u03b4\u03bf\u03c2 \u03b4\u03b5\u03bd \u03b5\u03c0\u03b9\u03b2\u03ac\u03bb\u03bb\u03b5\u03b9 \u03cc\u03c4\u03b9 \u03b1\u03c5\u03c4\u03ac \u03c0\u03c1\u03ad\u03c0\u03b5\u03b9 \u03bd\u03b1 \u03b5\u03af\u03bd\u03b1\u03b9 \u03c3\u03c5\u03bd\u03b8\u03b5\u03c4\u03b9\u03ba\u03ac\u00b7 \u03b8\u03b1 \u03bc\u03c0\u03bf\u03c1\u03bf\u03cd\u03c3\u03b1\u03bd \u03bd\u03b1 \u03b3\u03c1\u03b1\u03c6\u03c4\u03bf\u03cd\u03bd \u03ba\u03b1\u03b9 \u03b1\u03c0\u03cc \u03b1\u03bd\u03b8\u03c1\u03ce\u03c0\u03bf\u03c5\u03c2. \u03a3\u03c4\u03b1 \u03c3\u03c5\u03b3\u03ba\u03b5\u03ba\u03c1\u03b9\u03bc\u03ad\u03bd\u03b1 \u03c0\u03b5\u03b9\u03c1\u03ac\u03bc\u03b1\u03c4\u03b1 \u03c7\u03c1\u03b7\u03c3\u03b9\u03bc\u03bf\u03c0\u03bf\u03b9\u03ae\u03b8\u03b7\u03ba\u03b5 \u03ba\u03bf\u03b9\u03bd\u03ae \u03c3\u03c5\u03bd\u03b8\u03b5\u03c4\u03b9\u03ba\u03ae \u03b4\u03b9\u03b1\u03b4\u03b9\u03ba\u03b1\u03c3\u03af\u03b1, \u03ce\u03c3\u03c4\u03b5 \u03bd\u03b1 \u03b1\u03c0\u03bf\u03bc\u03bf\u03bd\u03c9\u03b8\u03b5\u03af \u03b7 \u03b1\u03be\u03af\u03b1 \u03c4\u03b7\u03c2 \u03b5\u03c0\u03b9\u03bb\u03bf\u03b3\u03ae\u03c2 \u03c4\u03c9\u03bd \u03b4\u03b5\u03b4\u03bf\u03bc\u03ad\u03bd\u03c9\u03bd.<\/p>\n<p>\u0393\u03b9\u03b1 \u03ba\u03ac\u03b8\u03b5 \u03b1\u03b4\u03cd\u03bd\u03b1\u03bc\u03b7 \u03b9\u03ba\u03b1\u03bd\u03cc\u03c4\u03b7\u03c4\u03b1 \u03b4\u03b7\u03bc\u03b9\u03bf\u03c5\u03c1\u03b3\u03ae\u03b8\u03b7\u03ba\u03b1\u03bd 40 prompts. \u039f generator \u03ad\u03bb\u03b1\u03b2\u03b5 \u03c4\u03b7\u03bd \u03c0\u03b5\u03c1\u03b9\u03b3\u03c1\u03b1\u03c6\u03ae \u03c4\u03bf\u03c5 \u03ba\u03cc\u03bc\u03b2\u03bf\u03c5, \u03c0\u03b1\u03c1\u03b1\u03b4\u03b5\u03af\u03b3\u03bc\u03b1\u03c4\u03b1 \u03b1\u03c0\u03cc \u03b1\u03c5\u03c4\u03cc\u03bd \u03ba\u03b1\u03b9 \u03ba\u03bf\u03bd\u03c4\u03b9\u03bd\u03cc context \u03b1\u03c0\u03cc \u03c4\u03bf \u03b4\u03ad\u03bd\u03c4\u03c1\u03bf, \u03cc\u03c0\u03c9\u03c2 descriptions \u03c4\u03bf\u03c5 parent \u03ba\u03b1\u03b9 \u03c4\u03c9\u03bd siblings. \u03a4\u03bf context \u03b5\u03af\u03c7\u03b5 \u03b4\u03b9\u03c0\u03bb\u03cc \u03c1\u03cc\u03bb\u03bf: \u03bd\u03b1 \u03ba\u03c1\u03b1\u03c4\u03ae\u03c3\u03b5\u03b9 \u03c4\u03b1 \u03bd\u03ad\u03b1 \u03c0\u03b1\u03c1\u03b1\u03b4\u03b5\u03af\u03b3\u03bc\u03b1\u03c4\u03b1 \u03ba\u03bf\u03bd\u03c4\u03ac \u03c3\u03c4\u03b7\u03bd \u03b5\u03c0\u03b9\u03b8\u03c5\u03bc\u03b7\u03c4\u03ae \u03b9\u03ba\u03b1\u03bd\u03cc\u03c4\u03b7\u03c4\u03b1 \u03ba\u03b1\u03b9 \u03c4\u03b1\u03c5\u03c4\u03cc\u03c7\u03c1\u03bf\u03bd\u03b1 \u03bd\u03b1 \u03c4\u03b1 \u03b4\u03b9\u03b1\u03c7\u03c9\u03c1\u03af\u03c3\u03b5\u03b9 \u03b1\u03c0\u03cc \u03b3\u03b5\u03b9\u03c4\u03bf\u03bd\u03b9\u03ba\u03ad\u03c2 \u03b9\u03ba\u03b1\u03bd\u03cc\u03c4\u03b7\u03c4\u03b5\u03c2. \u0391\u03ba\u03bf\u03bb\u03bf\u03cd\u03b8\u03b7\u03c3\u03b5 \u03b5\u03bb\u03b1\u03c6\u03c1\u03cd \u03c6\u03b9\u03bb\u03c4\u03c1\u03ac\u03c1\u03b9\u03c3\u03bc\u03b1, \u03cc\u03c0\u03c9\u03c2 \u03b1\u03c6\u03b1\u03af\u03c1\u03b5\u03c3\u03b7 near-duplicates \u03bc\u03b5 word-overlap Jaccard similarity \u03ba\u03b1\u03b9 \u03c0\u03b5\u03c1\u03b9\u03bf\u03c1\u03b9\u03c3\u03bc\u03bf\u03af \u03c3\u03b5 \u03b5\u03c0\u03b1\u03bd\u03b1\u03bb\u03b1\u03bc\u03b2\u03b1\u03bd\u03cc\u03bc\u03b5\u03bd\u03b1 \u03bc\u03bf\u03c4\u03af\u03b2\u03b1 prompts.<\/p>\n<p>\u039f\u03b9 \u03b1\u03c0\u03b1\u03bd\u03c4\u03ae\u03c3\u03b5\u03b9\u03c2 \u03b4\u03b7\u03bc\u03b9\u03bf\u03c5\u03c1\u03b3\u03ae\u03b8\u03b7\u03ba\u03b1\u03bd \u03b1\u03c0\u03cc teacher model \u03bc\u03b5 \u03c4\u03b7\u03bd \u03af\u03b4\u03b9\u03b1 \u03c3\u03c5\u03bd\u03c4\u03b1\u03b3\u03ae \u03b3\u03b9\u03b1 \u03cc\u03bb\u03b5\u03c2 \u03c4\u03b9\u03c2 \u03c3\u03c5\u03b3\u03ba\u03c1\u03b9\u03bd\u03cc\u03bc\u03b5\u03bd\u03b5\u03c2 \u03bc\u03b5\u03b8\u03cc\u03b4\u03bf\u03c5\u03c2. \u03a4\u03bf budget \u03ae\u03c4\u03b1\u03bd 1.000 \u03c0\u03b1\u03c1\u03b1\u03b4\u03b5\u03af\u03b3\u03bc\u03b1\u03c4\u03b1 \u03ba\u03b1\u03b9 \u03b4\u03b9\u03b1\u03c4\u03b7\u03c1\u03ae\u03b8\u03b7\u03ba\u03b1\u03bd \u03af\u03b4\u03b9\u03b1 \u03c4\u03b1 hyperparameters \u03c4\u03bf\u03c5 supervised fine-tuning \u03ba\u03b1\u03b9 \u03c4\u03bf evaluation protocol. \u0391\u03c5\u03c4\u03ae \u03b7 \u03c0\u03b5\u03b9\u03c1\u03b1\u03bc\u03b1\u03c4\u03b9\u03ba\u03ae \u03b9\u03c3\u03cc\u03c4\u03b7\u03c4\u03b1 \u03b5\u03af\u03bd\u03b1\u03b9 \u03ba\u03b5\u03bd\u03c4\u03c1\u03b9\u03ba\u03ae: \u03c4\u03bf CRAFT, \u03c4\u03bf EvalTree \u03ba\u03b1\u03b9 \u03c4\u03bf Random \u03b4\u03b9\u03ad\u03c6\u03b5\u03c1\u03b1\u03bd \u03c3\u03c4\u03bf\u03bd \u03c4\u03c1\u03cc\u03c0\u03bf \u03c3\u03c4\u03cc\u03c7\u03b5\u03c5\u03c3\u03b7\u03c2 \u03c4\u03c9\u03bd \u03b4\u03b5\u03b4\u03bf\u03bc\u03ad\u03bd\u03c9\u03bd, \u03cc\u03c7\u03b9 \u03c3\u03c4\u03bf \u03c0\u03cc\u03c3\u03b1 \u03b4\u03b5\u03b4\u03bf\u03bc\u03ad\u03bd\u03b1 \u03ad\u03bb\u03b1\u03b2\u03b1\u03bd \u03ae \u03c3\u03c4\u03bf \u03c0\u03ce\u03c2 \u03b5\u03ba\u03c0\u03b1\u03b9\u03b4\u03b5\u03cd\u03c4\u03b7\u03ba\u03b1\u03bd.<\/p>\n<h2 id=\"baselines-evaltree-random\">\u039f\u03b9 \u03b4\u03cd\u03bf baselines \u03ba\u03b1\u03b9 \u03c4\u03b9 \u03b1\u03ba\u03c1\u03b9\u03b2\u03ce\u03c2 \u03b5\u03bb\u03ad\u03b3\u03c7\u03bf\u03c5\u03bd<\/h2>\n<p>\u03a4\u03bf EvalTree \u03b5\u03af\u03bd\u03b1\u03b9 \u03b7 \u03c0\u03b9\u03bf \u03ba\u03bf\u03bd\u03c4\u03b9\u03bd\u03ae \u03c3\u03cd\u03b3\u03ba\u03c1\u03b9\u03c3\u03b7. \u039a\u03b1\u03b9 \u03b1\u03c5\u03c4\u03cc \u03b4\u03b7\u03bc\u03b9\u03bf\u03c5\u03c1\u03b3\u03b5\u03af \u03b9\u03b5\u03c1\u03b1\u03c1\u03c7\u03af\u03b1 \u03ba\u03b1\u03b9 \u03b5\u03bd\u03c4\u03bf\u03c0\u03af\u03b6\u03b5\u03b9 \u03b1\u03b4\u03cd\u03bd\u03b1\u03bc\u03b5\u03c2 \u03c0\u03b5\u03c1\u03b9\u03bf\u03c7\u03ad\u03c2, \u03b1\u03bb\u03bb\u03ac \u03b7 \u03bc\u03bf\u03bd\u03ac\u03b4\u03b1 \u03c4\u03bf\u03c5 \u03b5\u03af\u03bd\u03b1\u03b9 \u03bf\u03bb\u03cc\u03ba\u03bb\u03b7\u03c1\u03bf \u03c4\u03bf prompt. \u0388\u03bd\u03b1\u03c2 prompt-level \u03ba\u03cc\u03bc\u03b2\u03bf\u03c2 \u03bc\u03c0\u03bf\u03c1\u03b5\u03af \u03bd\u03b1 \u03b4\u03b5\u03af\u03be\u03b5\u03b9 \u03cc\u03c4\u03b9 \u03ad\u03bd\u03b1 \u03bc\u03bf\u03bd\u03c4\u03ad\u03bb\u03bf \u03b4\u03c5\u03c3\u03ba\u03bf\u03bb\u03b5\u03cd\u03b5\u03c4\u03b1\u03b9 \u03c3\u03b5 \u03ad\u03bd\u03b1 \u03b5\u03af\u03b4\u03bf\u03c2 \u03b5\u03c1\u03ce\u03c4\u03b7\u03c3\u03b7\u03c2, \u03cc\u03c7\u03b9 \u03c0\u03bf\u03b9\u03b1 \u03b1\u03c0\u03b1\u03af\u03c4\u03b7\u03c3\u03b7 \u03c4\u03b7\u03c2 \u03b1\u03c0\u03ac\u03bd\u03c4\u03b7\u03c3\u03b7\u03c2 \u03c0\u03b1\u03c1\u03ad\u03bb\u03b5\u03b9\u03c8\u03b5. \u0397 \u03c3\u03cd\u03b3\u03ba\u03c1\u03b9\u03c3\u03b7 \u03bc\u03b5 \u03c4\u03bf CRAFT \u03b5\u03bb\u03ad\u03b3\u03c7\u03b5\u03b9 \u03b5\u03c0\u03bf\u03bc\u03ad\u03bd\u03c9\u03c2 \u03b1\u03bd \u03c4\u03bf \u03ba\u03c1\u03b9\u03c4\u03ae\u03c1\u03b9\u03bf \u03c4\u03b7\u03c2 rubric \u03c0\u03c1\u03bf\u03c3\u03b8\u03ad\u03c4\u03b5\u03b9 \u03b4\u03b9\u03b1\u03b3\u03bd\u03c9\u03c3\u03c4\u03b9\u03ba\u03ae \u03c0\u03bb\u03b7\u03c1\u03bf\u03c6\u03bf\u03c1\u03af\u03b1 \u03c0\u03ad\u03c1\u03b1 \u03b1\u03c0\u03cc \u03c4\u03bf prompt. \u03a0\u03c1\u03cc\u03ba\u03b5\u03b9\u03c4\u03b1\u03b9 \u03b3\u03b9\u03b1 \u03b4\u03b9\u03b1\u03c6\u03bf\u03c1\u03b5\u03c4\u03b9\u03ba\u03cc \u03b5\u03c0\u03af\u03c0\u03b5\u03b4\u03bf attribution \u03b1\u03c0\u03cc \u03c4\u03bf <a href=\"https:\/\/twodots.gr\/lotapo-ai-agents-anazitisi-credit-assignment\/\">LOTAPO, \u03c0\u03bf\u03c5 \u03bc\u03b5\u03c4\u03c1\u03ac \u03c0\u03bf\u03b9\u03bf \u03b2\u03ae\u03bc\u03b1 \u03b1\u03bd\u03b1\u03b6\u03ae\u03c4\u03b7\u03c3\u03b7\u03c2 \u03b5\u03bd\u03cc\u03c2 AI agent \u03b2\u03bf\u03ae\u03b8\u03b7\u03c3\u03b5 \u03c4\u03bf \u03c4\u03b5\u03bb\u03b9\u03ba\u03cc \u03b1\u03c0\u03bf\u03c4\u03ad\u03bb\u03b5\u03c3\u03bc\u03b1<\/a>: \u03b5\u03b4\u03ce \u03bf \u03c3\u03c4\u03cc\u03c7\u03bf\u03c2 \u03b5\u03af\u03bd\u03b1\u03b9 \u03b7 \u03b5\u03c0\u03b1\u03bd\u03b1\u03bb\u03b1\u03bc\u03b2\u03b1\u03bd\u03cc\u03bc\u03b5\u03bd\u03b7 \u03b9\u03ba\u03b1\u03bd\u03cc\u03c4\u03b7\u03c4\u03b1 \u03c4\u03bf\u03c5 \u03bc\u03bf\u03bd\u03c4\u03ad\u03bb\u03bf\u03c5.<\/p>\n<p>\u03a4\u03bf Random baseline \u03b1\u03c6\u03b1\u03b9\u03c1\u03b5\u03af \u03ba\u03b1\u03b9 \u03c4\u03b7\u03bd \u03b9\u03b5\u03c1\u03b1\u03c1\u03c7\u03af\u03b1 \u03ba\u03b1\u03b9 \u03c4\u03b7 \u03c3\u03c4\u03cc\u03c7\u03b5\u03c5\u03c3\u03b7 \u03b1\u03b4\u03c5\u03bd\u03b1\u03bc\u03b9\u03ce\u03bd. \u0395\u03c0\u03b9\u03bb\u03ad\u03b3\u03b5\u03b9 seed prompts \u03bf\u03bc\u03bf\u03b9\u03cc\u03bc\u03bf\u03c1\u03c6\u03b1 \u03b1\u03c0\u03cc \u03c4\u03bf \u03af\u03b4\u03b9\u03bf domain \u03ba\u03b1\u03b9 \u03c4\u03b1 \u03c0\u03b5\u03c1\u03bd\u03ac \u03c3\u03c4\u03b7\u03bd \u03af\u03b4\u03b9\u03b1 downstream \u03b4\u03b9\u03b1\u03b4\u03b9\u03ba\u03b1\u03c3\u03af\u03b1. \u0391\u03bd \u03c4\u03bf CRAFT \u03c5\u03c0\u03b5\u03c1\u03ad\u03c7\u03b5\u03b9 \u03c4\u03bf\u03c5 Random, \u03b7 \u03b4\u03b9\u03b1\u03c6\u03bf\u03c1\u03ac \u03b4\u03b5\u03bd \u03bc\u03c0\u03bf\u03c1\u03b5\u03af \u03bd\u03b1 \u03b1\u03c0\u03bf\u03b4\u03bf\u03b8\u03b5\u03af \u03b1\u03c0\u03bb\u03ce\u03c2 \u03c3\u03c4\u03b7\u03bd \u03c0\u03c1\u03bf\u03c3\u03b8\u03ae\u03ba\u03b7 \u03c0\u03b5\u03c1\u03b9\u03c3\u03c3\u03cc\u03c4\u03b5\u03c1\u03c9\u03bd \u03c3\u03c5\u03bd\u03b8\u03b5\u03c4\u03b9\u03ba\u03ce\u03bd \u03c0\u03b1\u03c1\u03b1\u03b4\u03b5\u03b9\u03b3\u03bc\u03ac\u03c4\u03c9\u03bd \u03c4\u03bf\u03c5 \u03af\u03b4\u03b9\u03bf\u03c5 domain. \u03a5\u03c0\u03bf\u03c3\u03c4\u03b7\u03c1\u03af\u03b6\u03b5\u03b9 \u03c4\u03b7\u03bd \u03b1\u03be\u03af\u03b1 \u03c4\u03b7\u03c2 \u03bf\u03c1\u03b3\u03b1\u03bd\u03c9\u03bc\u03ad\u03bd\u03b7\u03c2 \u03b4\u03b9\u03ac\u03b3\u03bd\u03c9\u03c3\u03b7\u03c2 \u03ba\u03b1\u03b9 \u03c4\u03b7\u03c2 \u03c3\u03c4\u03bf\u03c7\u03b5\u03c5\u03bc\u03ad\u03bd\u03b7\u03c2 \u03b5\u03c0\u03b9\u03bb\u03bf\u03b3\u03ae\u03c2.<\/p>\n<h2 id=\"montela-domains-benchmarks\">\u03a4\u03bf \u03c0\u03b5\u03b9\u03c1\u03b1\u03bc\u03b1\u03c4\u03b9\u03ba\u03cc \u03c0\u03b5\u03b4\u03af\u03bf: \u03c4\u03ad\u03c3\u03c3\u03b5\u03c1\u03b1 \u03bc\u03bf\u03bd\u03c4\u03ad\u03bb\u03b1 \u03ba\u03b1\u03b9 \u03b4\u03cd\u03bf \u03b1\u03c0\u03b1\u03b9\u03c4\u03b7\u03c4\u03b9\u03ba\u03ac domains<\/h2>\n<p>\u039f\u03b9 \u03b5\u03c1\u03b5\u03c5\u03bd\u03b7\u03c4\u03ad\u03c2 \u03b1\u03be\u03b9\u03bf\u03bb\u03cc\u03b3\u03b7\u03c3\u03b1\u03bd \u03c4\u03b1 Qwen3-4B, Qwen3-8B, Gemma-3-4B \u03ba\u03b1\u03b9 Llama-3.1-8B-Instruct \u03c3\u03b5 \u03c7\u03c1\u03b7\u03bc\u03b1\u03c4\u03bf\u03bf\u03b9\u03ba\u03bf\u03bd\u03bf\u03bc\u03b9\u03ba\u03ac \u03ba\u03b1\u03b9 \u03bd\u03bf\u03bc\u03b9\u03ba\u03ac tasks. \u0393\u03b9\u03b1 \u03c4\u03b7 \u03b4\u03b9\u03ac\u03b3\u03bd\u03c9\u03c3\u03b7 \u03c7\u03c1\u03b7\u03c3\u03b9\u03bc\u03bf\u03c0\u03bf\u03af\u03b7\u03c3\u03b1\u03bd \u03c4\u03bf finance subset \u03c4\u03bf\u03c5 PRBench \u03bc\u03b5 629 prompts \u03ba\u03b1\u03b9 10.806 \u03ba\u03c1\u03b9\u03c4\u03ae\u03c1\u03b9\u03b1, \u03ba\u03b1\u03b9 \u03c4\u03bf legal subset \u03bc\u03b5 532 prompts \u03ba\u03b1\u03b9 9.637 \u03ba\u03c1\u03b9\u03c4\u03ae\u03c1\u03b9\u03b1. \u03a4\u03b1 \u03b4\u03b5\u03b4\u03bf\u03bc\u03ad\u03bd\u03b1 \u03b4\u03b9\u03ac\u03b3\u03bd\u03c9\u03c3\u03b7\u03c2 \u03ae\u03c4\u03b1\u03bd \u03be\u03b5\u03c7\u03c9\u03c1\u03b9\u03c3\u03c4\u03ac \u03b1\u03c0\u03cc \u03c4\u03b1 benchmarks \u03c4\u03b5\u03bb\u03b9\u03ba\u03ae\u03c2 \u03b1\u03be\u03b9\u03bf\u03bb\u03cc\u03b3\u03b7\u03c3\u03b7\u03c2.<\/p>\n<p>\u0397 \u03c4\u03b5\u03bb\u03b9\u03ba\u03ae \u03b4\u03bf\u03ba\u03b9\u03bc\u03ae \u03c0\u03b5\u03c1\u03b9\u03bb\u03ac\u03bc\u03b2\u03b1\u03bd\u03b5 13 held-out benchmarks: \u03b5\u03c0\u03c4\u03ac \u03bd\u03bf\u03bc\u03b9\u03ba\u03ac \u03ba\u03b1\u03b9 \u03ad\u03be\u03b9 \u03c7\u03c1\u03b7\u03bc\u03b1\u03c4\u03bf\u03bf\u03b9\u03ba\u03bf\u03bd\u03bf\u03bc\u03b9\u03ba\u03ac. \u03a4\u03bf \u03bd\u03bf\u03bc\u03b9\u03ba\u03cc suite \u03ba\u03ac\u03bb\u03c5\u03c8\u03b5 \u03bc\u03b5\u03c4\u03b1\u03be\u03cd \u03ac\u03bb\u03bb\u03c9\u03bd doctrinal knowledge, \u03b5\u03c0\u03b9\u03bb\u03bf\u03b3\u03ae case holdings, \u03b5\u03c6\u03b1\u03c1\u03bc\u03bf\u03b3\u03ae \u03bd\u03cc\u03bc\u03bf\u03c5 \u03c3\u03b5 \u03bd\u03ad\u03b1 \u03c0\u03b5\u03c1\u03b9\u03c3\u03c4\u03b1\u03c4\u03b9\u03ba\u03ac \u03ba\u03b1\u03b9 \u03b5\u03c1\u03bc\u03b7\u03bd\u03b5\u03af\u03b1 \u03c3\u03c5\u03bc\u03b2\u03ac\u03c3\u03b5\u03c9\u03bd. \u03a4\u03bf \u03c7\u03c1\u03b7\u03bc\u03b1\u03c4\u03bf\u03bf\u03b9\u03ba\u03bf\u03bd\u03bf\u03bc\u03b9\u03ba\u03cc suite \u03ba\u03ac\u03bb\u03c5\u03c8\u03b5 \u03b5\u03c1\u03c9\u03c4\u03b1\u03c0\u03b1\u03bd\u03c4\u03ae\u03c3\u03b5\u03b9\u03c2 \u03c0\u03ac\u03bd\u03c9 \u03c3\u03b5 filings, \u03b1\u03c1\u03b9\u03b8\u03bc\u03b7\u03c4\u03b9\u03ba\u03cc \u03c3\u03c5\u03bb\u03bb\u03bf\u03b3\u03b9\u03c3\u03bc\u03cc, \u03c3\u03c5\u03bd\u03b4\u03c5\u03b1\u03c3\u03bc\u03cc \u03c0\u03b9\u03bd\u03ac\u03ba\u03c9\u03bd \u03ba\u03b1\u03b9 \u03ba\u03b5\u03b9\u03bc\u03ad\u03bd\u03bf\u03c5, sentiment, \u03c3\u03c4\u03ac\u03c3\u03b7 \u03bd\u03bf\u03bc\u03b9\u03c3\u03bc\u03b1\u03c4\u03b9\u03ba\u03ae\u03c2 \u03c0\u03bf\u03bb\u03b9\u03c4\u03b9\u03ba\u03ae\u03c2 \u03ba\u03b1\u03b9 ESG classification.<\/p>\n<p>\u03a3\u03b7\u03bc\u03b1\u03bd\u03c4\u03b9\u03ba\u03cc \u03b5\u03af\u03bd\u03b1\u03b9 \u03cc\u03c4\u03b9 \u03c4\u03b1 held-out tasks \u03b4\u03b5\u03bd \u03b5\u03af\u03c7\u03b1\u03bd \u03c4\u03bf \u03af\u03b4\u03b9\u03bf long-form format \u03bc\u03b5 \u03c4\u03b1 \u03b4\u03b5\u03b4\u03bf\u03bc\u03ad\u03bd\u03b1 \u03b4\u03b9\u03ac\u03b3\u03bd\u03c9\u03c3\u03b7\u03c2. \u03a0\u03b5\u03c1\u03b9\u03bb\u03ac\u03bc\u03b2\u03b1\u03bd\u03b1\u03bd multiple choice, classification, extraction \u03ba\u03b1\u03b9 numeric answers. \u0386\u03c1\u03b1 \u03c4\u03b1 \u03b1\u03c0\u03bf\u03c4\u03b5\u03bb\u03ad\u03c3\u03bc\u03b1\u03c4\u03b1 \u03c3\u03c7\u03b5\u03b4\u03b9\u03ac\u03c3\u03c4\u03b7\u03ba\u03b1\u03bd \u03b3\u03b9\u03b1 \u03bd\u03b1 \u03bc\u03b5\u03c4\u03c1\u03ae\u03c3\u03bf\u03c5\u03bd \u03bc\u03b5\u03c4\u03b1\u03c6\u03bf\u03c1\u03ac \u03b9\u03ba\u03b1\u03bd\u03cc\u03c4\u03b7\u03c4\u03b1\u03c2, \u03cc\u03c7\u03b9 \u03b1\u03c0\u03bf\u03bc\u03bd\u03b7\u03bc\u03cc\u03bd\u03b5\u03c5\u03c3\u03b7 \u03cd\u03c6\u03bf\u03c5\u03c2 \u03b1\u03c0\u03ac\u03bd\u03c4\u03b7\u03c3\u03b7\u03c2. \u039a\u03ac\u03b8\u03b5 checkpoint \u03b1\u03be\u03b9\u03bf\u03bb\u03bf\u03b3\u03ae\u03b8\u03b7\u03ba\u03b5 \u03c3\u03b5 \u03b5\u03c0\u03b1\u03bd\u03b1\u03bb\u03b1\u03bc\u03b2\u03b1\u03bd\u03cc\u03bc\u03b5\u03bd\u03b1 temperature-decoding runs \u03bc\u03b5 \u03b4\u03b9\u03b1\u03c6\u03bf\u03c1\u03b5\u03c4\u03b9\u03ba\u03ac seeds \u03ba\u03b1\u03b9 \u03bf\u03b9 \u03c0\u03af\u03bd\u03b1\u03ba\u03b5\u03c2 \u03b1\u03bd\u03b1\u03c6\u03ad\u03c1\u03bf\u03c5\u03bd \u03bc\u03ad\u03c3\u03bf \u03cc\u03c1\u03bf \u03ba\u03b1\u03b9 sample standard deviation.<\/p>\n<h2 id=\"apotelesmata-nomika-benchmarks\">\u03a4\u03b9 \u03ad\u03b4\u03b5\u03b9\u03be\u03b1\u03bd \u03c4\u03b1 \u03b1\u03c0\u03bf\u03c4\u03b5\u03bb\u03ad\u03c3\u03bc\u03b1\u03c4\u03b1 \u03c3\u03c4\u03b1 \u03bd\u03bf\u03bc\u03b9\u03ba\u03ac benchmarks<\/h2>\n<p>\u03a3\u03c4\u03bf \u03bd\u03bf\u03bc\u03b9\u03ba\u03cc domain, \u03c4\u03bf CRAFT \u03c0\u03ad\u03c4\u03c5\u03c7\u03b5 \u03c4\u03bf\u03bd \u03c5\u03c8\u03b7\u03bb\u03cc\u03c4\u03b5\u03c1\u03bf \u03bc\u03ad\u03c3\u03bf \u03cc\u03c1\u03bf \u03c3\u03b5 \u03c4\u03c1\u03af\u03b1 \u03b1\u03c0\u03cc \u03c4\u03b1 \u03c4\u03ad\u03c3\u03c3\u03b5\u03c1\u03b1 \u03bc\u03bf\u03bd\u03c4\u03ad\u03bb\u03b1. \u0393\u03b9\u03b1 \u03c4\u03bf Qwen3-4B \u03bf \u03bc\u03ad\u03c3\u03bf\u03c2 \u03cc\u03c1\u03bf\u03c2 \u03ae\u03c4\u03b1\u03bd 53,0, \u03ad\u03bd\u03b1\u03bd\u03c4\u03b9 51,7 \u03b3\u03b9\u03b1 \u03c4\u03bf EvalTree \u03ba\u03b1\u03b9 49,9 \u03b3\u03b9\u03b1 \u03c4\u03bf Random. \u0393\u03b9\u03b1 \u03c4\u03bf Gemma-3-4B \u03b7 \u03b4\u03b9\u03b1\u03c6\u03bf\u03c1\u03ac \u03ae\u03c4\u03b1\u03bd \u03bc\u03b9\u03ba\u03c1\u03cc\u03c4\u03b5\u03c1\u03b7: 51,7 \u03ad\u03bd\u03b1\u03bd\u03c4\u03b9 50,7 \u03ba\u03b1\u03b9 51,4. \u0393\u03b9\u03b1 \u03c4\u03bf Llama-3.1-8B, \u03c4\u03bf CRAFT \u03ad\u03c6\u03c4\u03b1\u03c3\u03b5 \u03c4\u03bf 51,0, \u03ad\u03bd\u03b1\u03bd\u03c4\u03b9 48,2 \u03c4\u03bf\u03c5 EvalTree \u03ba\u03b1\u03b9 46,9 \u03c4\u03bf\u03c5 Random.<\/p>\n<p>\u0397 \u03b5\u03be\u03b1\u03af\u03c1\u03b5\u03c3\u03b7 \u03ae\u03c4\u03b1\u03bd \u03c4\u03bf Qwen3-8B. \u0395\u03ba\u03b5\u03af \u03c4\u03bf EvalTree \u03b5\u03af\u03c7\u03b5 \u03bc\u03ad\u03c3\u03bf \u03cc\u03c1\u03bf 62,1 \u03ba\u03b1\u03b9 \u03c4\u03bf CRAFT 61,5, \u03b5\u03bd\u03ce \u03c4\u03bf Random 61,2. \u0397 \u03b4\u03b9\u03b1\u03c6\u03bf\u03c1\u03ac \u03c4\u03c9\u03bd 0,6 \u03bc\u03bf\u03bd\u03ac\u03b4\u03c9\u03bd \u03b1\u03bd\u03ac\u03bc\u03b5\u03c3\u03b1 \u03c3\u03c4\u03b9\u03c2 \u03b4\u03cd\u03bf \u03b9\u03b5\u03c1\u03b1\u03c1\u03c7\u03b9\u03ba\u03ad\u03c2 \u03bc\u03b5\u03b8\u03cc\u03b4\u03bf\u03c5\u03c2 \u03b2\u03c1\u03b9\u03c3\u03ba\u03cc\u03c4\u03b1\u03bd \u03bc\u03ad\u03c3\u03b1 \u03c3\u03c4\u03b1 \u03b5\u03c0\u03b9\u03ba\u03b1\u03bb\u03c5\u03c0\u03c4\u03cc\u03bc\u03b5\u03bd\u03b1 variance bands. \u039f\u03b9 \u03af\u03b4\u03b9\u03bf\u03b9 \u03bf\u03b9 \u03c3\u03c5\u03b3\u03b3\u03c1\u03b1\u03c6\u03b5\u03af\u03c2 \u03b4\u03b5\u03bd \u03c0\u03b1\u03c1\u03bf\u03c5\u03c3\u03b9\u03ac\u03b6\u03bf\u03c5\u03bd \u03c4\u03bf \u03b1\u03c0\u03bf\u03c4\u03ad\u03bb\u03b5\u03c3\u03bc\u03b1 \u03c9\u03c2 \u03ba\u03b1\u03b8\u03bf\u03bb\u03b9\u03ba\u03ae \u03bd\u03af\u03ba\u03b7 \u03c3\u03b5 \u03ba\u03ac\u03b8\u03b5 task. \u03a4\u03b1 \u03b5\u03c0\u03b9\u03bc\u03ad\u03c1\u03bf\u03c5\u03c2 benchmarks \u03b5\u03af\u03c7\u03b1\u03bd \u03bc\u03b9\u03ba\u03c4\u03ae \u03b5\u03b9\u03ba\u03cc\u03bd\u03b1, \u03ba\u03ac\u03c4\u03b9 \u03b1\u03bd\u03b1\u03bc\u03b5\u03bd\u03cc\u03bc\u03b5\u03bd\u03bf \u03c3\u03b5 \u03ad\u03bd\u03b1 suite \u03c0\u03bf\u03c5 \u03c3\u03c5\u03bd\u03b4\u03c5\u03ac\u03b6\u03b5\u03b9 \u03b4\u03b9\u03b1\u03c6\u03bf\u03c1\u03b5\u03c4\u03b9\u03ba\u03ad\u03c2 \u03bd\u03bf\u03bc\u03b9\u03ba\u03ad\u03c2 \u03b9\u03ba\u03b1\u03bd\u03cc\u03c4\u03b7\u03c4\u03b5\u03c2.<\/p>\n<p>\u0397 \u03c0\u03c1\u03bf\u03c3\u03b5\u03ba\u03c4\u03b9\u03ba\u03ae \u03b1\u03bd\u03ac\u03b3\u03bd\u03c9\u03c3\u03b7 \u03b5\u03af\u03bd\u03b1\u03b9 \u03cc\u03c4\u03b9 \u03c4\u03bf CRAFT \u03b2\u03b5\u03bb\u03c4\u03af\u03c9\u03c3\u03b5 \u03c4\u03b7 \u03c3\u03c5\u03bd\u03bf\u03bb\u03b9\u03ba\u03ae \u03bc\u03b5\u03c4\u03b1\u03c6\u03bf\u03c1\u03ac \u03c3\u03c4\u03b9\u03c2 \u03c0\u03b5\u03c1\u03b9\u03c3\u03c3\u03cc\u03c4\u03b5\u03c1\u03b5\u03c2 model-domain \u03c3\u03c5\u03bd\u03b8\u03ae\u03ba\u03b5\u03c2, \u03cc\u03c7\u03b9 \u03cc\u03c4\u03b9 \u03ba\u03ad\u03c1\u03b4\u03b9\u03c3\u03b5 \u03ba\u03ac\u03b8\u03b5 \u03b3\u03c1\u03b1\u03bc\u03bc\u03ae. \u0391\u03c5\u03c4\u03ae \u03b7 \u03b4\u03b9\u03ac\u03ba\u03c1\u03b9\u03c3\u03b7 \u03ad\u03c7\u03b5\u03b9 \u03b1\u03be\u03af\u03b1 \u03ba\u03b1\u03b9 \u03b3\u03b9\u03b1 \u03b5\u03c0\u03b9\u03c7\u03b5\u03b9\u03c1\u03b7\u03c3\u03b9\u03b1\u03ba\u03ac pilots: \u03ad\u03bd\u03b1\u03c2 \u03ba\u03b1\u03bb\u03cd\u03c4\u03b5\u03c1\u03bf\u03c2 \u03c3\u03c5\u03bd\u03bf\u03bb\u03b9\u03ba\u03cc\u03c2 \u03bc\u03ad\u03c3\u03bf\u03c2 \u03cc\u03c1\u03bf\u03c2 \u03b4\u03b5\u03bd \u03ba\u03b1\u03c4\u03b1\u03c1\u03b3\u03b5\u03af \u03c4\u03b7\u03bd \u03b1\u03bd\u03ac\u03b3\u03ba\u03b7 \u03b5\u03bb\u03ad\u03b3\u03c7\u03bf\u03c5 \u03c4\u03c9\u03bd \u03c3\u03c5\u03b3\u03ba\u03b5\u03ba\u03c1\u03b9\u03bc\u03ad\u03bd\u03c9\u03bd \u03ba\u03c1\u03af\u03c3\u03b9\u03bc\u03c9\u03bd tasks \u03c4\u03b7\u03c2 \u03b5\u03ba\u03ac\u03c3\u03c4\u03bf\u03c4\u03b5 \u03b5\u03c6\u03b1\u03c1\u03bc\u03bf\u03b3\u03ae\u03c2.<\/p>\n<h2 id=\"apotelesmata-chrimatooikonomika\">\u0397 \u03c0\u03b9\u03bf \u03ba\u03b1\u03b8\u03b1\u03c1\u03ae \u03b5\u03b9\u03ba\u03cc\u03bd\u03b1 \u03c3\u03c4\u03b1 \u03c7\u03c1\u03b7\u03bc\u03b1\u03c4\u03bf\u03bf\u03b9\u03ba\u03bf\u03bd\u03bf\u03bc\u03b9\u03ba\u03ac<\/h2>\n<p>\u03a3\u03c4\u03b1 \u03c7\u03c1\u03b7\u03bc\u03b1\u03c4\u03bf\u03bf\u03b9\u03ba\u03bf\u03bd\u03bf\u03bc\u03b9\u03ba\u03ac, \u03c4\u03bf CRAFT \u03b5\u03af\u03c7\u03b5 \u03c4\u03bf\u03bd \u03ba\u03b1\u03bb\u03cd\u03c4\u03b5\u03c1\u03bf domain average \u03ba\u03b1\u03b9 \u03b3\u03b9\u03b1 \u03c4\u03b1 \u03c4\u03ad\u03c3\u03c3\u03b5\u03c1\u03b1 \u03bc\u03bf\u03bd\u03c4\u03ad\u03bb\u03b1. \u03a3\u03c4\u03bf Qwen3-4B \u03ad\u03c6\u03c4\u03b1\u03c3\u03b5 \u03c4\u03bf 46,0, \u03ad\u03bd\u03b1\u03bd\u03c4\u03b9 40,8 \u03b3\u03b9\u03b1 \u03c4\u03bf EvalTree \u03ba\u03b1\u03b9 37,6 \u03b3\u03b9\u03b1 \u03c4\u03bf Random. \u03a3\u03c4\u03bf Qwen3-8B \u03c3\u03b7\u03bc\u03b5\u03af\u03c9\u03c3\u03b5 45,1 \u03ad\u03bd\u03b1\u03bd\u03c4\u03b9 43,5 \u03ba\u03b1\u03b9 44,1. \u03a3\u03c4\u03bf Gemma-3-4B \u03c3\u03b7\u03bc\u03b5\u03af\u03c9\u03c3\u03b5 39,8 \u03ad\u03bd\u03b1\u03bd\u03c4\u03b9 38,6 \u03ba\u03b1\u03b9 38,4. \u03a3\u03c4\u03bf Llama-3.1-8B \u03b7 \u03b4\u03b9\u03b1\u03c6\u03bf\u03c1\u03ac \u03ae\u03c4\u03b1\u03bd \u03ad\u03bd\u03c4\u03bf\u03bd\u03b7: 42,5 \u03ad\u03bd\u03b1\u03bd\u03c4\u03b9 36,0 \u03ba\u03b1\u03b9 36,6.<\/p>\n<p>\u039a\u03b1\u03b9 \u03b5\u03b4\u03ce, \u03cc\u03bc\u03c9\u03c2, \u03b4\u03b5\u03bd \u03c5\u03c0\u03ae\u03c1\u03be\u03b5 \u03bf\u03bc\u03bf\u03b9\u03cc\u03bc\u03bf\u03c1\u03c6\u03b7 \u03b5\u03c0\u03b9\u03ba\u03c1\u03ac\u03c4\u03b7\u03c3\u03b7 \u03c3\u03b5 \u03ba\u03ac\u03b8\u03b5 benchmark. \u0393\u03b9\u03b1 \u03c0\u03b1\u03c1\u03ac\u03b4\u03b5\u03b9\u03b3\u03bc\u03b1, \u03c3\u03c4\u03bf Qwen3-8B \u03c4\u03b1 \u03ba\u03ad\u03c1\u03b4\u03b7 \u03c3\u03c4\u03b1 FiQA-SA, FOMC \u03ba\u03b1\u03b9 MLESG \u03b1\u03bd\u03c4\u03b9\u03c3\u03c4\u03ac\u03b8\u03bc\u03b9\u03c3\u03b1\u03bd \u03b1\u03c3\u03b8\u03b5\u03bd\u03ad\u03c3\u03c4\u03b5\u03c1\u03b1 \u03b1\u03c0\u03bf\u03c4\u03b5\u03bb\u03ad\u03c3\u03bc\u03b1\u03c4\u03b1 \u03c3\u03c4\u03b1 FinanceBench \u03ba\u03b1\u03b9 TAT-QA. \u03a4\u03b1 open-ended numerical tasks \u03c0\u03b1\u03c1\u03ad\u03bc\u03b5\u03b9\u03bd\u03b1\u03bd \u03c0\u03b9\u03bf \u03b8\u03bf\u03c1\u03c5\u03b2\u03ce\u03b4\u03b7, \u03b5\u03c0\u03b5\u03b9\u03b4\u03ae \u03b5\u03be\u03b1\u03c1\u03c4\u03ce\u03bd\u03c4\u03b1\u03b9 \u03b1\u03c0\u03cc \u03bc\u03b5\u03b3\u03ac\u03bb\u03b1 \u03b1\u03c0\u03bf\u03c3\u03c0\u03ac\u03c3\u03bc\u03b1\u03c4\u03b1, \u03c5\u03c0\u03bf\u03bb\u03bf\u03b3\u03b9\u03c3\u03bc\u03bf\u03cd\u03c2 \u03ba\u03b1\u03b9 judge-based scoring.<\/p>\n<p>\u03a3\u03c5\u03bd\u03bf\u03bb\u03b9\u03ba\u03ac, \u03c4\u03bf CRAFT \u03b5\u03af\u03c7\u03b5 \u03c4\u03b7\u03bd \u03b9\u03c3\u03c7\u03c5\u03c1\u03cc\u03c4\u03b5\u03c1\u03b7 \u03bc\u03ad\u03c3\u03b7 \u03b5\u03c0\u03af\u03b4\u03bf\u03c3\u03b7 \u03c3\u03b5 \u03b5\u03c0\u03c4\u03ac \u03b1\u03c0\u03cc \u03c4\u03b9\u03c2 \u03bf\u03ba\u03c4\u03ce \u03c3\u03c5\u03bd\u03b8\u03ae\u03ba\u03b5\u03c2 \u03bc\u03bf\u03bd\u03c4\u03ad\u03bb\u03bf\u03c5\u2013domain. \u0397 \u03b5\u03c1\u03b3\u03b1\u03c3\u03af\u03b1 \u03b1\u03c0\u03bf\u03b4\u03af\u03b4\u03b5\u03b9 \u03c4\u03b7 \u03b4\u03b9\u03b1\u03c6\u03bf\u03c1\u03ac \u03c3\u03b5 \u03b4\u03cd\u03bf \u03c3\u03c4\u03bf\u03b9\u03c7\u03b5\u03af\u03b1: \u03b7 \u03c5\u03c0\u03b5\u03c1\u03bf\u03c7\u03ae \u03b1\u03c0\u03ad\u03bd\u03b1\u03bd\u03c4\u03b9 \u03c3\u03c4\u03bf Random \u03c5\u03c0\u03bf\u03c3\u03c4\u03b7\u03c1\u03af\u03b6\u03b5\u03b9 \u03c4\u03b7\u03bd \u03b1\u03be\u03af\u03b1 \u03c4\u03b7\u03c2 \u03b9\u03b5\u03c1\u03b1\u03c1\u03c7\u03af\u03b1\u03c2 \u03ba\u03b1\u03b9 \u03c4\u03b7\u03c2 \u03c3\u03c4\u03cc\u03c7\u03b5\u03c5\u03c3\u03b7\u03c2 \u03b1\u03b4\u03c5\u03bd\u03b1\u03bc\u03b9\u03ce\u03bd, \u03b5\u03bd\u03ce \u03b7 \u03c5\u03c0\u03b5\u03c1\u03bf\u03c7\u03ae \u03b1\u03c0\u03ad\u03bd\u03b1\u03bd\u03c4\u03b9 \u03c3\u03c4\u03bf EvalTree \u03c5\u03c0\u03bf\u03c3\u03c4\u03b7\u03c1\u03af\u03b6\u03b5\u03b9 \u03cc\u03c4\u03b9 \u03c4\u03b1 rubric criteria \u03b5\u03af\u03bd\u03b1\u03b9 \u03c7\u03c1\u03b7\u03c3\u03b9\u03bc\u03cc\u03c4\u03b5\u03c1\u03b7 \u03b4\u03b9\u03b1\u03b3\u03bd\u03c9\u03c3\u03c4\u03b9\u03ba\u03ae \u03bc\u03bf\u03bd\u03ac\u03b4\u03b1 \u03b1\u03c0\u03cc \u03bf\u03bb\u03cc\u03ba\u03bb\u03b7\u03c1\u03b1 prompts \u03c3\u03c4\u03bf \u03c3\u03c5\u03b3\u03ba\u03b5\u03ba\u03c1\u03b9\u03bc\u03ad\u03bd\u03bf setup.<\/p>\n<div class=\"td-chart td-chart--metrics\">\n<div class=\"td-chart-head\">\n<p class=\"td-chart-title\">\u0397 \u03ba\u03bb\u03af\u03bc\u03b1\u03ba\u03b1 \u03ba\u03b1\u03b9 \u03c4\u03bf \u03b2\u03b1\u03c3\u03b9\u03ba\u03cc \u03b1\u03c0\u03bf\u03c4\u03ad\u03bb\u03b5\u03c3\u03bc\u03b1 \u03c4\u03b7\u03c2 \u03bc\u03b5\u03bb\u03ad\u03c4\u03b7\u03c2 CRAFT<\/p>\n<p class=\"td-chart-subtitle\">\u038c\u03bb\u03b1 \u03c4\u03b1 \u03bc\u03b5\u03b3\u03ad\u03b8\u03b7 \u03c0\u03c1\u03bf\u03ad\u03c1\u03c7\u03bf\u03bd\u03c4\u03b1\u03b9 \u03b1\u03c0\u03cc \u03c4\u03b7\u03bd \u03ad\u03ba\u03b4\u03bf\u03c3\u03b7 \u03c4\u03b7\u03c2 \u03b5\u03c1\u03b3\u03b1\u03c3\u03af\u03b1\u03c2 arXiv 2607.16122 \u03c4\u03b7\u03c2 17\u03b7\u03c2 \u0399\u03bf\u03c5\u03bb\u03af\u03bf\u03c5 2026 \u03ba\u03b1\u03b9 \u03b1\u03c6\u03bf\u03c1\u03bf\u03cd\u03bd \u03b1\u03c0\u03bf\u03ba\u03bb\u03b5\u03b9\u03c3\u03c4\u03b9\u03ba\u03ac \u03c4\u03bf \u03c3\u03c5\u03b3\u03ba\u03b5\u03ba\u03c1\u03b9\u03bc\u03ad\u03bd\u03bf experimental setup.<\/p>\n<\/div>\n<div class=\"td-chart-body\">\n<div class=\"td-metric-grid\">\n<div class=\"td-metric-card\"><span class=\"td-metric-value\">20.443<\/span><span class=\"td-metric-label\">\u03b6\u03b5\u03cd\u03b3\u03b7 prompt\u2013rubric \u03c3\u03c4\u03bf capability tree<\/span><span class=\"td-metric-note\">10.806 finance \u03ba\u03b1\u03b9 9.637 legal criteria<\/span><\/div>\n<div class=\"td-metric-card\"><span class=\"td-metric-value\">13<\/span><span class=\"td-metric-label\">held-out benchmarks<\/span><span class=\"td-metric-note\">\u0395\u03c0\u03c4\u03ac \u03bd\u03bf\u03bc\u03b9\u03ba\u03ac \u03ba\u03b1\u03b9 \u03ad\u03be\u03b9 \u03c7\u03c1\u03b7\u03bc\u03b1\u03c4\u03bf\u03bf\u03b9\u03ba\u03bf\u03bd\u03bf\u03bc\u03b9\u03ba\u03ac<\/span><\/div>\n<div class=\"td-metric-card\"><span class=\"td-metric-value\">1.000<\/span><span class=\"td-metric-label\">\u03c0\u03b1\u03c1\u03b1\u03b4\u03b5\u03af\u03b3\u03bc\u03b1\u03c4\u03b1 \u03b1\u03bd\u03ac fine-tuning \u03c3\u03cd\u03b3\u03ba\u03c1\u03b9\u03c3\u03b7<\/span><span class=\"td-metric-note\">\u038a\u03b4\u03b9\u03bf budget \u03ba\u03b1\u03b9 protocol \u03b3\u03b9\u03b1 \u03cc\u03bb\u03b5\u03c2 \u03c4\u03b9\u03c2 \u03bc\u03b5\u03b8\u03cc\u03b4\u03bf\u03c5\u03c2<\/span><\/div>\n<div class=\"td-metric-card\"><span class=\"td-metric-value\">7 \u03b1\u03c0\u03cc 8<\/span><span class=\"td-metric-label\">\u03ba\u03b1\u03bb\u03cd\u03c4\u03b5\u03c1\u03bf\u03b9 model\u2013domain \u03bc\u03ad\u03c3\u03bf\u03b9 \u03cc\u03c1\u03bf\u03b9<\/span><span class=\"td-metric-note\">\u03a4\u03ad\u03c3\u03c3\u03b5\u03c1\u03b9\u03c2 finance \u03ba\u03b1\u03b9 \u03c4\u03c1\u03b5\u03b9\u03c2 legal \u03c3\u03c5\u03bd\u03b8\u03ae\u03ba\u03b5\u03c2<\/span><\/div>\n<\/div>\n<\/div>\n<\/div>\n<h2 id=\"dynamiki-epilogi-vathous\">\u0393\u03b9\u03b1\u03c4\u03af \u03b7 \u03b4\u03c5\u03bd\u03b1\u03bc\u03b9\u03ba\u03ae \u03b5\u03c0\u03b9\u03bb\u03bf\u03b3\u03ae \u03b2\u03ac\u03b8\u03bf\u03c5\u03c2 \u03ad\u03c7\u03b5\u03b9 \u03c0\u03c1\u03b1\u03ba\u03c4\u03b9\u03ba\u03ae \u03b1\u03be\u03af\u03b1<\/h2>\n<p>\u0397 ablation \u03bc\u03b5\u03bb\u03ad\u03c4\u03b7 \u03c3\u03cd\u03b3\u03ba\u03c1\u03b9\u03bd\u03b5 \u03c4\u03b7 \u03b4\u03c5\u03bd\u03b1\u03bc\u03b9\u03ba\u03ae \u03b5\u03c0\u03b9\u03bb\u03bf\u03b3\u03ae \u03bc\u03b5 fixed selection \u03c3\u03c4\u03b1 \u03b5\u03c0\u03af\u03c0\u03b5\u03b4\u03b1 L4 \u03ba\u03b1\u03b9 L5, \u03ba\u03c1\u03b1\u03c4\u03ce\u03bd\u03c4\u03b1\u03c2 \u03cc\u03bb\u03b1 \u03c4\u03b1 \u03ac\u03bb\u03bb\u03b1 \u03af\u03b4\u03b9\u03b1. \u0388\u03b3\u03b9\u03bd\u03b5 \u03c3\u03b5 Qwen3-4B \u03ba\u03b1\u03b9 Llama-3.1-8B, \u03c4\u03cc\u03c3\u03bf \u03c3\u03c4\u03bf legal \u03cc\u03c3\u03bf \u03ba\u03b1\u03b9 \u03c3\u03c4\u03bf finance domain. \u0397 \u03b4\u03c5\u03bd\u03b1\u03bc\u03b9\u03ba\u03ae \u03b4\u03b9\u03b1\u03b4\u03b9\u03ba\u03b1\u03c3\u03af\u03b1 \u03c0\u03ad\u03c4\u03c5\u03c7\u03b5 \u03c4\u03bf \u03ba\u03b1\u03bb\u03cd\u03c4\u03b5\u03c1\u03bf domain average \u03ba\u03b1\u03b9 \u03c3\u03c4\u03b9\u03c2 \u03c4\u03ad\u03c3\u03c3\u03b5\u03c1\u03b9\u03c2 \u03c3\u03c5\u03b3\u03ba\u03c1\u03af\u03c3\u03b5\u03b9\u03c2.<\/p>\n<p>\u03a4\u03bf \u03c0\u03b9\u03bf \u03c7\u03c1\u03ae\u03c3\u03b9\u03bc\u03bf \u03b5\u03cd\u03c1\u03b7\u03bc\u03b1 \u03b4\u03b5\u03bd \u03b5\u03af\u03bd\u03b1\u03b9 \u03bc\u03cc\u03bd\u03bf \u03cc\u03c4\u03b9 \u03ba\u03ad\u03c1\u03b4\u03b9\u03c3\u03b5 \u03b7 dynamic \u03b5\u03c0\u03b9\u03bb\u03bf\u03b3\u03ae. \u0395\u03af\u03bd\u03b1\u03b9 \u03cc\u03c4\u03b9 \u03bf\u03b9 \u03b4\u03cd\u03bf fixed-depth \u03b5\u03bd\u03b1\u03bb\u03bb\u03b1\u03ba\u03c4\u03b9\u03ba\u03ad\u03c2 \u03b4\u03b5\u03bd \u03b5\u03af\u03c7\u03b1\u03bd \u03c3\u03c4\u03b1\u03b8\u03b5\u03c1\u03ae \u03c3\u03b5\u03b9\u03c1\u03ac: \u03ac\u03bb\u03bb\u03bf\u03c4\u03b5 \u03ae\u03c4\u03b1\u03bd \u03ba\u03b1\u03bb\u03cd\u03c4\u03b5\u03c1\u03bf \u03c4\u03bf L4 \u03ba\u03b1\u03b9 \u03ac\u03bb\u03bb\u03bf\u03c4\u03b5 \u03c4\u03bf L5. \u0391\u03c5\u03c4\u03cc \u03b5\u03bd\u03b9\u03c3\u03c7\u03cd\u03b5\u03b9 \u03c4\u03bf \u03b2\u03b1\u03c3\u03b9\u03ba\u03cc \u03c3\u03ba\u03b5\u03c0\u03c4\u03b9\u03ba\u03cc \u03c4\u03bf\u03c5 CRAFT. \u039f\u03b9 \u03b1\u03b4\u03c5\u03bd\u03b1\u03bc\u03af\u03b5\u03c2 \u03b4\u03b9\u03b1\u03c6\u03bf\u03c1\u03b5\u03c4\u03b9\u03ba\u03ce\u03bd \u03bc\u03bf\u03bd\u03c4\u03ad\u03bb\u03c9\u03bd \u03b4\u03b5\u03bd \u03b6\u03bf\u03c5\u03bd \u03c5\u03c0\u03bf\u03c7\u03c1\u03b5\u03c9\u03c4\u03b9\u03ba\u03ac \u03c3\u03c4\u03bf \u03af\u03b4\u03b9\u03bf \u03b5\u03c0\u03af\u03c0\u03b5\u03b4\u03bf \u03b1\u03c6\u03b1\u03af\u03c1\u03b5\u03c3\u03b7\u03c2. \u0395\u03c0\u03bf\u03bc\u03ad\u03bd\u03c9\u03c2, \u03bc\u03b9\u03b1 \u03b5\u03bd\u03b9\u03b1\u03af\u03b1 \u03c0\u03bf\u03bb\u03b9\u03c4\u03b9\u03ba\u03ae granularity \u03bc\u03c0\u03bf\u03c1\u03b5\u03af \u03b5\u03af\u03c4\u03b5 \u03bd\u03b1 \u03c3\u03c0\u03ac\u03c3\u03b5\u03b9 \u03bc\u03b9\u03b1 \u03b5\u03c5\u03c1\u03b5\u03af\u03b1 \u03b1\u03b4\u03c5\u03bd\u03b1\u03bc\u03af\u03b1 \u03c3\u03b5 \u03c0\u03bf\u03bb\u03cd \u03bc\u03b9\u03ba\u03c1\u03ac, \u03b1\u03c3\u03c4\u03b1\u03b8\u03ae \u03ba\u03bf\u03bc\u03bc\u03ac\u03c4\u03b9\u03b1 \u03b5\u03af\u03c4\u03b5 \u03bd\u03b1 \u03ba\u03c1\u03cd\u03c8\u03b5\u03b9 \u03bc\u03b9\u03b1 \u03c3\u03c4\u03b5\u03bd\u03ae \u03b1\u03b4\u03c5\u03bd\u03b1\u03bc\u03af\u03b1 \u03bc\u03ad\u03c3\u03b1 \u03c3\u03b5 \u03ad\u03bd\u03b1\u03bd \u03b3\u03b5\u03bd\u03b9\u03ba\u03cc \u03ba\u03cc\u03bc\u03b2\u03bf.<\/p>\n<h2 id=\"craft-epicheirisiako-ai-workflow\">\u03a0\u03ce\u03c2 \u03bc\u03b5\u03c4\u03b1\u03c6\u03c1\u03ac\u03b6\u03b5\u03c4\u03b1\u03b9 \u03b7 \u03ad\u03c1\u03b5\u03c5\u03bd\u03b1 \u03c3\u03b5 \u03b5\u03c0\u03b9\u03c7\u03b5\u03b9\u03c1\u03b7\u03c3\u03b9\u03b1\u03ba\u03cc AI workflow<\/h2>\n<p>\u0393\u03b9\u03b1 \u03bc\u03b9\u03b1 \u03bf\u03bc\u03ac\u03b4\u03b1 \u03c0\u03bf\u03c5 \u03c7\u03c4\u03af\u03b6\u03b5\u03b9 AI assistant \u03c3\u03b5 e-commerce, customer support, marketing operations \u03ae \u03b5\u03c4\u03b1\u03b9\u03c1\u03b9\u03ba\u03ae \u03b3\u03bd\u03ce\u03c3\u03b7 \u03c3\u03c4\u03bf \u03c0\u03bb\u03b1\u03af\u03c3\u03b9\u03bf \u03ad\u03c1\u03b3\u03c9\u03bd <a href=\"https:\/\/twodots.gr\/aftomatismoi-epicheiriseon-ai\/\">\u0391\u03c5\u03c4\u03bf\u03bc\u03b1\u03c4\u03b9\u03c3\u03bc\u03ce\u03bd \u0395\u03c0\u03b9\u03c7\u03b5\u03b9\u03c1\u03ae\u03c3\u03b5\u03c9\u03bd &amp; AI<\/a>, \u03c4\u03bf CRAFT \u03b4\u03b5\u03bd \u03b5\u03af\u03bd\u03b1\u03b9 \u03ad\u03c4\u03bf\u03b9\u03bc\u03bf plugin. \u0395\u03af\u03bd\u03b1\u03b9 \u03b1\u03c1\u03c7\u03b9\u03c4\u03b5\u03ba\u03c4\u03bf\u03bd\u03b9\u03ba\u03cc \u03c0\u03c1\u03cc\u03c4\u03c5\u03c0\u03bf \u03b3\u03b9\u03b1 \u03c4\u03bf \u03c0\u03ce\u03c2 \u03bc\u03c0\u03bf\u03c1\u03b5\u03af \u03bd\u03b1 \u03c3\u03c5\u03bd\u03b4\u03b5\u03b8\u03b5\u03af \u03c4\u03bf quality assurance \u03bc\u03b5 \u03c4\u03b7 \u03b2\u03b5\u03bb\u03c4\u03af\u03c9\u03c3\u03b7 \u03c4\u03bf\u03c5 \u03c3\u03c5\u03c3\u03c4\u03ae\u03bc\u03b1\u03c4\u03bf\u03c2. \u03a0\u03c1\u03ce\u03c4\u03b1 \u03c7\u03c1\u03b5\u03b9\u03ac\u03b6\u03bf\u03bd\u03c4\u03b1\u03b9 prompts \u03c0\u03bf\u03c5 \u03b1\u03bd\u03c4\u03b9\u03c0\u03c1\u03bf\u03c3\u03c9\u03c0\u03b5\u03cd\u03bf\u03c5\u03bd \u03c0\u03c1\u03b1\u03b3\u03bc\u03b1\u03c4\u03b9\u03ba\u03ad\u03c2 \u03b5\u03c1\u03b3\u03b1\u03c3\u03af\u03b5\u03c2 \u03ba\u03b1\u03b9 \u03ba\u03c1\u03b9\u03c4\u03ae\u03c1\u03b9\u03b1 \u03c0\u03bf\u03c5 \u03c0\u03b5\u03c1\u03b9\u03b3\u03c1\u03ac\u03c6\u03bf\u03c5\u03bd \u03c4\u03b9 \u03c3\u03b7\u03bc\u03b1\u03af\u03bd\u03b5\u03b9 \u03b1\u03c0\u03bf\u03b4\u03b5\u03ba\u03c4\u03ae \u03b1\u03c0\u03ac\u03bd\u03c4\u03b7\u03c3\u03b7. \u039c\u03b5\u03c4\u03ac, \u03bf\u03b9 \u03b1\u03c0\u03bf\u03c4\u03c5\u03c7\u03af\u03b5\u03c2 \u03c0\u03c1\u03ad\u03c0\u03b5\u03b9 \u03bd\u03b1 \u03b1\u03bd\u03b1\u03bb\u03cd\u03bf\u03bd\u03c4\u03b1\u03b9 \u03c3\u03b5 \u03b5\u03c0\u03af\u03c0\u03b5\u03b4\u03bf \u03b1\u03c0\u03b1\u03b9\u03c4\u03ae\u03c3\u03b5\u03c9\u03bd, \u03cc\u03c7\u03b9 \u03bc\u03cc\u03bd\u03bf \u03c3\u03b5 \u03b5\u03c0\u03af\u03c0\u03b5\u03b4\u03bf ticket \u03ae \u03ba\u03b1\u03c4\u03b7\u03b3\u03bf\u03c1\u03af\u03b1\u03c2. \u0391\u03c5\u03c4\u03ae \u03b7 \u03c0\u03b5\u03b9\u03b8\u03b1\u03c1\u03c7\u03af\u03b1 \u03c3\u03c5\u03bc\u03c0\u03bb\u03b7\u03c1\u03ce\u03bd\u03b5\u03b9 \u03ba\u03b1\u03b9 \u03c4\u03b7\u03bd <a href=\"https:\/\/twodots.gr\/ai-agents-poios-ftaiei-otan-aftomatopoiisi-apotygchanei\/\">\u03b1\u03c0\u03cc\u03b4\u03bf\u03c3\u03b7 \u03b5\u03c5\u03b8\u03cd\u03bd\u03b7\u03c2 \u03cc\u03c4\u03b1\u03bd \u03bc\u03b9\u03b1 \u03b1\u03c5\u03c4\u03bf\u03bc\u03b1\u03c4\u03bf\u03c0\u03bf\u03af\u03b7\u03c3\u03b7 \u03b1\u03c0\u03bf\u03c4\u03c5\u03b3\u03c7\u03ac\u03bd\u03b5\u03b9<\/a>, \u03b5\u03c0\u03b5\u03b9\u03b4\u03ae \u03be\u03b5\u03c7\u03c9\u03c1\u03af\u03b6\u03b5\u03b9 \u03c4\u03b7\u03bd \u03b1\u03b4\u03c5\u03bd\u03b1\u03bc\u03af\u03b1 \u03c4\u03bf\u03c5 \u03bc\u03bf\u03bd\u03c4\u03ad\u03bb\u03bf\u03c5 \u03b1\u03c0\u03cc \u03c4\u03bf \u03c0\u03c1\u03cc\u03b2\u03bb\u03b7\u03bc\u03b1 \u03b4\u03b5\u03b4\u03bf\u03bc\u03ad\u03bd\u03c9\u03bd, \u03b5\u03c1\u03b3\u03b1\u03bb\u03b5\u03af\u03bf\u03c5 \u03ae \u03b4\u03b9\u03b1\u03b4\u03b9\u03ba\u03b1\u03c3\u03af\u03b1\u03c2.<\/p>\n<div class=\"td-step-list\">\n<p class=\"td-step-list-title\">\u0388\u03be\u03b9 \u03b2\u03ae\u03bc\u03b1\u03c4\u03b1 \u03b3\u03b9\u03b1 \u03ad\u03bd\u03b1 CRAFT-inspired \u03b5\u03c0\u03b9\u03c7\u03b5\u03b9\u03c1\u03b7\u03c3\u03b9\u03b1\u03ba\u03cc pilot<\/p>\n<ol>\n<li><span class=\"td-step-kicker\">Step 1<\/span><strong>\u039f\u03c1\u03af\u03c3\u03c4\u03b5 \u03c0\u03c1\u03b1\u03b3\u03bc\u03b1\u03c4\u03b9\u03ba\u03ac production tasks \u03ba\u03b1\u03b9 \u03be\u03b5\u03c7\u03c9\u03c1\u03b9\u03c3\u03c4\u03cc held-out set<\/strong>\n<p>\u03a7\u03c1\u03b7\u03c3\u03b9\u03bc\u03bf\u03c0\u03bf\u03b9\u03ae\u03c3\u03c4\u03b5 \u03b1\u03bd\u03c4\u03b9\u03c0\u03c1\u03bf\u03c3\u03c9\u03c0\u03b5\u03c5\u03c4\u03b9\u03ba\u03ac \u03b1\u03b9\u03c4\u03ae\u03bc\u03b1\u03c4\u03b1 \u03b1\u03c0\u03cc e-commerce, support \u03ae operations, \u03b1\u03bb\u03bb\u03ac \u03ba\u03c1\u03b1\u03c4\u03ae\u03c3\u03c4\u03b5 \u03b1\u03bd\u03b5\u03be\u03ac\u03c1\u03c4\u03b7\u03c4\u03b5\u03c2 \u03c0\u03b5\u03c1\u03b9\u03c0\u03c4\u03ce\u03c3\u03b5\u03b9\u03c2 \u03b3\u03b9\u03b1 \u03c4\u03b7\u03bd \u03c4\u03b5\u03bb\u03b9\u03ba\u03ae \u03b5\u03c0\u03b1\u03bb\u03ae\u03b8\u03b5\u03c5\u03c3\u03b7 \u03c4\u03b7\u03c2 \u03bc\u03b5\u03c4\u03b1\u03c6\u03bf\u03c1\u03ac\u03c2.<\/p>\n<\/li>\n<li><span class=\"td-step-kicker\">Step 2<\/span><strong>\u0393\u03c1\u03ac\u03c8\u03c4\u03b5 \u03b1\u03c4\u03bf\u03bc\u03b9\u03ba\u03ac \u03ba\u03b1\u03b9 \u03b5\u03bb\u03ad\u03b3\u03be\u03b9\u03bc\u03b1 rubric criteria<\/strong>\n<p>\u039a\u03ac\u03b8\u03b5 \u03ba\u03c1\u03b9\u03c4\u03ae\u03c1\u03b9\u03bf \u03c0\u03c1\u03ad\u03c0\u03b5\u03b9 \u03bd\u03b1 \u03b5\u03bb\u03ad\u03b3\u03c7\u03b5\u03b9 \u03bc\u03af\u03b1 \u03b1\u03c0\u03b1\u03af\u03c4\u03b7\u03c3\u03b7, \u03cc\u03c0\u03c9\u03c2 \u03b5\u03c6\u03b1\u03c1\u03bc\u03bf\u03b3\u03ae \u03c0\u03bf\u03bb\u03b9\u03c4\u03b9\u03ba\u03ae\u03c2 \u03b5\u03c0\u03b9\u03c3\u03c4\u03c1\u03bf\u03c6\u03ce\u03bd, \u03c4\u03b5\u03ba\u03bc\u03b7\u03c1\u03af\u03c9\u03c3\u03b7 claim, \u03c0\u03bb\u03b7\u03c1\u03cc\u03c4\u03b7\u03c4\u03b1 \u03b2\u03b7\u03bc\u03ac\u03c4\u03c9\u03bd \u03ae \u03c3\u03c9\u03c3\u03c4\u03ae \u03bc\u03bf\u03c1\u03c6\u03ae \u03b1\u03c0\u03bf\u03c4\u03b5\u03bb\u03ad\u03c3\u03bc\u03b1\u03c4\u03bf\u03c2.<\/p>\n<\/li>\n<li><span class=\"td-step-kicker\">Step 3<\/span><strong>\u0392\u03b1\u03b8\u03bc\u03bf\u03bd\u03bf\u03bc\u03ae\u03c3\u03c4\u03b5 judge \u03ba\u03b1\u03b9 \u03b1\u03bd\u03b8\u03c1\u03ce\u03c0\u03b9\u03bd\u03bf review<\/strong>\n<p>\u039c\u03b5\u03c4\u03c1\u03ae\u03c3\u03c4\u03b5 \u03c4\u03b7 \u03c3\u03c5\u03bc\u03c6\u03c9\u03bd\u03af\u03b1 \u03c3\u03b5 \u03b4\u03b5\u03af\u03b3\u03bc\u03b1 \u03b1\u03c0\u03b1\u03bd\u03c4\u03ae\u03c3\u03b5\u03c9\u03bd, \u03b5\u03be\u03b5\u03c4\u03ac\u03c3\u03c4\u03b5 \u03b1\u03bc\u03c6\u03af\u03c3\u03b7\u03bc\u03b1 criteria \u03ba\u03b1\u03b9 \u03bc\u03b7\u03bd \u03b5\u03c0\u03b9\u03c4\u03c1\u03ad\u03c8\u03b5\u03c4\u03b5 \u03c3\u03b5 \u03ad\u03bd\u03b1\u03bd \u03b1\u03c3\u03c4\u03b1\u03b8\u03ae evaluator \u03bd\u03b1 \u03b3\u03af\u03bd\u03b5\u03b9 \u03b7 \u03bc\u03bf\u03bd\u03b1\u03b4\u03b9\u03ba\u03ae \u03c0\u03b7\u03b3\u03ae node scores.<\/p>\n<\/li>\n<li><span class=\"td-step-kicker\">Step 4<\/span><strong>\u039f\u03bc\u03b1\u03b4\u03bf\u03c0\u03bf\u03b9\u03ae\u03c3\u03c4\u03b5 \u03b5\u03c0\u03b1\u03bd\u03b1\u03bb\u03b1\u03bc\u03b2\u03b1\u03bd\u03cc\u03bc\u03b5\u03bd\u03b5\u03c2 \u03b1\u03b4\u03c5\u03bd\u03b1\u03bc\u03af\u03b5\u03c2 \u03bc\u03b5 \u03b5\u03c0\u03b1\u03c1\u03ba\u03ae \u03c5\u03c0\u03bf\u03c3\u03c4\u03ae\u03c1\u03b9\u03be\u03b7<\/strong>\n<p>\u03a3\u03c5\u03bd\u03b4\u03ad\u03c3\u03c4\u03b5 \u03ba\u03ac\u03b8\u03b5 cluster \u03bc\u03b5 \u03c4\u03b1 \u03b1\u03c1\u03c7\u03b9\u03ba\u03ac prompts \u03ba\u03b1\u03b9 failures \u03ba\u03b1\u03b9 \u03b1\u03c0\u03bf\u03c6\u03cd\u03b3\u03b5\u03c4\u03b5 \u03c3\u03c5\u03bc\u03c0\u03b5\u03c1\u03ac\u03c3\u03bc\u03b1\u03c4\u03b1 \u03b1\u03c0\u03cc \u03c0\u03bf\u03bb\u03cd \u03bc\u03b9\u03ba\u03c1\u03bf\u03cd\u03c2 \u03ae \u03b1\u03bb\u03bb\u03b7\u03bb\u03bf\u03b5\u03c0\u03b9\u03ba\u03b1\u03bb\u03c5\u03c0\u03c4\u03cc\u03bc\u03b5\u03bd\u03bf\u03c5\u03c2 \u03ba\u03cc\u03bc\u03b2\u03bf\u03c5\u03c2.<\/p>\n<\/li>\n<li><span class=\"td-step-kicker\">Step 5<\/span><strong>\u0394\u03b7\u03bc\u03b9\u03bf\u03c5\u03c1\u03b3\u03ae\u03c3\u03c4\u03b5 \u03b4\u03b5\u03b4\u03bf\u03bc\u03ad\u03bd\u03b1 \u03bc\u03cc\u03bd\u03bf \u03b3\u03b9\u03b1 \u03c4\u03bf\u03c5\u03c2 \u03b5\u03c0\u03b9\u03bb\u03b5\u03b3\u03bc\u03ad\u03bd\u03bf\u03c5\u03c2 \u03c3\u03c4\u03cc\u03c7\u03bf\u03c5\u03c2<\/strong>\n<p>\u03a3\u03c5\u03bd\u03b4\u03c5\u03ac\u03c3\u03c4\u03b5 \u03b1\u03bd\u03b8\u03c1\u03ce\u03c0\u03b9\u03bd\u03b1 \u03ba\u03b1\u03b9 \u03c3\u03c5\u03bd\u03b8\u03b5\u03c4\u03b9\u03ba\u03ac \u03c0\u03b1\u03c1\u03b1\u03b4\u03b5\u03af\u03b3\u03bc\u03b1\u03c4\u03b1, \u03c6\u03b9\u03bb\u03c4\u03c1\u03ac\u03c1\u03b5\u03c4\u03b5 duplicates \u03ba\u03b1\u03b9 leakage \u03ba\u03b1\u03b9 \u03ba\u03c1\u03b1\u03c4\u03ae\u03c3\u03c4\u03b5 \u03c3\u03c4\u03b1\u03b8\u03b5\u03c1\u03cc budget \u03b3\u03b9\u03b1 \u03b4\u03af\u03ba\u03b1\u03b9\u03b7 \u03c3\u03cd\u03b3\u03ba\u03c1\u03b9\u03c3\u03b7 \u03bc\u03b5 \u03c4\u03bf \u03c5\u03c0\u03ac\u03c1\u03c7\u03bf\u03bd baseline.<\/p>\n<\/li>\n<li><span class=\"td-step-kicker\">Step 6<\/span><strong>\u0395\u03c0\u03b1\u03bd\u03b1\u03be\u03b9\u03bf\u03bb\u03bf\u03b3\u03ae\u03c3\u03c4\u03b5 \u03b1\u03bd\u03ac \u03ba\u03c1\u03af\u03c3\u03b9\u03bc\u03b7 \u03b9\u03ba\u03b1\u03bd\u03cc\u03c4\u03b7\u03c4\u03b1 \u03c0\u03c1\u03b9\u03bd \u03b1\u03c0\u03cc deployment<\/strong>\n<p>\u0395\u03bb\u03ad\u03b3\u03be\u03c4\u03b5 domain average, task-specific acceptance criteria, variance, \u03ba\u03cc\u03c3\u03c4\u03bf\u03c2 \u03ba\u03b1\u03b9 failure regressions. \u0397 \u03c0\u03b1\u03c1\u03b1\u03b3\u03c9\u03b3\u03ae \u03c7\u03c1\u03b5\u03b9\u03ac\u03b6\u03b5\u03c4\u03b1\u03b9 \u03b1\u03bd\u03b8\u03c1\u03ce\u03c0\u03b9\u03bd\u03b1 gates \u03cc\u03c4\u03b1\u03bd \u03bc\u03b9\u03b1 \u03bb\u03ac\u03b8\u03bf\u03c2 \u03b5\u03bd\u03ad\u03c1\u03b3\u03b5\u03b9\u03b1 \u03ad\u03c7\u03b5\u03b9 \u03b5\u03c0\u03b9\u03c7\u03b5\u03b9\u03c1\u03b7\u03c3\u03b9\u03b1\u03ba\u03ae \u03b5\u03c0\u03af\u03c0\u03c4\u03c9\u03c3\u03b7.<\/p>\n<\/li>\n<\/ol>\n<\/div>\n<p>\u03a3\u03b5 \u03ad\u03bd\u03b1 e-commerce use case, \u03bc\u03b9\u03b1 rubric \u03b8\u03b1 \u03bc\u03c0\u03bf\u03c1\u03bf\u03cd\u03c3\u03b5 \u03bd\u03b1 \u03be\u03b5\u03c7\u03c9\u03c1\u03af\u03b6\u03b5\u03b9 \u03c4\u03b7\u03bd \u03bf\u03c1\u03b8\u03ae \u03b1\u03bd\u03ac\u03ba\u03c4\u03b7\u03c3\u03b7 \u03c7\u03b1\u03c1\u03b1\u03ba\u03c4\u03b7\u03c1\u03b9\u03c3\u03c4\u03b9\u03ba\u03ce\u03bd \u03c0\u03c1\u03bf\u03ca\u03cc\u03bd\u03c4\u03bf\u03c2, \u03c4\u03b7 \u03c3\u03c5\u03bc\u03bc\u03cc\u03c1\u03c6\u03c9\u03c3\u03b7 \u03bc\u03b5 \u03c4\u03b7\u03bd \u03c0\u03bf\u03bb\u03b9\u03c4\u03b9\u03ba\u03ae \u03b5\u03c0\u03b9\u03c3\u03c4\u03c1\u03bf\u03c6\u03ce\u03bd, \u03c4\u03b7 \u03c3\u03b1\u03c6\u03ae\u03bd\u03b5\u03b9\u03b1 \u03c4\u03b7\u03c2 \u03b1\u03c0\u03ac\u03bd\u03c4\u03b7\u03c3\u03b7\u03c2 \u03ba\u03b1\u03b9 \u03c4\u03b7\u03bd \u03b1\u03c0\u03bf\u03c6\u03c5\u03b3\u03ae \u03bc\u03b7 \u03c4\u03b5\u03ba\u03bc\u03b7\u03c1\u03b9\u03c9\u03bc\u03ad\u03bd\u03c9\u03bd \u03b9\u03c3\u03c7\u03c5\u03c1\u03b9\u03c3\u03bc\u03ce\u03bd. \u0391\u03bd \u03c4\u03bf \u03bc\u03bf\u03bd\u03c4\u03ad\u03bb\u03bf \u03b1\u03c0\u03bf\u03c4\u03c5\u03b3\u03c7\u03ac\u03bd\u03b5\u03b9 \u03ba\u03c5\u03c1\u03af\u03c9\u03c2 \u03c3\u03c4\u03b7\u03bd \u03b5\u03c6\u03b1\u03c1\u03bc\u03bf\u03b3\u03ae \u03b5\u03be\u03b1\u03b9\u03c1\u03ad\u03c3\u03b5\u03c9\u03bd \u03c4\u03b7\u03c2 \u03c0\u03bf\u03bb\u03b9\u03c4\u03b9\u03ba\u03ae\u03c2 \u03ba\u03b1\u03b9 \u03cc\u03c7\u03b9 \u03c3\u03c4\u03b7\u03bd \u03b1\u03bd\u03ac\u03ba\u03c4\u03b7\u03c3\u03b7 \u03c0\u03c1\u03bf\u03ca\u03cc\u03bd\u03c4\u03c9\u03bd, \u03ad\u03bd\u03b1 \u03b5\u03bd\u03b9\u03b1\u03af\u03bf score \u00abcustomer support\u00bb \u03b8\u03b1 \u03ba\u03b1\u03c4\u03b5\u03cd\u03b8\u03c5\u03bd\u03b5 \u03bb\u03ac\u03b8\u03bf\u03c2 \u03c4\u03b7\u03bd \u03b5\u03c0\u03cc\u03bc\u03b5\u03bd\u03b7 \u03c3\u03c5\u03bb\u03bb\u03bf\u03b3\u03ae \u03b4\u03b5\u03b4\u03bf\u03bc\u03ad\u03bd\u03c9\u03bd. \u0397 \u03bb\u03bf\u03b3\u03b9\u03ba\u03ae \u03c4\u03bf\u03c5 CRAFT \u03b8\u03b1 \u03c3\u03c4\u03cc\u03c7\u03b5\u03c5\u03b5 \u03c4\u03b7 \u03c3\u03c5\u03b3\u03ba\u03b5\u03ba\u03c1\u03b9\u03bc\u03ad\u03bd\u03b7 \u03b9\u03ba\u03b1\u03bd\u03cc\u03c4\u03b7\u03c4\u03b1.<\/p>\n<p>\u0393\u03b9\u03b1 marketing \u03ba\u03b1\u03b9 content teams, \u03b7 \u03af\u03b4\u03b9\u03b1 \u03b1\u03c1\u03c7\u03ae \u03bc\u03c0\u03bf\u03c1\u03b5\u03af \u03bd\u03b1 \u03b5\u03c6\u03b1\u03c1\u03bc\u03bf\u03c3\u03c4\u03b5\u03af \u03c3\u03b5 brand voice, factual grounding, \u03c7\u03c1\u03ae\u03c3\u03b7 \u03b5\u03b3\u03ba\u03b5\u03ba\u03c1\u03b9\u03bc\u03ad\u03bd\u03c9\u03bd claims, \u03c0\u03bb\u03b7\u03c1\u03cc\u03c4\u03b7\u03c4\u03b1 brief \u03ae \u03c0\u03c1\u03bf\u03c3\u03b1\u03c1\u03bc\u03bf\u03b3\u03ae \u03b1\u03bd\u03ac funnel stage. \u03a0\u03c1\u03cc\u03ba\u03b5\u03b9\u03c4\u03b1\u03b9 \u03b3\u03b9\u03b1 \u03c0\u03c1\u03bf\u03c3\u03b5\u03ba\u03c4\u03b9\u03ba\u03ae \u03bc\u03b5\u03c4\u03b1\u03c6\u03bf\u03c1\u03ac \u03c4\u03b7\u03c2 \u03bc\u03b5\u03b8\u03cc\u03b4\u03bf\u03c5, \u03cc\u03c7\u03b9 \u03b3\u03b9\u03b1 \u03b1\u03c0\u03bf\u03c4\u03ad\u03bb\u03b5\u03c3\u03bc\u03b1 \u03c0\u03bf\u03c5 \u03b4\u03bf\u03ba\u03b9\u03bc\u03ac\u03c3\u03c4\u03b7\u03ba\u03b5 \u03c3\u03c4\u03b7\u03bd \u03b5\u03c1\u03b3\u03b1\u03c3\u03af\u03b1: \u03c4\u03b1 \u03c0\u03b5\u03b9\u03c1\u03ac\u03bc\u03b1\u03c4\u03b1 \u03b1\u03c6\u03bf\u03c1\u03bf\u03cd\u03bd \u03b1\u03c0\u03bf\u03ba\u03bb\u03b5\u03b9\u03c3\u03c4\u03b9\u03ba\u03ac finance \u03ba\u03b1\u03b9 legal datasets. \u0397 \u03b5\u03c0\u03b9\u03c7\u03b5\u03b9\u03c1\u03b7\u03c3\u03b9\u03b1\u03ba\u03ae \u03b1\u03be\u03af\u03b1 \u03b2\u03c1\u03af\u03c3\u03ba\u03b5\u03c4\u03b1\u03b9 \u03c3\u03c4\u03bf design pattern, \u03c4\u03bf \u03bf\u03c0\u03bf\u03af\u03bf \u03b8\u03b1 \u03c0\u03c1\u03ad\u03c0\u03b5\u03b9 \u03bd\u03b1 \u03b5\u03c0\u03b9\u03ba\u03c5\u03c1\u03c9\u03b8\u03b5\u03af \u03bc\u03b5 domain-specific held-out tests \u03c0\u03c1\u03b9\u03bd \u03c7\u03c1\u03b7\u03c3\u03b9\u03bc\u03bf\u03c0\u03bf\u03b9\u03b7\u03b8\u03b5\u03af \u03c3\u03b5 \u03c0\u03b1\u03c1\u03b1\u03b3\u03c9\u03b3\u03ae.<\/p>\n<h2 id=\"proypotheseis-oria-craft\">\u039f\u03b9 \u03c0\u03c1\u03bf\u03cb\u03c0\u03bf\u03b8\u03ad\u03c3\u03b5\u03b9\u03c2 \u03ba\u03b1\u03b9 \u03c4\u03b1 \u03cc\u03c1\u03b9\u03b1 \u03c0\u03bf\u03c5 \u03b4\u03b5\u03bd \u03c0\u03c1\u03ad\u03c0\u03b5\u03b9 \u03bd\u03b1 \u03b1\u03b3\u03bd\u03bf\u03b7\u03b8\u03bf\u03cd\u03bd<\/h2>\n<p>\u03a4\u03bf CRAFT \u03b1\u03c0\u03b1\u03b9\u03c4\u03b5\u03af \u03c0\u03bb\u03bf\u03cd\u03c3\u03b9\u03b1 rubric-based \u03b4\u03b5\u03b4\u03bf\u03bc\u03ad\u03bd\u03b1. \u0391\u03bd \u03c4\u03b1 \u03ba\u03c1\u03b9\u03c4\u03ae\u03c1\u03b9\u03b1 \u03b5\u03af\u03bd\u03b1\u03b9 \u03b1\u03c3\u03b1\u03c6\u03ae, \u03b1\u03bb\u03bb\u03b7\u03bb\u03bf\u03b5\u03c0\u03b9\u03ba\u03b1\u03bb\u03c5\u03c0\u03c4\u03cc\u03bc\u03b5\u03bd\u03b1 \u03ae \u03b4\u03b5\u03bd \u03b1\u03bd\u03c4\u03b1\u03bd\u03b1\u03ba\u03bb\u03bf\u03cd\u03bd \u03c4\u03b7\u03bd \u03c0\u03c1\u03b1\u03b3\u03bc\u03b1\u03c4\u03b9\u03ba\u03ae \u03b5\u03c0\u03b9\u03c7\u03b5\u03b9\u03c1\u03b7\u03c3\u03b9\u03b1\u03ba\u03ae \u03c0\u03bf\u03b9\u03cc\u03c4\u03b7\u03c4\u03b1, \u03b7 \u03b9\u03b5\u03c1\u03b1\u03c1\u03c7\u03af\u03b1 \u03b8\u03b1 \u03bf\u03c1\u03b3\u03b1\u03bd\u03ce\u03c3\u03b5\u03b9 \u03bb\u03ac\u03b8\u03bf\u03c2 \u03c3\u03ae\u03bc\u03b1. \u0391\u03c0\u03b1\u03b9\u03c4\u03b5\u03af \u03b5\u03c0\u03af\u03c3\u03b7\u03c2 \u03b1\u03be\u03b9\u03cc\u03c0\u03b9\u03c3\u03c4\u03bf judge, \u03b5\u03c0\u03b5\u03b9\u03b4\u03ae \u03ba\u03ac\u03b8\u03b5 node score \u03ba\u03bb\u03b7\u03c1\u03bf\u03bd\u03bf\u03bc\u03b5\u03af \u03c4\u03b9\u03c2 pass\/fail \u03b1\u03c0\u03bf\u03c6\u03ac\u03c3\u03b5\u03b9\u03c2 \u03c4\u03bf\u03c5. \u0397 \u03bc\u03ad\u03b8\u03bf\u03b4\u03bf\u03c2 \u03c7\u03c1\u03b7\u03c3\u03b9\u03bc\u03bf\u03c0\u03bf\u03b9\u03b5\u03af \u03c0\u03bf\u03bb\u03bb\u03b1\u03c0\u03bb\u03ac LLM calls \u03b3\u03b9\u03b1 capability extraction, naming, consolidation, routing, refinement \u03ba\u03b1\u03b9 grading, \u03ac\u03c1\u03b1 \u03bc\u03b9\u03b1 \u03b5\u03c6\u03b1\u03c1\u03bc\u03bf\u03b3\u03ae \u03c0\u03c1\u03ad\u03c0\u03b5\u03b9 \u03bd\u03b1 \u03b5\u03be\u03b5\u03c4\u03ac\u03c3\u03b5\u03b9 \u03ba\u03cc\u03c3\u03c4\u03bf\u03c2, latency \u03ba\u03b1\u03b9 \u03b1\u03bd\u03b1\u03c0\u03b1\u03c1\u03b1\u03b3\u03c9\u03b3\u03b9\u03bc\u03cc\u03c4\u03b7\u03c4\u03b1.<\/p>\n<div class=\"td-decision-band\">\n<p class=\"td-decision-label\">\u03a4\u03bf go\/no-go \u03ba\u03c1\u03b9\u03c4\u03ae\u03c1\u03b9\u03bf<\/p>\n<p><strong>\u039c\u03b7\u03bd \u03b5\u03ba\u03c0\u03b1\u03b9\u03b4\u03b5\u03cd\u03b5\u03c4\u03b5 \u03c0\u03ac\u03bd\u03c9 \u03c3\u03b5 \u03ad\u03bd\u03b1 capability tree \u03c0\u03c1\u03b9\u03bd \u03b1\u03c0\u03bf\u03b4\u03b5\u03af\u03be\u03b5\u03c4\u03b5 \u03cc\u03c4\u03b9 \u03c4\u03b1 rubrics \u03ba\u03b1\u03b9 \u03bf judge \u03bc\u03b5\u03c4\u03c1\u03bf\u03cd\u03bd \u03c4\u03b7\u03bd \u03c0\u03bf\u03b9\u03cc\u03c4\u03b7\u03c4\u03b1 \u03c0\u03bf\u03c5 \u03c7\u03c1\u03b5\u03b9\u03ac\u03b6\u03b5\u03c4\u03b1\u03b9 \u03c0\u03c1\u03b1\u03b3\u03bc\u03b1\u03c4\u03b9\u03ba\u03ac \u03b7 \u03c0\u03b1\u03c1\u03b1\u03b3\u03c9\u03b3\u03ae.<\/strong><\/p>\n<p>\u0388\u03bd\u03b1 \u03c7\u03b1\u03bc\u03b7\u03bb\u03cc node score \u03b5\u03af\u03bd\u03b1\u03b9 \u03c7\u03c1\u03ae\u03c3\u03b9\u03bc\u03bf training target \u03bc\u03cc\u03bd\u03bf \u03cc\u03c4\u03b1\u03bd \u03c3\u03c4\u03b7\u03c1\u03af\u03b6\u03b5\u03c4\u03b1\u03b9 \u03c3\u03b5 \u03b1\u03c1\u03ba\u03b5\u03c4\u03ac \u03c0\u03b1\u03c1\u03b1\u03b4\u03b5\u03af\u03b3\u03bc\u03b1\u03c4\u03b1, \u03c3\u03c4\u03b1\u03b8\u03b5\u03c1\u03ae \u03b2\u03b1\u03b8\u03bc\u03bf\u03bb\u03cc\u03b3\u03b7\u03c3\u03b7, \u03c3\u03b1\u03c6\u03ae \u03b5\u03c0\u03b9\u03c7\u03b5\u03b9\u03c1\u03b7\u03c3\u03b9\u03b1\u03ba\u03ae \u03b5\u03c0\u03af\u03c0\u03c4\u03c9\u03c3\u03b7 \u03ba\u03b1\u03b9 held-out tests \u03c0\u03bf\u03c5 \u03b4\u03b5\u03bd \u03b1\u03bd\u03b1\u03ba\u03c5\u03ba\u03bb\u03ce\u03bd\u03bf\u03c5\u03bd \u03c4\u03b1 \u03b4\u03b5\u03b4\u03bf\u03bc\u03ad\u03bd\u03b1 \u03b2\u03b5\u03bb\u03c4\u03af\u03c9\u03c3\u03b7\u03c2.<\/p>\n<\/div>\n<p>\u03a4\u03b1 \u03b1\u03c0\u03bf\u03c4\u03b5\u03bb\u03ad\u03c3\u03bc\u03b1\u03c4\u03b1 \u03c0\u03c1\u03bf\u03ad\u03c1\u03c7\u03bf\u03bd\u03c4\u03b1\u03b9 \u03b1\u03c0\u03cc \u03c4\u03ad\u03c3\u03c3\u03b5\u03c1\u03b1 open-source \u03bc\u03bf\u03bd\u03c4\u03ad\u03bb\u03b1, \u03b4\u03cd\u03bf \u03b5\u03c0\u03b1\u03b3\u03b3\u03b5\u03bb\u03bc\u03b1\u03c4\u03b9\u03ba\u03ac domains \u03ba\u03b1\u03b9 \u03c3\u03c5\u03b3\u03ba\u03b5\u03ba\u03c1\u03b9\u03bc\u03ad\u03bd\u03b7 \u03c3\u03c5\u03bd\u03b8\u03b5\u03c4\u03b9\u03ba\u03ae \u03b4\u03b9\u03b1\u03b4\u03b9\u03ba\u03b1\u03c3\u03af\u03b1. \u0394\u03b5\u03bd \u03b1\u03c0\u03bf\u03b4\u03b5\u03b9\u03ba\u03bd\u03cd\u03bf\u03c5\u03bd \u03cc\u03c4\u03b9 \u03ba\u03ac\u03b8\u03b5 \u03bc\u03bf\u03bd\u03c4\u03ad\u03bb\u03bf, \u03ba\u03ac\u03b8\u03b5 \u03ba\u03bb\u03ac\u03b4\u03bf\u03c2 \u03ae \u03ba\u03ac\u03b8\u03b5 generator \u03b8\u03b1 \u03ad\u03c7\u03b5\u03b9 \u03c4\u03bf \u03af\u03b4\u03b9\u03bf \u03cc\u03c6\u03b5\u03bb\u03bf\u03c2. \u0395\u03c0\u03af\u03c3\u03b7\u03c2, \u03bf\u03b9 benchmark-level \u03b5\u03c0\u03b9\u03b4\u03cc\u03c3\u03b5\u03b9\u03c2 \u03ae\u03c4\u03b1\u03bd \u03b5\u03c4\u03b5\u03c1\u03bf\u03b3\u03b5\u03bd\u03b5\u03af\u03c2. \u0393\u03b9\u03b1 high-stakes \u03c7\u03c1\u03ae\u03c3\u03b7, \u03bf \u03c3\u03c5\u03bd\u03bf\u03bb\u03b9\u03ba\u03cc\u03c2 \u03bc\u03ad\u03c3\u03bf\u03c2 \u03cc\u03c1\u03bf\u03c2 \u03c0\u03c1\u03ad\u03c0\u03b5\u03b9 \u03bd\u03b1 \u03c3\u03c5\u03bc\u03c0\u03bb\u03b7\u03c1\u03ce\u03bd\u03b5\u03c4\u03b1\u03b9 \u03bc\u03b5 task-specific acceptance criteria, \u03b1\u03bd\u03b8\u03c1\u03ce\u03c0\u03b9\u03bd\u03bf review \u03ba\u03b1\u03b9 \u03b5\u03bb\u03ad\u03b3\u03c7\u03bf\u03c5\u03c2 \u03c3\u03b5 \u03c0\u03c1\u03b1\u03b3\u03bc\u03b1\u03c4\u03b9\u03ba\u03ad\u03c2 \u03ba\u03b1\u03c4\u03b1\u03bd\u03bf\u03bc\u03ad\u03c2 \u03b4\u03b5\u03b4\u03bf\u03bc\u03ad\u03bd\u03c9\u03bd.<\/p>\n<p>\u03a4\u03ad\u03bb\u03bf\u03c2, \u03c4\u03bf CRAFT \u03b4\u03b5\u03bd \u03b5\u03af\u03bd\u03b1\u03b9 \u03bd\u03ad\u03b1 \u03bc\u03ad\u03b8\u03bf\u03b4\u03bf\u03c2 fine-tuning. \u039f\u03b9 \u03c3\u03c5\u03b3\u03b3\u03c1\u03b1\u03c6\u03b5\u03af\u03c2 \u03c4\u03bf \u03bf\u03c1\u03af\u03b6\u03bf\u03c5\u03bd \u03c9\u03c2 \u03c4\u03c1\u03cc\u03c0\u03bf \u03c0\u03b1\u03c1\u03b1\u03b3\u03c9\u03b3\u03ae\u03c2 model-specific training targets \u03b1\u03c0\u03cc \u03bc\u03b5\u03c4\u03c1\u03b7\u03bc\u03ad\u03bd\u03b5\u03c2 \u03b1\u03c0\u03bf\u03c4\u03c5\u03c7\u03af\u03b5\u03c2. \u0397 \u03c0\u03bf\u03b9\u03cc\u03c4\u03b7\u03c4\u03b1 \u03c4\u03c9\u03bd \u03c4\u03b5\u03bb\u03b9\u03ba\u03ce\u03bd \u03b4\u03b5\u03b4\u03bf\u03bc\u03ad\u03bd\u03c9\u03bd, \u03c4\u03bf\u03c5 teacher model \u03ba\u03b1\u03b9 \u03c4\u03b7\u03c2 \u03b5\u03ba\u03c0\u03b1\u03af\u03b4\u03b5\u03c5\u03c3\u03b7\u03c2 \u03b5\u03be\u03b1\u03ba\u03bf\u03bb\u03bf\u03c5\u03b8\u03b5\u03af \u03bd\u03b1 \u03ad\u03c7\u03b5\u03b9 \u03ba\u03b1\u03b8\u03bf\u03c1\u03b9\u03c3\u03c4\u03b9\u03ba\u03cc \u03c1\u03cc\u03bb\u03bf. \u0397 \u03bc\u03ad\u03b8\u03bf\u03b4\u03bf\u03c2 \u03b2\u03b5\u03bb\u03c4\u03b9\u03ce\u03bd\u03b5\u03b9 \u03c4\u03bf \u03b5\u03c1\u03ce\u03c4\u03b7\u03bc\u03b1 \u00ab\u03c3\u03b5 \u03c4\u03b9 \u03bd\u03b1 \u03b5\u03ba\u03c0\u03b1\u03b9\u03b4\u03b5\u03cd\u03c3\u03bf\u03c5\u03bc\u03b5 \u03bc\u03b5\u03c4\u03ac;\u00bb, \u03b4\u03b5\u03bd \u03b1\u03c5\u03c4\u03bf\u03bc\u03b1\u03c4\u03bf\u03c0\u03bf\u03b9\u03b5\u03af \u03bf\u03bb\u03cc\u03ba\u03bb\u03b7\u03c1\u03bf \u03c4\u03bf\u03bd \u03ba\u03cd\u03ba\u03bb\u03bf \u03bc\u03b5 \u03b5\u03b3\u03b3\u03c5\u03b7\u03bc\u03ad\u03bd\u03bf \u03b1\u03c0\u03bf\u03c4\u03ad\u03bb\u03b5\u03c3\u03bc\u03b1.<\/p>\n<h2 id=\"dashboard-systima-synechous-mathisis\">\u0391\u03c0\u03cc dashboard \u03b5\u03c0\u03af\u03b4\u03bf\u03c3\u03b7\u03c2 \u03c3\u03b5 \u03c3\u03cd\u03c3\u03c4\u03b7\u03bc\u03b1 \u03c3\u03c5\u03bd\u03b5\u03c7\u03bf\u03cd\u03c2 \u03bc\u03ac\u03b8\u03b7\u03c3\u03b7\u03c2<\/h2>\n<p>\u0397 \u03bf\u03c5\u03c3\u03af\u03b1 \u03c4\u03b7\u03c2 \u03b5\u03c1\u03b3\u03b1\u03c3\u03af\u03b1\u03c2 \u03b5\u03af\u03bd\u03b1\u03b9 \u03bf\u03c1\u03b3\u03b1\u03bd\u03c9\u03c4\u03b9\u03ba\u03ae \u03cc\u03c3\u03bf \u03ba\u03b1\u03b9 \u03c4\u03b5\u03c7\u03bd\u03b9\u03ba\u03ae. \u03a0\u03bf\u03bb\u03bb\u03ad\u03c2 \u03bf\u03bc\u03ac\u03b4\u03b5\u03c2 \u03b1\u03be\u03b9\u03bf\u03bb\u03bf\u03b3\u03bf\u03cd\u03bd \u03ad\u03bd\u03b1 AI \u03c3\u03cd\u03c3\u03c4\u03b7\u03bc\u03b1, \u03ba\u03b1\u03c4\u03b1\u03b3\u03c1\u03ac\u03c6\u03bf\u03c5\u03bd \u03ad\u03bd\u03b1 aggregate score \u03ba\u03b1\u03b9 \u03ad\u03c0\u03b5\u03b9\u03c4\u03b1 \u03c3\u03c5\u03bb\u03bb\u03ad\u03b3\u03bf\u03c5\u03bd \u03c0\u03b5\u03c1\u03b9\u03c3\u03c3\u03cc\u03c4\u03b5\u03c1\u03b1 \u03b4\u03b5\u03b4\u03bf\u03bc\u03ad\u03bd\u03b1 \u03bc\u03b5 \u03b2\u03ac\u03c3\u03b7 \u03b3\u03b5\u03bd\u03b9\u03ba\u03ad\u03c2 \u03ba\u03b1\u03c4\u03b7\u03b3\u03bf\u03c1\u03af\u03b5\u03c2 \u03ae \u03c4\u03b1 \u03c0\u03b9\u03bf \u03b5\u03bc\u03c6\u03b1\u03bd\u03ae failures. \u03a4\u03bf CRAFT \u03c0\u03c1\u03bf\u03c4\u03b5\u03af\u03bd\u03b5\u03b9 \u03bc\u03b9\u03b1 \u03b1\u03c5\u03c3\u03c4\u03b7\u03c1\u03cc\u03c4\u03b5\u03c1\u03b7 \u03b1\u03bb\u03c5\u03c3\u03af\u03b4\u03b1: \u03c1\u03b7\u03c4\u03ac \u03ba\u03c1\u03b9\u03c4\u03ae\u03c1\u03b9\u03b1, \u03c3\u03b7\u03bc\u03b1\u03c3\u03b9\u03bf\u03bb\u03bf\u03b3\u03b9\u03ba\u03ae \u03bf\u03c1\u03b3\u03ac\u03bd\u03c9\u03c3\u03b7 \u03c4\u03c9\u03bd \u03b9\u03ba\u03b1\u03bd\u03bf\u03c4\u03ae\u03c4\u03c9\u03bd, model-specific scoring, \u03b5\u03c0\u03b9\u03bb\u03bf\u03b3\u03ae \u03c4\u03b7\u03c2 \u03c3\u03c9\u03c3\u03c4\u03ae\u03c2 \u03ba\u03bb\u03af\u03bc\u03b1\u03ba\u03b1\u03c2 \u03b1\u03b4\u03c5\u03bd\u03b1\u03bc\u03af\u03b1\u03c2 \u03ba\u03b1\u03b9 \u03c0\u03b1\u03c1\u03b1\u03b3\u03c9\u03b3\u03ae \u03b4\u03b5\u03b4\u03bf\u03bc\u03ad\u03bd\u03c9\u03bd \u03c0\u03bf\u03c5 \u03c3\u03c4\u03bf\u03c7\u03b5\u03cd\u03bf\u03c5\u03bd \u03b1\u03ba\u03c1\u03b9\u03b2\u03ce\u03c2 \u03b5\u03ba\u03b5\u03af.<\/p>\n<p>\u0393\u03b9\u03b1 \u03c4\u03bf\u03c5\u03c2 \u03b5\u03c0\u03b1\u03b3\u03b3\u03b5\u03bb\u03bc\u03b1\u03c4\u03af\u03b5\u03c2 \u03ba\u03b1\u03b9 \u03c4\u03bf\u03c5\u03c2 \u03b9\u03b4\u03b9\u03bf\u03ba\u03c4\u03ae\u03c4\u03b5\u03c2 \u03b5\u03c0\u03b9\u03c7\u03b5\u03b9\u03c1\u03ae\u03c3\u03b5\u03c9\u03bd, \u03c4\u03bf \u03c3\u03c5\u03bc\u03c0\u03ad\u03c1\u03b1\u03c3\u03bc\u03b1 \u03b4\u03b5\u03bd \u03b5\u03af\u03bd\u03b1\u03b9 \u03cc\u03c4\u03b9 \u03c7\u03c1\u03b5\u03b9\u03ac\u03b6\u03bf\u03bd\u03c4\u03b1\u03b9 \u03b1\u03c0\u03b1\u03c1\u03b1\u03af\u03c4\u03b7\u03c4\u03b1 \u03ad\u03bd\u03b1 \u03b5\u03be\u03b1\u03b5\u03c0\u03af\u03c0\u03b5\u03b4\u03bf capability tree \u03b1\u03c0\u03cc \u03c4\u03b7\u03bd \u03c0\u03c1\u03ce\u03c4\u03b7 \u03b7\u03bc\u03ad\u03c1\u03b1. \u0395\u03af\u03bd\u03b1\u03b9 \u03cc\u03c4\u03b9 \u03ad\u03bd\u03b1 AI quality program \u03b1\u03c0\u03bf\u03b4\u03af\u03b4\u03b5\u03b9 \u03c0\u03b5\u03c1\u03b9\u03c3\u03c3\u03cc\u03c4\u03b5\u03c1\u03bf \u03cc\u03c4\u03b1\u03bd \u03ba\u03ac\u03b8\u03b5 \u03b1\u03c0\u03bf\u03c4\u03c5\u03c7\u03af\u03b1 \u03bc\u03b5\u03c4\u03b1\u03c6\u03c1\u03ac\u03b6\u03b5\u03c4\u03b1\u03b9 \u03c3\u03b5 \u03c3\u03c5\u03b3\u03ba\u03b5\u03ba\u03c1\u03b9\u03bc\u03ad\u03bd\u03b7, \u03b5\u03c0\u03b1\u03bd\u03b1\u03c7\u03c1\u03b7\u03c3\u03b9\u03bc\u03bf\u03c0\u03bf\u03b9\u03ae\u03c3\u03b9\u03bc\u03b7 \u03b1\u03c0\u03b1\u03af\u03c4\u03b7\u03c3\u03b7. \u0391\u03ba\u03cc\u03bc\u03b7 \u03ba\u03b1\u03b9 \u03bc\u03b9\u03b1 \u03bc\u03b9\u03ba\u03c1\u03cc\u03c4\u03b5\u03c1\u03b7 \u03b5\u03c6\u03b1\u03c1\u03bc\u03bf\u03b3\u03ae \u03bc\u03c0\u03bf\u03c1\u03b5\u03af \u03bd\u03b1 \u03be\u03b5\u03ba\u03b9\u03bd\u03ae\u03c3\u03b5\u03b9 \u03bc\u03b5 \u03ba\u03b1\u03bb\u03ac \u03ba\u03c1\u03b9\u03c4\u03ae\u03c1\u03b9\u03b1, \u03b4\u03b9\u03b1\u03c7\u03c9\u03c1\u03b9\u03c3\u03bc\u03cc \u03c4\u03cd\u03c0\u03c9\u03bd \u03b1\u03c0\u03bf\u03c4\u03c5\u03c7\u03af\u03b1\u03c2 \u03ba\u03b1\u03b9 held-out tests \u03c0\u03bf\u03c5 \u03b4\u03b5\u03bd \u03b1\u03bd\u03b1\u03ba\u03c5\u03ba\u03bb\u03ce\u03bd\u03bf\u03c5\u03bd \u03c4\u03b1 \u03b4\u03b5\u03b4\u03bf\u03bc\u03ad\u03bd\u03b1 \u03b2\u03b5\u03bb\u03c4\u03af\u03c9\u03c3\u03b7\u03c2.<\/p>\n<p>\u03a4\u03bf CRAFT \u03b4\u03b5\u03af\u03c7\u03bd\u03b5\u03b9 \u03cc\u03c4\u03b9 \u03b7 \u03b1\u03be\u03b9\u03bf\u03bb\u03cc\u03b3\u03b7\u03c3\u03b7 \u03bc\u03c0\u03bf\u03c1\u03b5\u03af \u03bd\u03b1 \u03b3\u03af\u03bd\u03b5\u03b9 \u03b5\u03bd\u03b5\u03c1\u03b3\u03cc \u03bc\u03ad\u03c1\u03bf\u03c2 \u03c4\u03bf\u03c5 post-training loop. \u0391\u03bd\u03c4\u03af \u03bd\u03b1 \u03b1\u03c0\u03b1\u03bd\u03c4\u03ac \u03bc\u03cc\u03bd\u03bf \u00ab\u03c0\u03cc\u03c3\u03bf \u03ba\u03b1\u03bb\u03cc \u03b5\u03af\u03bd\u03b1\u03b9 \u03c4\u03bf \u03bc\u03bf\u03bd\u03c4\u03ad\u03bb\u03bf;\u00bb, \u03bc\u03c0\u03bf\u03c1\u03b5\u03af \u03bd\u03b1 \u03b1\u03c0\u03b1\u03bd\u03c4\u03ac \u00ab\u03c0\u03bf\u03b9\u03b1 \u03b9\u03ba\u03b1\u03bd\u03cc\u03c4\u03b7\u03c4\u03b1 \u03bb\u03b5\u03af\u03c0\u03b5\u03b9, \u03c3\u03b5 \u03c0\u03bf\u03b9\u03bf \u03b5\u03c0\u03af\u03c0\u03b5\u03b4\u03bf \u03b5\u03bc\u03c6\u03b1\u03bd\u03af\u03b6\u03b5\u03c4\u03b1\u03b9 \u03ba\u03b1\u03b9 \u03c0\u03bf\u03b9\u03b1 \u03b4\u03b5\u03b4\u03bf\u03bc\u03ad\u03bd\u03b1 \u03b1\u03be\u03af\u03b6\u03b5\u03b9 \u03bd\u03b1 \u03b4\u03b7\u03bc\u03b9\u03bf\u03c5\u03c1\u03b3\u03ae\u03c3\u03bf\u03c5\u03bc\u03b5 \u03c3\u03c4\u03b7 \u03c3\u03c5\u03bd\u03ad\u03c7\u03b5\u03b9\u03b1;\u00bb. \u0391\u03c5\u03c4\u03ae \u03b7 \u03bc\u03b5\u03c4\u03ac\u03b2\u03b1\u03c3\u03b7 \u03b5\u03af\u03bd\u03b1\u03b9 \u03c0\u03b9\u03b8\u03b1\u03bd\u03cc \u03bd\u03b1 \u03be\u03b5\u03c7\u03c9\u03c1\u03af\u03c3\u03b5\u03b9 \u03c4\u03b1 AI projects \u03c0\u03bf\u03c5 \u03b1\u03c0\u03bb\u03ce\u03c2 \u03c0\u03b1\u03c1\u03b1\u03ba\u03bf\u03bb\u03bf\u03c5\u03b8\u03bf\u03cd\u03bd \u03b4\u03b5\u03af\u03ba\u03c4\u03b5\u03c2 \u03b1\u03c0\u03cc \u03b5\u03ba\u03b5\u03af\u03bd\u03b1 \u03c0\u03bf\u03c5 \u03bc\u03b1\u03b8\u03b1\u03af\u03bd\u03bf\u03c5\u03bd \u03c3\u03c5\u03c3\u03c4\u03b7\u03bc\u03b1\u03c4\u03b9\u03ba\u03ac \u03b1\u03c0\u03cc \u03c4\u03b9\u03c2 \u03b1\u03c0\u03bf\u03c4\u03c5\u03c7\u03af\u03b5\u03c2 \u03c4\u03bf\u03c5\u03c2.<\/p>\n<section class=\"td-service-cta\">\n<div class=\"td-service-cta-content\">\n<p class=\"td-service-cta-eyebrow\">Business Automation &amp; AI by TWO DOTS<\/p>\n<p class=\"td-service-cta-title\">\u039c\u03b5\u03c4\u03b1\u03c4\u03c1\u03ad\u03c8\u03c4\u03b5 \u03c4\u03b1 failures \u03c4\u03bf\u03c5 AI \u03c3\u03c5\u03c3\u03c4\u03ae\u03bc\u03b1\u03c4\u03cc\u03c2 \u03c3\u03b1\u03c2 \u03c3\u03b5 \u03c3\u03c5\u03b3\u03ba\u03b5\u03ba\u03c1\u03b9\u03bc\u03ad\u03bd\u03bf backlog \u03b2\u03b5\u03bb\u03c4\u03af\u03c9\u03c3\u03b7\u03c2.<\/p>\n<p>\u0397 TWO DOTS \u03c3\u03c5\u03bd\u03b4\u03ad\u03b5\u03b9 production use cases, rubrics, \u03b1\u03be\u03b9\u03bf\u03bb\u03cc\u03b3\u03b7\u03c3\u03b7, \u03b4\u03b5\u03b4\u03bf\u03bc\u03ad\u03bd\u03b1, \u03b1\u03bd\u03b8\u03c1\u03ce\u03c0\u03b9\u03bd\u03b1 approval gates \u03ba\u03b1\u03b9 \u03bc\u03b5\u03c4\u03c1\u03ae\u03c3\u03b5\u03b9\u03c2, \u03ce\u03c3\u03c4\u03b5 \u03ba\u03ac\u03b8\u03b5 \u03b1\u03bb\u03bb\u03b1\u03b3\u03ae \u03bd\u03b1 \u03b1\u03c0\u03b1\u03bd\u03c4\u03ac \u03c3\u03b5 \u03c0\u03c1\u03b1\u03b3\u03bc\u03b1\u03c4\u03b9\u03ba\u03ae \u03b5\u03c0\u03b9\u03c7\u03b5\u03b9\u03c1\u03b7\u03c3\u03b9\u03b1\u03ba\u03ae \u03b1\u03b4\u03c5\u03bd\u03b1\u03bc\u03af\u03b1 \u03ba\u03b1\u03b9 \u03cc\u03c7\u03b9 \u03bc\u03cc\u03bd\u03bf \u03c3\u03b5 \u03ad\u03bd\u03b1\u03bd \u03b3\u03b5\u03bd\u03b9\u03ba\u03cc \u03b4\u03b5\u03af\u03ba\u03c4\u03b7.<\/p>\n<div class=\"td-service-cta-actions\"><a class=\"td-service-cta-button\" href=\"https:\/\/twodots.gr\/aftomatismoi-epicheiriseon-ai\/\">\u03a3\u03c7\u03b5\u03b4\u03b9\u03ac\u03c3\u03c4\u03b5 \u03c4\u03bf AI quality workflow \u03bc\u03b5 \u03c4\u03b7\u03bd TWO DOTS<\/a><\/div>\n<\/div>\n<\/section>\n<section id=\"sychnes-erotiseis\" class=\"td-faq-section\">\n<div class=\"td-faq\">\n<p class=\"td-faq-heading\">Frequently Asked Questions (FAQs)<\/p>\n<details class=\"td-faq-item\">\n<summary class=\"td-faq-title\">\u03a4\u03b9 \u03b5\u03af\u03bd\u03b1\u03b9 \u03c4\u03bf CRAFT;<\/summary>\n<div class=\"td-faq-content\">\n<p>\u0395\u03af\u03bd\u03b1\u03b9 \u03bc\u03ad\u03b8\u03bf\u03b4\u03bf\u03c2 \u03c0\u03bf\u03c5 \u03bc\u03b5\u03c4\u03b1\u03c4\u03c1\u03ad\u03c0\u03b5\u03b9 rubric-based evaluation data \u03c3\u03b5 model-specific \u03b4\u03b9\u03ac\u03b3\u03bd\u03c9\u03c3\u03b7 \u03b1\u03b4\u03cd\u03bd\u03b1\u03bc\u03c9\u03bd \u03b9\u03ba\u03b1\u03bd\u03bf\u03c4\u03ae\u03c4\u03c9\u03bd \u03ba\u03b1\u03b9 \u03c7\u03c1\u03b7\u03c3\u03b9\u03bc\u03bf\u03c0\u03bf\u03b9\u03b5\u03af \u03b1\u03c5\u03c4\u03ae \u03c4\u03b7 \u03b4\u03b9\u03ac\u03b3\u03bd\u03c9\u03c3\u03b7 \u03b3\u03b9\u03b1 \u03bd\u03b1 \u03c3\u03c4\u03bf\u03c7\u03b5\u03cd\u03c3\u03b5\u03b9 \u03c4\u03b7 \u03b4\u03b7\u03bc\u03b9\u03bf\u03c5\u03c1\u03b3\u03af\u03b1 supervised fine-tuning \u03b4\u03b5\u03b4\u03bf\u03bc\u03ad\u03bd\u03c9\u03bd.<\/p>\n<\/div>\n<\/details>\n<details class=\"td-faq-item\">\n<summary class=\"td-faq-title\">\u03a0\u03bf\u03b9\u03b1 \u03b5\u03af\u03bd\u03b1\u03b9 \u03b7 \u03b2\u03b1\u03c3\u03b9\u03ba\u03ae \u03b4\u03b9\u03b1\u03c6\u03bf\u03c1\u03ac \u03b1\u03c0\u03cc \u03ad\u03bd\u03b1 \u03c3\u03c5\u03bd\u03b7\u03b8\u03b9\u03c3\u03bc\u03ad\u03bd\u03bf benchmark;<\/summary>\n<div class=\"td-faq-content\">\n<p>\u0388\u03bd\u03b1 benchmark \u03c3\u03c5\u03bd\u03ae\u03b8\u03c9\u03c2 \u03b1\u03bd\u03b1\u03c6\u03ad\u03c1\u03b5\u03b9 \u03b5\u03c0\u03af\u03b4\u03bf\u03c3\u03b7 \u03b1\u03bd\u03ac task \u03ae \u03ba\u03b1\u03c4\u03b7\u03b3\u03bf\u03c1\u03af\u03b1. \u03a4\u03bf CRAFT \u03b1\u03bd\u03b1\u03bb\u03cd\u03b5\u03b9 \u03ba\u03ac\u03b8\u03b5 \u03b5\u03c0\u03b9\u03bc\u03ad\u03c1\u03bf\u03c5\u03c2 \u03ba\u03c1\u03b9\u03c4\u03ae\u03c1\u03b9\u03bf \u03c9\u03c2 capability probe, \u03ce\u03c3\u03c4\u03b5 \u03bd\u03b1 \u03b5\u03bd\u03c4\u03bf\u03c0\u03af\u03c3\u03b5\u03b9 \u03c0\u03bf\u03b9\u03b1 \u03b1\u03c0\u03b1\u03af\u03c4\u03b7\u03c3\u03b7 \u03c4\u03b7\u03c2 \u03b1\u03c0\u03ac\u03bd\u03c4\u03b7\u03c3\u03b7\u03c2 \u03b1\u03c0\u03bf\u03c4\u03c5\u03b3\u03c7\u03ac\u03bd\u03b5\u03b9 \u03c3\u03c5\u03c3\u03c4\u03b7\u03bc\u03b1\u03c4\u03b9\u03ba\u03ac.<\/p>\n<\/div>\n<\/details>\n<details class=\"td-faq-item\">\n<summary class=\"td-faq-title\">\u0393\u03b9\u03b1\u03c4\u03af \u03b4\u03b5\u03bd \u03bf\u03bc\u03b1\u03b4\u03bf\u03c0\u03bf\u03b9\u03b5\u03af \u03b1\u03c0\u03bb\u03ce\u03c2 \u03bf\u03bb\u03cc\u03ba\u03bb\u03b7\u03c1\u03b1 prompts;<\/summary>\n<div class=\"td-faq-content\">\n<p>\u0388\u03bd\u03b1 prompt \u03bc\u03c0\u03bf\u03c1\u03b5\u03af \u03bd\u03b1 \u03b5\u03bb\u03ad\u03b3\u03c7\u03b5\u03b9 \u03c0\u03bf\u03bb\u03bb\u03ad\u03c2 \u03b4\u03b9\u03b1\u03c6\u03bf\u03c1\u03b5\u03c4\u03b9\u03ba\u03ad\u03c2 \u03b9\u03ba\u03b1\u03bd\u03cc\u03c4\u03b7\u03c4\u03b5\u03c2. \u0397 \u03bf\u03bc\u03b1\u03b4\u03bf\u03c0\u03bf\u03af\u03b7\u03c3\u03b7 \u03b1\u03bd\u03ac criterion \u03b5\u03c0\u03b9\u03c4\u03c1\u03ad\u03c0\u03b5\u03b9 \u03bd\u03b1 \u03b4\u03b9\u03b1\u03c7\u03c9\u03c1\u03b9\u03c3\u03c4\u03b5\u03af, \u03b3\u03b9\u03b1 \u03c0\u03b1\u03c1\u03ac\u03b4\u03b5\u03b9\u03b3\u03bc\u03b1, \u03b7 \u03c3\u03c9\u03c3\u03c4\u03ae \u03b1\u03bd\u03ac\u03ba\u03bb\u03b7\u03c3\u03b7 \u03c0\u03bb\u03b7\u03c1\u03bf\u03c6\u03bf\u03c1\u03af\u03b1\u03c2 \u03b1\u03c0\u03cc \u03c4\u03b7\u03bd \u03b5\u03c6\u03b1\u03c1\u03bc\u03bf\u03b3\u03ae \u03b5\u03bd\u03cc\u03c2 \u03ba\u03b1\u03bd\u03cc\u03bd\u03b1 \u03ae \u03c4\u03b7 \u03bc\u03bf\u03c1\u03c6\u03bf\u03c0\u03bf\u03af\u03b7\u03c3\u03b7 \u03c4\u03bf\u03c5 \u03c4\u03b5\u03bb\u03b9\u03ba\u03bf\u03cd \u03b1\u03c0\u03bf\u03c4\u03b5\u03bb\u03ad\u03c3\u03bc\u03b1\u03c4\u03bf\u03c2.<\/p>\n<\/div>\n<\/details>\n<details class=\"td-faq-item\">\n<summary class=\"td-faq-title\">\u03a0\u03ce\u03c2 \u03b5\u03c0\u03b9\u03bb\u03ad\u03b3\u03b5\u03b9 \u03c4\u03bf CRAFT \u03c4\u03bf \u03c3\u03c9\u03c3\u03c4\u03cc \u03b5\u03c0\u03af\u03c0\u03b5\u03b4\u03bf \u03b1\u03b4\u03c5\u03bd\u03b1\u03bc\u03af\u03b1\u03c2;<\/summary>\n<div class=\"td-faq-content\">\n<p>\u03a7\u03c1\u03b7\u03c3\u03b9\u03bc\u03bf\u03c0\u03bf\u03b9\u03b5\u03af top-down \u03b4\u03c5\u03bd\u03b1\u03bc\u03b9\u03ba\u03ae \u03b1\u03bd\u03b1\u03b6\u03ae\u03c4\u03b7\u03c3\u03b7 \u03c3\u03c4\u03bf capability tree. \u0395\u03c0\u03b9\u03bb\u03ad\u03b3\u03b5\u03b9 \u03ba\u03cc\u03bc\u03b2\u03bf\u03c5\u03c2 \u03bc\u03b5 \u03b5\u03c0\u03b1\u03c1\u03ba\u03ae \u03c5\u03c0\u03bf\u03c3\u03c4\u03ae\u03c1\u03b9\u03be\u03b7 \u03ba\u03b1\u03b9 \u03c7\u03b1\u03bc\u03b7\u03bb\u03cc pass rate, \u03b5\u03bd\u03ce \u03b1\u03c0\u03bf\u03ba\u03bb\u03b5\u03af\u03b5\u03b9 \u03c4\u03bf\u03c5\u03c2 \u03c0\u03c1\u03bf\u03b3\u03cc\u03bd\u03bf\u03c5\u03c2 \u03ba\u03b1\u03b9 \u03c4\u03bf\u03c5\u03c2 \u03b1\u03c0\u03bf\u03b3\u03cc\u03bd\u03bf\u03c5\u03c2 \u03b5\u03bd\u03cc\u03c2 \u03ae\u03b4\u03b7 \u03b5\u03c0\u03b9\u03bb\u03b5\u03b3\u03bc\u03ad\u03bd\u03bf\u03c5 \u03ba\u03cc\u03bc\u03b2\u03bf\u03c5 \u03b3\u03b9\u03b1 \u03bd\u03b1 \u03b1\u03c0\u03bf\u03c6\u03cd\u03b3\u03b5\u03b9 \u03b5\u03c0\u03b9\u03ba\u03b1\u03bb\u03cd\u03c8\u03b5\u03b9\u03c2.<\/p>\n<\/div>\n<\/details>\n<details class=\"td-faq-item\">\n<summary class=\"td-faq-title\">\u03a3\u03b5 \u03c0\u03bf\u03b9\u03b1 \u03bc\u03bf\u03bd\u03c4\u03ad\u03bb\u03b1 \u03ba\u03b1\u03b9 domains \u03b4\u03bf\u03ba\u03b9\u03bc\u03ac\u03c3\u03c4\u03b7\u03ba\u03b5;<\/summary>\n<div class=\"td-faq-content\">\n<p>\u0394\u03bf\u03ba\u03b9\u03bc\u03ac\u03c3\u03c4\u03b7\u03ba\u03b5 \u03c3\u03c4\u03b1 Qwen3-4B, Qwen3-8B, Gemma-3-4B \u03ba\u03b1\u03b9 Llama-3.1-8B-Instruct, \u03bc\u03b5 \u03b4\u03b5\u03b4\u03bf\u03bc\u03ad\u03bd\u03b1 \u03ba\u03b1\u03b9 benchmarks \u03b1\u03c0\u03cc \u03c4\u03bf\u03c5\u03c2 \u03c7\u03c1\u03b7\u03bc\u03b1\u03c4\u03bf\u03bf\u03b9\u03ba\u03bf\u03bd\u03bf\u03bc\u03b9\u03ba\u03bf\u03cd\u03c2 \u03ba\u03b1\u03b9 \u03bd\u03bf\u03bc\u03b9\u03ba\u03bf\u03cd\u03c2 \u03c4\u03bf\u03bc\u03b5\u03af\u03c2.<\/p>\n<\/div>\n<\/details>\n<details class=\"td-faq-item\">\n<summary class=\"td-faq-title\">\u0389\u03c4\u03b1\u03bd \u03ba\u03b1\u03bb\u03cd\u03c4\u03b5\u03c1\u03bf \u03c3\u03b5 \u03cc\u03bb\u03b5\u03c2 \u03c4\u03b9\u03c2 \u03b4\u03bf\u03ba\u03b9\u03bc\u03ad\u03c2;<\/summary>\n<div class=\"td-faq-content\">\n<p>\u038c\u03c7\u03b9 \u03c3\u03b5 \u03ba\u03ac\u03b8\u03b5 \u03b5\u03c0\u03b9\u03bc\u03ad\u03c1\u03bf\u03c5\u03c2 benchmark. \u0395\u03af\u03c7\u03b5 \u03cc\u03bc\u03c9\u03c2 \u03c4\u03bf\u03bd \u03ba\u03b1\u03bb\u03cd\u03c4\u03b5\u03c1\u03bf \u03c7\u03c1\u03b7\u03bc\u03b1\u03c4\u03bf\u03bf\u03b9\u03ba\u03bf\u03bd\u03bf\u03bc\u03b9\u03ba\u03cc \u03bc\u03ad\u03c3\u03bf \u03cc\u03c1\u03bf \u03ba\u03b1\u03b9 \u03c3\u03c4\u03b1 \u03c4\u03ad\u03c3\u03c3\u03b5\u03c1\u03b1 \u03bc\u03bf\u03bd\u03c4\u03ad\u03bb\u03b1 \u03ba\u03b1\u03b9 \u03c4\u03bf\u03bd \u03ba\u03b1\u03bb\u03cd\u03c4\u03b5\u03c1\u03bf \u03bd\u03bf\u03bc\u03b9\u03ba\u03cc \u03bc\u03ad\u03c3\u03bf \u03cc\u03c1\u03bf \u03c3\u03c4\u03b1 \u03c4\u03c1\u03af\u03b1 \u03b1\u03c0\u03cc \u03c4\u03b1 \u03c4\u03ad\u03c3\u03c3\u03b5\u03c1\u03b1. \u03a3\u03c4\u03b7\u03bd \u03c4\u03ad\u03c4\u03b1\u03c1\u03c4\u03b7 \u03bd\u03bf\u03bc\u03b9\u03ba\u03ae \u03c0\u03b5\u03c1\u03af\u03c0\u03c4\u03c9\u03c3\u03b7, \u03c4\u03bf EvalTree \u03c0\u03c1\u03bf\u03b7\u03b3\u03ae\u03b8\u03b7\u03ba\u03b5 \u03ba\u03b1\u03c4\u03ac 0,6 \u03bc\u03bf\u03bd\u03ac\u03b4\u03b5\u03c2 \u03bc\u03b5 \u03b5\u03c0\u03b9\u03ba\u03b1\u03bb\u03c5\u03c0\u03c4\u03cc\u03bc\u03b5\u03bd\u03b1 variance bands.<\/p>\n<\/div>\n<\/details>\n<details class=\"td-faq-item\">\n<summary class=\"td-faq-title\">\u039c\u03c0\u03bf\u03c1\u03b5\u03af \u03bd\u03b1 \u03b5\u03c6\u03b1\u03c1\u03bc\u03bf\u03c3\u03c4\u03b5\u03af \u03c3\u03b5 e-commerce \u03ae marketing;<\/summary>\n<div class=\"td-faq-content\">\n<p>\u0397 \u03bb\u03bf\u03b3\u03b9\u03ba\u03ae \u03bc\u03c0\u03bf\u03c1\u03b5\u03af \u03bd\u03b1 \u03c0\u03c1\u03bf\u03c3\u03b1\u03c1\u03bc\u03bf\u03c3\u03c4\u03b5\u03af \u03c3\u03b5 domain-specific rubrics, \u03b1\u03bb\u03bb\u03ac \u03b1\u03c5\u03c4\u03cc \u03b1\u03c0\u03bf\u03c4\u03b5\u03bb\u03b5\u03af \u03b5\u03c0\u03b9\u03c7\u03b5\u03b9\u03c1\u03b7\u03c3\u03b9\u03b1\u03ba\u03ae \u03b5\u03c6\u03b1\u03c1\u03bc\u03bf\u03b3\u03ae \u03c4\u03b7\u03c2 \u03b9\u03b4\u03ad\u03b1\u03c2 \u03ba\u03b1\u03b9 \u03cc\u03c7\u03b9 \u03b1\u03c0\u03bf\u03c4\u03ad\u03bb\u03b5\u03c3\u03bc\u03b1 \u03c0\u03bf\u03c5 \u03b5\u03c0\u03b1\u03bb\u03b7\u03b8\u03b5\u03cd\u03c4\u03b7\u03ba\u03b5 \u03c3\u03c4\u03b7\u03bd \u03b5\u03c1\u03b3\u03b1\u03c3\u03af\u03b1. \u03a7\u03c1\u03b5\u03b9\u03ac\u03b6\u03bf\u03bd\u03c4\u03b1\u03b9 \u03b4\u03b9\u03ba\u03ac \u03c3\u03b1\u03c2 \u03ba\u03c1\u03b9\u03c4\u03ae\u03c1\u03b9\u03b1, \u03c0\u03c1\u03b1\u03b3\u03bc\u03b1\u03c4\u03b9\u03ba\u03ac prompts \u03ba\u03b1\u03b9 \u03b1\u03bd\u03b5\u03be\u03ac\u03c1\u03c4\u03b7\u03c4\u03b1 held-out tests.<\/p>\n<\/div>\n<\/details>\n<details class=\"td-faq-item\">\n<summary class=\"td-faq-title\">\u03a0\u03bf\u03b9\u03bf\u03c2 \u03b5\u03af\u03bd\u03b1\u03b9 \u03bf \u03c3\u03b7\u03bc\u03b1\u03bd\u03c4\u03b9\u03ba\u03cc\u03c4\u03b5\u03c1\u03bf\u03c2 \u03ba\u03af\u03bd\u03b4\u03c5\u03bd\u03bf\u03c2 \u03ba\u03b1\u03c4\u03ac \u03c4\u03b7\u03bd \u03c5\u03bb\u03bf\u03c0\u03bf\u03af\u03b7\u03c3\u03b7;<\/summary>\n<div class=\"td-faq-content\">\n<p>\u0397 \u03c0\u03bf\u03b9\u03cc\u03c4\u03b7\u03c4\u03b1 \u03c4\u03bf\u03c5 \u03c3\u03ae\u03bc\u03b1\u03c4\u03bf\u03c2. \u0391\u03c3\u03b1\u03c6\u03ae rubrics \u03ae \u03b1\u03c3\u03c5\u03bd\u03b5\u03c0\u03ae\u03c2 LLM judge \u03bc\u03c0\u03bf\u03c1\u03bf\u03cd\u03bd \u03bd\u03b1 \u03b4\u03b7\u03bc\u03b9\u03bf\u03c5\u03c1\u03b3\u03ae\u03c3\u03bf\u03c5\u03bd \u03c0\u03b1\u03c1\u03b1\u03c0\u03bb\u03b1\u03bd\u03b7\u03c4\u03b9\u03ba\u03ac node scores. \u0393\u03b9\u03b1 \u03ba\u03c1\u03af\u03c3\u03b9\u03bc\u03b5\u03c2 \u03c7\u03c1\u03ae\u03c3\u03b5\u03b9\u03c2 \u03c7\u03c1\u03b5\u03b9\u03ac\u03b6\u03bf\u03bd\u03c4\u03b1\u03b9 \u03ad\u03bb\u03b5\u03b3\u03c7\u03bf\u03b9 \u03c3\u03c5\u03bd\u03ad\u03c0\u03b5\u03b9\u03b1\u03c2, \u03b1\u03bd\u03b8\u03c1\u03ce\u03c0\u03b9\u03bd\u03bf review \u03ba\u03b1\u03b9 task-specific \u03b1\u03c0\u03bf\u03b4\u03bf\u03c7\u03ae.<\/p>\n<\/div>\n<\/details>\n<\/div>\n<\/section>\n<div class=\"td-source-list\">\n<p id=\"piges\" class=\"td-source-list-title\">Sources<\/p>\n<ul>\n<li><a href=\"https:\/\/arxiv.org\/abs\/2607.16122\" target=\"_blank\" rel=\"noopener\">CRAFT: Clustering Rubrics to Diagnose Weak LLM Capabilities and Generate Targeted Fine-Tuning Data<\/a><\/li>\n<li><a href=\"https:\/\/arxiv.org\/abs\/2511.11562\" target=\"_blank\" rel=\"noopener\">PRBench: Large-Scale Expert Rubrics for Evaluating High-Stakes Professional Reasoning<\/a><\/li>\n<li><a href=\"https:\/\/www.nist.gov\/publications\/artificial-intelligence-risk-management-framework-generative-artificial-intelligence\" target=\"_blank\" rel=\"noopener\">NIST AI 600-1: Generative Artificial Intelligence Profile<\/a><\/li>\n<li><a href=\"https:\/\/airc.nist.gov\/airmf-resources\/airmf\/5-sec-core\/\" target=\"_blank\" rel=\"noopener\">NIST AI RMF Core: Measure \u03ba\u03b1\u03b9 Manage \u03b3\u03b9\u03b1 \u03b1\u03be\u03b9\u03bf\u03bb\u03cc\u03b3\u03b7\u03c3\u03b7 \u03ba\u03b1\u03b9 \u03c0\u03b1\u03c1\u03b1\u03ba\u03bf\u03bb\u03bf\u03cd\u03b8\u03b7\u03c3\u03b7 AI<\/a><\/li>\n<\/ul>\n<\/div>","protected":false},"excerpt":{"rendered":"<p>CRAFT converts rubric-based failures into targeted training data. How it works, what 13 benchmarks revealed, and what an enterprise AI pilot needs.<\/p>","protected":false},"author":1,"featured_media":86986,"comment_status":"open","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"_gspb_post_css":"","content-type":"","footnotes":""},"categories":[17366],"tags":[18888,17781,9225,7477,3597],"class_list":["post-86975","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-digital-marketing","tag-ai-evaluation","tag-fine-tuning","tag-llm","tag-machine-learning","tag-techniti-noimosyni"],"blocksy_meta":[],"_links":{"self":[{"href":"https:\/\/twodots.gr\/en\/wp-json\/wp\/v2\/posts\/86975","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/twodots.gr\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/twodots.gr\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/twodots.gr\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/twodots.gr\/en\/wp-json\/wp\/v2\/comments?post=86975"}],"version-history":[{"count":0,"href":"https:\/\/twodots.gr\/en\/wp-json\/wp\/v2\/posts\/86975\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/twodots.gr\/en\/wp-json\/wp\/v2\/media\/86986"}],"wp:attachment":[{"href":"https:\/\/twodots.gr\/en\/wp-json\/wp\/v2\/media?parent=86975"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/twodots.gr\/en\/wp-json\/wp\/v2\/categories?post=86975"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/twodots.gr\/en\/wp-json\/wp\/v2\/tags?post=86975"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}