{"id":97801,"date":"2026-09-18T12:14:17","date_gmt":"2026-09-18T09:14:17","guid":{"rendered":"https:\/\/twodots.gr\/?p=97801"},"modified":"2026-09-18T12:14:19","modified_gmt":"2026-09-18T09:14:19","slug":"k-bench-ai-agents-epistimoniki-akriveia-artifacts","status":"publish","type":"post","link":"https:\/\/twodots.gr\/en\/k-bench-ai-agents-epistimoniki-akriveia-artifacts\/","title":{"rendered":"K-Bench: \u03b3\u03b9\u03b1\u03c4\u03af \u03b7 \u03ba\u03b1\u03bb\u03ae \u03b1\u03c0\u03ac\u03bd\u03c4\u03b7\u03c3\u03b7 \u03b5\u03bd\u03cc\u03c2 AI agent \u03b4\u03b5\u03bd \u03c3\u03b7\u03bc\u03b1\u03af\u03bd\u03b5\u03b9 \u03c3\u03c9\u03c3\u03c4\u03ae \u03b4\u03bf\u03c5\u03bb\u03b5\u03b9\u03ac"},"content":{"rendered":"<div class=\"td-article-lede\">\n<p><strong>Answer first:<\/strong> \u03cc\u03c7\u03b9, \u03bc\u03b9\u03b1 \u03ac\u03c8\u03bf\u03b3\u03b1 \u03b4\u03b9\u03b1\u03c4\u03c5\u03c0\u03c9\u03bc\u03ad\u03bd\u03b7 \u03b1\u03c0\u03ac\u03bd\u03c4\u03b7\u03c3\u03b7 \u03b4\u03b5\u03bd \u03b1\u03c0\u03bf\u03b4\u03b5\u03b9\u03ba\u03bd\u03cd\u03b5\u03b9 \u03cc\u03c4\u03b9 \u03ad\u03bd\u03b1\u03c2 AI agent \u03ad\u03ba\u03b1\u03bd\u03b5 \u03c3\u03c9\u03c3\u03c4\u03ac \u03c4\u03b7 \u03b4\u03bf\u03c5\u03bb\u03b5\u03b9\u03ac. \u03a3\u03c4\u03bf K-Bench 01, \u03b5\u03bd\u03bd\u03ad\u03b1 \u03bc\u03bf\u03bd\u03c4\u03ad\u03bb\u03b1 \u03b5\u03ba\u03c4\u03ad\u03bb\u03b5\u03c3\u03b1\u03bd 178 \u03c0\u03c1\u03b1\u03b3\u03bc\u03b1\u03c4\u03b9\u03ba\u03ac \u03b5\u03c0\u03b9\u03c3\u03c4\u03b7\u03bc\u03bf\u03bd\u03b9\u03ba\u03ac \u03b1\u03b9\u03c4\u03ae\u03bc\u03b1\u03c4\u03b1 \u03c3\u03b5 \u03af\u03b4\u03b9\u03b1 sandboxes, \u03cc\u03bc\u03c9\u03c2 \u03b7 \u03b5\u03c0\u03b9\u03ba\u03bf\u03b9\u03bd\u03c9\u03bd\u03af\u03b1 \u03b2\u03b1\u03b8\u03bc\u03bf\u03bb\u03bf\u03b3\u03ae\u03b8\u03b7\u03ba\u03b5 \u03ba\u03b1\u03c4\u03ac \u03bc\u03ad\u03c3\u03bf \u03cc\u03c1\u03bf \u03bc\u03b5 7,33\/10 \u03ba\u03b1\u03b9 \u03b7 \u03b5\u03c0\u03b9\u03c3\u03c4\u03b7\u03bc\u03bf\u03bd\u03b9\u03ba\u03ae \u03b1\u03ba\u03c1\u03af\u03b2\u03b5\u03b9\u03b1 \u03bc\u03b5 6,22\/10.<\/p>\n<p>\u03a4\u03bf \u03c0\u03c1\u03b1\u03ba\u03c4\u03b9\u03ba\u03cc \u03bc\u03ac\u03b8\u03b7\u03bc\u03b1 \u03b3\u03b9\u03b1 \u03bc\u03b9\u03b1 \u03b5\u03c0\u03b9\u03c7\u03b5\u03af\u03c1\u03b7\u03c3\u03b7 \u03b5\u03af\u03bd\u03b1\u03b9 \u03bd\u03b1 \u03b5\u03bb\u03ad\u03b3\u03c7\u03b5\u03b9 \u03c7\u03c9\u03c1\u03b9\u03c3\u03c4\u03ac \u03c4\u03bf \u03c4\u03b5\u03bb\u03b9\u03ba\u03cc \u03ba\u03b5\u03af\u03bc\u03b5\u03bd\u03bf, \u03c4\u03b1 \u03b4\u03b5\u03b4\u03bf\u03bc\u03ad\u03bd\u03b1, \u03c4\u03b9\u03c2 \u03c0\u03b7\u03b3\u03ad\u03c2, \u03c4\u03b1 tool outputs \u03ba\u03b1\u03b9 \u03c4\u03b1 \u03c0\u03b1\u03c1\u03b1\u03b4\u03bf\u03c4\u03ad\u03b1 \u03b1\u03c1\u03c7\u03b5\u03af\u03b1. \u03a4\u03bf K-Bench \u03b4\u03b5\u03bd \u03b4\u03af\u03bd\u03b5\u03b9 \u03bc\u03b9\u03b1 \u03ba\u03b1\u03b8\u03bf\u03bb\u03b9\u03ba\u03ae \u03ba\u03b1\u03c4\u03ac\u03c4\u03b1\u03be\u03b7 \u03bc\u03bf\u03bd\u03c4\u03ad\u03bb\u03c9\u03bd: \u03bc\u03b5\u03c4\u03c1\u03ac \u03c3\u03c5\u03b3\u03ba\u03b5\u03ba\u03c1\u03b9\u03bc\u03ad\u03bd\u03b1 \u03c3\u03c5\u03c3\u03c4\u03ae\u03bc\u03b1\u03c4\u03b1, \u03c3\u03b5 \u03ad\u03bd\u03b1 stock harness, \u03bc\u03b5 \u03ad\u03bd\u03b1 run \u03b1\u03bd\u03ac task \u03ba\u03b1\u03b9 \u03c4\u03c1\u03b5\u03b9\u03c2 LLM judges \u03c7\u03c9\u03c1\u03af\u03c2 \u03b1\u03bd\u03b8\u03c1\u03ce\u03c0\u03b9\u03bd\u03bf expert baseline. \u0394\u03b5\u03af\u03c7\u03bd\u03b5\u03b9 \u03cc\u03bc\u03c9\u03c2 \u03ba\u03b1\u03b8\u03b1\u03c1\u03ac \u03b3\u03b9\u03b1\u03c4\u03af \u03b7 \u03b5\u03c5\u03c6\u03c1\u03ac\u03b4\u03b5\u03b9\u03b1 \u03b4\u03b5\u03bd \u03c0\u03c1\u03ad\u03c0\u03b5\u03b9 \u03bd\u03b1 \u03bb\u03b5\u03b9\u03c4\u03bf\u03c5\u03c1\u03b3\u03b5\u03af \u03c9\u03c2 proxy \u03bf\u03c1\u03b8\u03cc\u03c4\u03b7\u03c4\u03b1\u03c2.<\/p>\n<\/div>\n<div class=\"td-article-toc\">\n<div class=\"td-toc-title\">Contents<\/div>\n<ul>\n<li><a href=\"#pragmatika-aitimata\">\u0391\u03c0\u03cc \u03c4\u03b9\u03c2 \u03b5\u03be\u03b5\u03c4\u03ac\u03c3\u03b5\u03b9\u03c2 \u03c3\u03c4\u03b1 \u03c0\u03c1\u03b1\u03b3\u03bc\u03b1\u03c4\u03b9\u03ba\u03ac \u03b5\u03c0\u03b9\u03c3\u03c4\u03b7\u03bc\u03bf\u03bd\u03b9\u03ba\u03ac \u03b1\u03b9\u03c4\u03ae\u03bc\u03b1\u03c4\u03b1<\/a><\/li>\n<li><a href=\"#pos-stithike\">\u03a0\u03ce\u03c2 \u03c3\u03c4\u03ae\u03b8\u03b7\u03ba\u03b5 \u03b7 \u03b1\u03be\u03b9\u03bf\u03bb\u03cc\u03b3\u03b7\u03c3\u03b7<\/a><\/li>\n<li><a href=\"#anoixti-koryfi\">\u0397 \u03ba\u03bf\u03c1\u03c5\u03c6\u03ae \u03c4\u03b7\u03c2 \u03ba\u03b1\u03c4\u03ac\u03c4\u03b1\u03be\u03b7\u03c2 \u03ad\u03bc\u03b5\u03b9\u03bd\u03b5 \u03b1\u03bd\u03bf\u03b9\u03c7\u03c4\u03ae<\/a><\/li>\n<li><a href=\"#epikoinonia-akriveia\">\u0397 \u03b5\u03c0\u03b9\u03ba\u03bf\u03b9\u03bd\u03c9\u03bd\u03af\u03b1 \u03c0\u03c1\u03bf\u03b7\u03b3\u03b5\u03af\u03c4\u03b1\u03b9 \u03c4\u03b7\u03c2 \u03b5\u03c0\u03b9\u03c3\u03c4\u03b7\u03bc\u03bf\u03bd\u03b9\u03ba\u03ae\u03c2 \u03b1\u03ba\u03c1\u03af\u03b2\u03b5\u03b9\u03b1\u03c2<\/a><\/li>\n<li><a href=\"#artifacts\">\u03a4\u03b1 artifacts \u03b5\u03af\u03bd\u03b1\u03b9 \u03bc\u03ad\u03c1\u03bf\u03c2 \u03c4\u03b7\u03c2 \u03b1\u03c0\u03ac\u03bd\u03c4\u03b7\u03c3\u03b7\u03c2<\/a><\/li>\n<li><a href=\"#overclaiming\">\u03a4\u03bf \u03c3\u03c5\u03c7\u03bd\u03cc\u03c4\u03b5\u03c1\u03bf failure mode \u03ae\u03c4\u03b1\u03bd \u03c4\u03bf overclaiming<\/a><\/li>\n<li><a href=\"#attachments-dyskolia\">Attachments \u03ba\u03b1\u03b9 \u03bc\u03b5\u03b3\u03ac\u03bb\u03b1 prompts: \u03c0\u03c1\u03b1\u03b3\u03bc\u03b1\u03c4\u03b9\u03ba\u03ae \u03b4\u03c5\u03c3\u03ba\u03bf\u03bb\u03af\u03b1, \u03cc\u03c7\u03b9 \u03b1\u03c0\u03bb\u03ae \u03b1\u03b9\u03c4\u03af\u03b1<\/a><\/li>\n<li><a href=\"#kostos-prospatheia\">\u03a4\u03bf \u03ba\u03cc\u03c3\u03c4\u03bf\u03c2 \u03ba\u03b1\u03b9 \u03b7 \u03c0\u03c1\u03bf\u03c3\u03c0\u03ac\u03b8\u03b5\u03b9\u03b1 \u03b4\u03b5\u03bd \u03b5\u03b3\u03b3\u03c5\u03ce\u03bd\u03c4\u03b1\u03b9 \u03c0\u03bf\u03b9\u03cc\u03c4\u03b7\u03c4\u03b1<\/a><\/li>\n<li><a href=\"#krities-organo\">\u039f\u03b9 \u03ba\u03c1\u03b9\u03c4\u03ad\u03c2 \u03b5\u03af\u03bd\u03b1\u03b9 \u03ba\u03b1\u03b9 \u03b1\u03c5\u03c4\u03bf\u03af \u03bc\u03ad\u03c1\u03bf\u03c2 \u03c4\u03bf\u03c5 \u03c3\u03c5\u03c3\u03c4\u03ae\u03bc\u03b1\u03c4\u03bf\u03c2<\/a><\/li>\n<li><a href=\"#oria-privacy\">\u038c\u03c1\u03b9\u03b1, privacy \u03ba\u03b1\u03b9 \u03b4\u03c5\u03bd\u03b1\u03c4\u03cc\u03c4\u03b7\u03c4\u03b1 \u03b1\u03bd\u03b1\u03c0\u03b1\u03c1\u03b1\u03b3\u03c9\u03b3\u03ae\u03c2<\/a><\/li>\n<li><a href=\"#evaluation-stack\">\u0388\u03bd\u03b1 \u03c0\u03c1\u03b1\u03ba\u03c4\u03b9\u03ba\u03cc evaluation stack \u03b3\u03b9\u03b1 \u03b5\u03c0\u03b9\u03c7\u03b5\u03b9\u03c1\u03ae\u03c3\u03b5\u03b9\u03c2<\/a><\/li>\n<li><a href=\"#symperasma\">\u03a4\u03bf \u03c3\u03c9\u03c3\u03c4\u03cc \u03c3\u03c5\u03bc\u03c0\u03ad\u03c1\u03b1\u03c3\u03bc\u03b1 \u03b4\u03b5\u03bd \u03b5\u03af\u03bd\u03b1\u03b9 \u03ad\u03bd\u03b1\u03c2 \u03bd\u03b9\u03ba\u03b7\u03c4\u03ae\u03c2<\/a><\/li>\n<\/ul>\n<\/div>\n<h2 id=\"pragmatika-aitimata\">\u0391\u03c0\u03cc \u03c4\u03b9\u03c2 \u03b5\u03be\u03b5\u03c4\u03ac\u03c3\u03b5\u03b9\u03c2 \u03c3\u03c4\u03b1 \u03c0\u03c1\u03b1\u03b3\u03bc\u03b1\u03c4\u03b9\u03ba\u03ac \u03b5\u03c0\u03b9\u03c3\u03c4\u03b7\u03bc\u03bf\u03bd\u03b9\u03ba\u03ac \u03b1\u03b9\u03c4\u03ae\u03bc\u03b1\u03c4\u03b1<\/h2>\n<p>\u03a0\u03bf\u03bb\u03bb\u03ac benchmarks \u03b5\u03af\u03bd\u03b1\u03b9 \u03c3\u03c7\u03b5\u03b4\u03b9\u03b1\u03c3\u03bc\u03ad\u03bd\u03b1 \u03ce\u03c3\u03c4\u03b5 \u03bd\u03b1 \u03b2\u03b1\u03b8\u03bc\u03bf\u03bb\u03bf\u03b3\u03bf\u03cd\u03bd\u03c4\u03b1\u03b9 \u03b5\u03cd\u03ba\u03bf\u03bb\u03b1: \u03c0\u03bf\u03bb\u03bb\u03b1\u03c0\u03bb\u03ad\u03c2 \u03b5\u03c0\u03b9\u03bb\u03bf\u03b3\u03ad\u03c2, \u03c0\u03c1\u03bf\u03b2\u03bb\u03ae\u03bc\u03b1\u03c4\u03b1 \u03bc\u03b5 \u03b3\u03bd\u03c9\u03c3\u03c4\u03ae \u03bb\u03cd\u03c3\u03b7 \u03ae \u03c0\u03c1\u03bf\u03c3\u03bf\u03bc\u03bf\u03b9\u03ce\u03c3\u03b5\u03b9\u03c2 \u03bc\u03b5 \u03b5\u03bb\u03b5\u03b3\u03c7\u03cc\u03bc\u03b5\u03bd\u03b7 \u03b4\u03bf\u03bc\u03ae. \u0391\u03c5\u03c4\u03ac \u03c0\u03b1\u03c1\u03b1\u03bc\u03ad\u03bd\u03bf\u03c5\u03bd \u03c7\u03c1\u03ae\u03c3\u03b9\u03bc\u03b1 \u03b3\u03b9\u03b1 \u03c4\u03b7 \u03b3\u03bd\u03ce\u03c3\u03b7 \u03ba\u03b1\u03b9 \u03c4\u03b7 \u03c3\u03c5\u03bb\u03bb\u03bf\u03b3\u03b9\u03c3\u03c4\u03b9\u03ba\u03ae \u03c0\u03bf\u03c5 \u03bc\u03b5\u03c4\u03c1\u03bf\u03cd\u03bd, \u03b1\u03bb\u03bb\u03ac \u03b4\u03b5\u03bd \u03b1\u03bd\u03b1\u03c0\u03b1\u03c1\u03b9\u03c3\u03c4\u03bf\u03cd\u03bd \u03c0\u03bb\u03ae\u03c1\u03c9\u03c2 \u03ad\u03bd\u03b1 \u03c0\u03c1\u03b1\u03b3\u03bc\u03b1\u03c4\u03b9\u03ba\u03cc \u03b1\u03af\u03c4\u03b7\u03bc\u03b1. \u0388\u03bd\u03b1\u03c2 \u03b5\u03c0\u03b9\u03c3\u03c4\u03ae\u03bc\u03bf\u03bd\u03b1\u03c2 \u03bc\u03c0\u03bf\u03c1\u03b5\u03af \u03bd\u03b1 \u03b1\u03bd\u03b5\u03b2\u03ac\u03c3\u03b5\u03b9 \u03c0\u03af\u03bd\u03b1\u03ba\u03b5\u03c2, papers \u03ae \u03b5\u03b9\u03ba\u03cc\u03bd\u03b5\u03c2, \u03bd\u03b1 \u03b6\u03b7\u03c4\u03ae\u03c3\u03b5\u03b9 \u03b1\u03bd\u03ac\u03bb\u03c5\u03c3\u03b7 \u03bc\u03b5 \u03bc\u03b9\u03c3\u03bf\u03b4\u03b9\u03b1\u03c4\u03c5\u03c0\u03c9\u03bc\u03ad\u03bd\u03bf \u03c3\u03c4\u03cc\u03c7\u03bf \u03ba\u03b1\u03b9 \u03bd\u03b1 \u03bc\u03b7\u03bd \u03c5\u03c0\u03ac\u03c1\u03c7\u03b5\u03b9 \u03bc\u03af\u03b1 \u03bc\u03bf\u03bd\u03b1\u03b4\u03b9\u03ba\u03ae \u03c3\u03c9\u03c3\u03c4\u03ae \u03b1\u03c0\u03ac\u03bd\u03c4\u03b7\u03c3\u03b7.<\/p>\n<p>\u03a4\u03bf K-Bench 01 \u03b5\u03c0\u03b9\u03bb\u03ad\u03b3\u03b5\u03b9 \u03c4\u03b7\u03bd \u03b1\u03bd\u03c4\u03af\u03b8\u03b5\u03c4\u03b7 \u03b1\u03bd\u03c4\u03b1\u03bb\u03bb\u03b1\u03b3\u03ae: \u03b4\u03b9\u03b1\u03c4\u03b7\u03c1\u03b5\u03af \u03c4\u03b1 \u03c0\u03c1\u03ce\u03c4\u03b1 \u03bc\u03b7\u03bd\u03cd\u03bc\u03b1\u03c4\u03b1 \u03c7\u03c1\u03b7\u03c3\u03c4\u03ce\u03bd \u03b1\u03c5\u03c4\u03bf\u03cd\u03c3\u03b9\u03b1, \u03c7\u03c9\u03c1\u03af\u03c2 \u03b1\u03bd\u03b1\u03ba\u03b1\u03c4\u03b1\u03c3\u03ba\u03b5\u03c5\u03ae \u03ba\u03b1\u03b9 \u03c7\u03c9\u03c1\u03af\u03c2 reference answer. \u039f\u03b9 \u03ba\u03c1\u03b9\u03c4\u03ad\u03c2 \u03b5\u03be\u03b5\u03c4\u03ac\u03b6\u03bf\u03c5\u03bd \u03c4\u03cc\u03c3\u03bf \u03c4\u03bf transcript \u03cc\u03c3\u03bf \u03ba\u03b1\u03b9 \u03c4\u03bf \u03b4\u03ad\u03bd\u03c4\u03c1\u03bf \u03b1\u03c1\u03c7\u03b5\u03af\u03c9\u03bd \u03c0\u03bf\u03c5 \u03ac\u03c6\u03b7\u03c3\u03b5 \u03ba\u03ac\u03b8\u03b5 run. \u0388\u03c4\u03c3\u03b9 \u03bc\u03b9\u03b1 \u03ba\u03b1\u03bb\u03ae \u03c0\u03b5\u03c1\u03b9\u03b3\u03c1\u03b1\u03c6\u03ae \u03b4\u03b5\u03bd \u03bc\u03c0\u03bf\u03c1\u03b5\u03af \u03b1\u03c0\u03cc \u03bc\u03cc\u03bd\u03b7 \u03c4\u03b7\u03c2 \u03bd\u03b1 \u03ba\u03b1\u03bb\u03cd\u03c8\u03b5\u03b9 \u03ad\u03bd\u03b1\u03bd \u03ac\u03b4\u03b5\u03b9\u03bf \u03c6\u03ac\u03ba\u03b5\u03bb\u03bf, \u03ad\u03bd\u03b1 \u03b5\u03bb\u03bb\u03b9\u03c0\u03ad\u03c2 spreadsheet \u03ae \u03ba\u03ce\u03b4\u03b9\u03ba\u03b1 \u03c0\u03bf\u03c5 \u03b4\u03b5\u03bd \u03b5\u03ba\u03c4\u03b5\u03bb\u03b5\u03af\u03c4\u03b1\u03b9.<\/p>\n<p>\u0391\u03c5\u03c4\u03cc \u03c6\u03ad\u03c1\u03bd\u03b5\u03b9 \u03c4\u03b7\u03bd \u03b1\u03be\u03b9\u03bf\u03bb\u03cc\u03b3\u03b7\u03c3\u03b7 \u03c0\u03b9\u03bf \u03ba\u03bf\u03bd\u03c4\u03ac \u03c3\u03b5 \u03c0\u03c1\u03b1\u03b3\u03bc\u03b1\u03c4\u03b9\u03ba\u03ae \u03b1\u03c0\u03bf\u03b4\u03bf\u03c7\u03ae \u03b5\u03c1\u03b3\u03b1\u03c3\u03af\u03b1\u03c2. \u0397 \u03af\u03b4\u03b9\u03b1 \u03b1\u03c1\u03c7\u03ae \u03b9\u03c3\u03c7\u03cd\u03b5\u03b9 \u03c3\u03c4\u03bf <a href=\"https:\/\/twodots.gr\/ai-agents-behavioral-testing\/\">behavioral testing \u03c4\u03c9\u03bd AI agents<\/a>: \u03c4\u03bf \u03ba\u03c1\u03af\u03c3\u03b9\u03bc\u03bf \u03b5\u03c1\u03ce\u03c4\u03b7\u03bc\u03b1 \u03b4\u03b5\u03bd \u03b5\u03af\u03bd\u03b1\u03b9 \u03b1\u03bd \u03b7 \u03b1\u03c0\u03ac\u03bd\u03c4\u03b7\u03c3\u03b7 \u03c6\u03b1\u03af\u03bd\u03b5\u03c4\u03b1\u03b9 \u03c0\u03b5\u03b9\u03c3\u03c4\u03b9\u03ba\u03ae, \u03b1\u03bb\u03bb\u03ac \u03b1\u03bd \u03c4\u03bf \u03c3\u03cd\u03c3\u03c4\u03b7\u03bc\u03b1 \u03c0\u03ad\u03c4\u03c5\u03c7\u03b5 \u03c4\u03bf \u03b6\u03b7\u03c4\u03bf\u03cd\u03bc\u03b5\u03bd\u03bf, \u03c4\u03ae\u03c1\u03b7\u03c3\u03b5 \u03c4\u03bf\u03c5\u03c2 \u03c0\u03b5\u03c1\u03b9\u03bf\u03c1\u03b9\u03c3\u03bc\u03bf\u03cd\u03c2 \u03ba\u03b1\u03b9 \u03ac\u03c6\u03b7\u03c3\u03b5 \u03b5\u03c0\u03b1\u03bb\u03b7\u03b8\u03b5\u03cd\u03c3\u03b9\u03bc\u03bf \u03b1\u03c0\u03bf\u03c4\u03ad\u03bb\u03b5\u03c3\u03bc\u03b1.<\/p>\n<div class=\"td-comparison td-comparison-cards td-comparison-cards--horizontal\">\n<div class=\"td-comparison-grid td-comparison-grid--three\">\n<div class=\"td-platform-card\">\n<p class=\"td-comparison-title\">\u0391\u03c0\u03ac\u03bd\u03c4\u03b7\u03c3\u03b7 \u00b7 \u03a4\u03b9 \u03b4\u03b9\u03b1\u03b2\u03ac\u03b6\u03b5\u03b9 \u03bf \u03c7\u03c1\u03ae\u03c3\u03c4\u03b7\u03c2<\/p>\n<p>\u0394\u03bf\u03bc\u03ae, \u03c3\u03b1\u03c6\u03ae\u03bd\u03b5\u03b9\u03b1, \u03b3\u03bb\u03ce\u03c3\u03c3\u03b1 \u03ba\u03b1\u03b9 \u03c4\u03cc\u03bd\u03bf\u03c2. \u0395\u03af\u03bd\u03b1\u03b9 \u03c3\u03b7\u03bc\u03b1\u03bd\u03c4\u03b9\u03ba\u03ac \u03b3\u03b9\u03b1 \u03c4\u03b7 \u03c7\u03c1\u03ae\u03c3\u03b7, \u03b1\u03bb\u03bb\u03ac \u03b4\u03b5\u03bd \u03b1\u03c0\u03bf\u03b4\u03b5\u03b9\u03ba\u03bd\u03cd\u03bf\u03c5\u03bd \u03b5\u03c0\u03b9\u03c3\u03c4\u03b7\u03bc\u03bf\u03bd\u03b9\u03ba\u03ae \u03b1\u03ba\u03c1\u03af\u03b2\u03b5\u03b9\u03b1 \u03ae \u03bf\u03bb\u03bf\u03ba\u03bb\u03ae\u03c1\u03c9\u03c3\u03b7.<\/p>\n<div class=\"td-badge-row\"><span class=\"td-badge\">Communication<\/span><span class=\"td-badge\">Claims<\/span><\/div>\n<\/div>\n<div class=\"td-platform-card td-platform-card--navy\">\n<p class=\"td-comparison-title\">\u0395\u03c1\u03b3\u03b1\u03c3\u03af\u03b1 \u00b7 \u03a4\u03b9 \u03ad\u03ba\u03b1\u03bd\u03b5 \u03bf agent<\/p>\n<p>\u03a0\u03b7\u03b3\u03ad\u03c2, \u03c5\u03c0\u03bf\u03bb\u03bf\u03b3\u03b9\u03c3\u03bc\u03bf\u03af, tool calls, sanity checks, \u03c7\u03b5\u03b9\u03c1\u03b9\u03c3\u03bc\u03cc\u03c2 \u03b4\u03b5\u03b4\u03bf\u03bc\u03ad\u03bd\u03c9\u03bd \u03ba\u03b1\u03b9 \u03b1\u03bd\u03b1\u03b3\u03bd\u03ce\u03c1\u03b9\u03c3\u03b7 \u03b1\u03c0\u03bf\u03c4\u03c5\u03c7\u03b9\u03ce\u03bd \u03bc\u03ad\u03c3\u03b1 \u03c3\u03c4\u03bf transcript.<\/p>\n<div class=\"td-badge-row\"><span class=\"td-badge\">Evidence<\/span><span class=\"td-badge\">Execution<\/span><\/div>\n<\/div>\n<div class=\"td-platform-card\">\n<p class=\"td-comparison-title\">\u03a0\u03b1\u03c1\u03b1\u03b4\u03bf\u03c4\u03ad\u03bf \u00b7 \u03a4\u03b9 \u03bc\u03c0\u03bf\u03c1\u03b5\u03af \u03bd\u03b1 \u03c7\u03c1\u03b7\u03c3\u03b9\u03bc\u03bf\u03c0\u03bf\u03b9\u03b7\u03b8\u03b5\u03af<\/p>\n<p>\u0391\u03c1\u03c7\u03b5\u03af\u03b1 \u03c0\u03bf\u03c5 \u03b1\u03bd\u03bf\u03af\u03b3\u03bf\u03c5\u03bd, \u03c3\u03c9\u03c3\u03c4\u03ac \u03b4\u03b5\u03b4\u03bf\u03bc\u03ad\u03bd\u03b1, \u03b5\u03ba\u03c4\u03b5\u03bb\u03ad\u03c3\u03b9\u03bc\u03bf\u03c2 \u03ba\u03ce\u03b4\u03b9\u03ba\u03b1\u03c2 \u03ba\u03b1\u03b9 \u03b1\u03c0\u03bf\u03c4\u03ad\u03bb\u03b5\u03c3\u03bc\u03b1 \u03c0\u03bf\u03c5 \u03ba\u03b1\u03bb\u03cd\u03c0\u03c4\u03b5\u03b9 \u03c4\u03bf \u03c0\u03c1\u03b1\u03b3\u03bc\u03b1\u03c4\u03b9\u03ba\u03cc \u03b1\u03af\u03c4\u03b7\u03bc\u03b1.<\/p>\n<div class=\"td-badge-row\"><span class=\"td-badge\">Artifacts<\/span><span class=\"td-badge\">Acceptance<\/span><\/div>\n<\/div>\n<\/div>\n<\/div>\n<h2 id=\"pos-stithike\">\u03a0\u03ce\u03c2 \u03c3\u03c4\u03ae\u03b8\u03b7\u03ba\u03b5 \u03b7 \u03b1\u03be\u03b9\u03bf\u03bb\u03cc\u03b3\u03b7\u03c3\u03b7<\/h2>\n<p>\u039f\u03b9 178 \u03b5\u03c1\u03b3\u03b1\u03c3\u03af\u03b5\u03c2 \u03c0\u03c1\u03bf\u03ae\u03bb\u03b8\u03b1\u03bd \u03b1\u03c0\u03cc \u03b4\u03cd\u03bf \u03bf\u03bc\u03bf\u03b9\u03cc\u03bc\u03bf\u03c1\u03c6\u03b1 \u03c4\u03c5\u03c7\u03b1\u03af\u03b1 \u03b4\u03b5\u03af\u03b3\u03bc\u03b1\u03c4\u03b1 live traffic \u03bc\u03ad\u03c3\u03b1 \u03c3\u03b5 \u03bc\u03af\u03b1 \u03b5\u03b2\u03b4\u03bf\u03bc\u03ac\u03b4\u03b1. \u039a\u03ac\u03b8\u03b5 \u03b5\u03c1\u03b3\u03b1\u03c3\u03af\u03b1 \u03b5\u03ba\u03c4\u03b5\u03bb\u03ad\u03c3\u03c4\u03b7\u03ba\u03b5 \u03bc\u03af\u03b1 \u03c6\u03bf\u03c1\u03ac \u03c3\u03b5 \u03ba\u03b1\u03b8\u03ad\u03bd\u03b1 \u03b1\u03c0\u03cc \u03b5\u03bd\u03bd\u03ad\u03b1 frontier models, \u03bc\u03ad\u03c3\u03b1 \u03c3\u03b5 \u03ba\u03bf\u03b9\u03bd\u03cc stock harness \u03ba\u03b1\u03b9 \u03af\u03b4\u03b9\u03b1 Modal sandboxes. \u03a4\u03bf \u03c0\u03b5\u03af\u03c1\u03b1\u03bc\u03b1 \u03bf\u03bb\u03bf\u03ba\u03bb\u03ae\u03c1\u03c9\u03c3\u03b5 1.602 agent runs.<\/p>\n<p>\u03a4\u03c1\u03b5\u03b9\u03c2 language-model judges \u2014gpt-5.6-sol, qwen3.8-max \u03ba\u03b1\u03b9 grok-4.5\u2014 \u03b2\u03b1\u03b8\u03bc\u03bf\u03bb\u03cc\u03b3\u03b7\u03c3\u03b1\u03bd \u03b1\u03bd\u03b5\u03be\u03ac\u03c1\u03c4\u03b7\u03c4\u03b1 \u03ba\u03ac\u03b8\u03b5 run \u03c7\u03c9\u03c1\u03af\u03c2 \u03bd\u03b1 \u03b2\u03bb\u03ad\u03c0\u03bf\u03c5\u03bd \u03c4\u03b7\u03bd \u03c4\u03b1\u03c5\u03c4\u03cc\u03c4\u03b7\u03c4\u03b1 \u03c4\u03bf\u03c5 \u03b4\u03b9\u03b1\u03b3\u03c9\u03bd\u03b9\u03b6\u03cc\u03bc\u03b5\u03bd\u03bf\u03c5 \u03c3\u03c5\u03c3\u03c4\u03ae\u03bc\u03b1\u03c4\u03bf\u03c2. \u0394\u03b9\u03ac\u03b2\u03b1\u03c3\u03b1\u03bd \u03c4\u03bf \u03b1\u03af\u03c4\u03b7\u03bc\u03b1, \u03c4\u03bf scrubbed transcript, \u03ad\u03bd\u03b1 deterministic execution digest, \u03c4\u03bf inventory \u03c4\u03c9\u03bd artifacts \u03ba\u03b1\u03b9 \u03c4\u03b1 \u03c0\u03c1\u03b1\u03b3\u03bc\u03b1\u03c4\u03b9\u03ba\u03ac output files. \u03a0\u03b1\u03c1\u03ae\u03b3\u03b1\u03b3\u03b1\u03bd 4.806 assessments \u03ba\u03b1\u03b9 39.934 \u03b5\u03c6\u03b1\u03c1\u03bc\u03bf\u03c3\u03bc\u03ad\u03bd\u03b5\u03c2 \u03b2\u03b1\u03b8\u03bc\u03bf\u03bb\u03bf\u03b3\u03ae\u03c3\u03b5\u03b9\u03c2 \u03b4\u03b9\u03b1\u03c3\u03c4\u03ac\u03c3\u03b5\u03c9\u03bd \u03ae \u03c3\u03c5\u03bd\u03bf\u03bb\u03b9\u03ba\u03bf\u03cd \u03b1\u03c0\u03bf\u03c4\u03b5\u03bb\u03ad\u03c3\u03bc\u03b1\u03c4\u03bf\u03c2.<\/p>\n<p>\u039f\u03b9 \u03bf\u03ba\u03c4\u03ce \u03b4\u03b9\u03b1\u03c3\u03c4\u03ac\u03c3\u03b5\u03b9\u03c2 \u03ae\u03c4\u03b1\u03bd \u03b5\u03ba\u03c0\u03bb\u03ae\u03c1\u03c9\u03c3\u03b7 \u03b1\u03b9\u03c4\u03ae\u03bc\u03b1\u03c4\u03bf\u03c2, \u03b5\u03c0\u03b9\u03c3\u03c4\u03b7\u03bc\u03bf\u03bd\u03b9\u03ba\u03ae \u03b1\u03ba\u03c1\u03af\u03b2\u03b5\u03b9\u03b1, \u03c0\u03bf\u03b9\u03cc\u03c4\u03b7\u03c4\u03b1 reasoning, \u03c7\u03c1\u03ae\u03c3\u03b7 \u03b5\u03c1\u03b3\u03b1\u03bb\u03b5\u03af\u03c9\u03bd, \u03b4\u03b9\u03b1\u03c7\u03b5\u03af\u03c1\u03b9\u03c3\u03b7 \u03b4\u03b5\u03b4\u03bf\u03bc\u03ad\u03bd\u03c9\u03bd, \u03c0\u03bf\u03b9\u03cc\u03c4\u03b7\u03c4\u03b1 artifacts, \u03b5\u03c0\u03b9\u03ba\u03bf\u03b9\u03bd\u03c9\u03bd\u03af\u03b1 \u03ba\u03b1\u03b9 \u03b5\u03b9\u03bb\u03b9\u03ba\u03c1\u03af\u03bd\u03b5\u03b9\u03b1\/\u03b2\u03b1\u03b8\u03bc\u03bf\u03bd\u03cc\u03bc\u03b7\u03c3\u03b7. \u03a4\u03bf 8 \u03bf\u03c1\u03af\u03c3\u03c4\u03b7\u03ba\u03b5 \u03c9\u03c2 \u03b4\u03bf\u03c5\u03bb\u03b5\u03b9\u03ac \u03c0\u03bf\u03c5 \u03ad\u03bd\u03b1\u03c2 domain scientist \u03b8\u03b1 \u03b1\u03c0\u03bf\u03b4\u03b5\u03c7\u03cc\u03c4\u03b1\u03bd \u03bc\u03b5 \u03bc\u03b9\u03ba\u03c1\u03ad\u03c2 \u03b4\u03b9\u03bf\u03c1\u03b8\u03ce\u03c3\u03b5\u03b9\u03c2. \u0397 \u03bf\u03bc\u03ac\u03b4\u03b1 \u03ba\u03b1\u03c4\u03ad\u03b3\u03c1\u03b1\u03c8\u03b5 \u03b5\u03c0\u03af\u03c3\u03b7\u03c2 16 failure tags, \u03bc\u03b5\u03c4\u03b1\u03be\u03cd \u03ac\u03bb\u03bb\u03c9\u03bd fabricated results, missing artifacts, \u03c3\u03c4\u03b1\u03c4\u03b9\u03c3\u03c4\u03b9\u03ba\u03ae \u03ba\u03b1\u03ba\u03bf\u03c0\u03c1\u03b1\u03ba\u03c4\u03b9\u03ba\u03ae, \u03c0\u03c1\u03cc\u03c9\u03c1\u03b7 \u03bf\u03bb\u03bf\u03ba\u03bb\u03ae\u03c1\u03c9\u03c3\u03b7, shallow analysis \u03ba\u03b1\u03b9 overclaiming.<\/p>\n<div class=\"td-chart td-chart--metrics\">\n<div class=\"td-chart-head\">\n<p class=\"td-chart-title\">\u03a4\u03ad\u03c3\u03c3\u03b5\u03c1\u03b9\u03c2 \u03b1\u03c1\u03b9\u03b8\u03bc\u03bf\u03af \u03c0\u03bf\u03c5 \u03bf\u03c1\u03af\u03b6\u03bf\u03c5\u03bd \u03c4\u03bf K-Bench 01<\/p>\n<p class=\"td-chart-subtitle\">\u039c\u03b5\u03c4\u03c1\u03ae\u03c3\u03b5\u03b9\u03c2 \u03c4\u03bf\u03c5 \u03c3\u03c5\u03b3\u03ba\u03b5\u03ba\u03c1\u03b9\u03bc\u03ad\u03bd\u03bf\u03c5 traffic-derived benchmark\u00b7 \u03b4\u03b5\u03bd \u03b5\u03af\u03bd\u03b1\u03b9 \u03ba\u03b1\u03b8\u03bf\u03bb\u03b9\u03ba\u03ac \u03c0\u03bf\u03c3\u03bf\u03c3\u03c4\u03ac \u03b1\u03c0\u03cc\u03b4\u03bf\u03c3\u03b7\u03c2 \u03cc\u03bb\u03c9\u03bd \u03c4\u03c9\u03bd AI agents.<\/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\">178 tasks<\/span><span class=\"td-metric-label\">\u03c0\u03c1\u03ce\u03c4\u03b1 \u03b1\u03b9\u03c4\u03ae\u03bc\u03b1\u03c4\u03b1 \u03b1\u03c0\u03cc \u03c0\u03c1\u03b1\u03b3\u03bc\u03b1\u03c4\u03b9\u03ba\u03ad\u03c2 \u03b5\u03c0\u03b9\u03c3\u03c4\u03b7\u03bc\u03bf\u03bd\u03b9\u03ba\u03ad\u03c2 \u03c3\u03c5\u03bd\u03b5\u03b4\u03c1\u03af\u03b5\u03c2<\/span><\/div>\n<div class=\"td-metric-card\"><span class=\"td-metric-value\">9 \u03bc\u03bf\u03bd\u03c4\u03ad\u03bb\u03b1<\/span><span class=\"td-metric-label\">\u03c3\u03c4\u03bf \u03af\u03b4\u03b9\u03bf stock harness \u03ba\u03b1\u03b9 \u03c3\u03b5 \u03af\u03b4\u03b9\u03b1 sandboxes<\/span><\/div>\n<div class=\"td-metric-card\"><span class=\"td-metric-value\">1.602 runs<\/span><span class=\"td-metric-label\">\u03ad\u03bd\u03b1 run \u03b1\u03bd\u03ac \u03bc\u03bf\u03bd\u03c4\u03ad\u03bb\u03bf \u03ba\u03b1\u03b9 task, \u03c7\u03c9\u03c1\u03af\u03c2 retry<\/span><\/div>\n<div class=\"td-metric-card\"><span class=\"td-metric-value\">4.806 \u03ba\u03c1\u03af\u03c3\u03b5\u03b9\u03c2<\/span><span class=\"td-metric-label\">\u03c4\u03c1\u03b5\u03b9\u03c2 \u03c4\u03c5\u03c6\u03bb\u03bf\u03af LLM judges \u03b3\u03b9\u03b1 \u03ba\u03ac\u03b8\u03b5 \u03bf\u03bb\u03bf\u03ba\u03bb\u03b7\u03c1\u03c9\u03bc\u03ad\u03bd\u03bf run<\/span><\/div>\n<\/div>\n<\/div>\n<\/div>\n<p>\u03a4\u03b1 125 \u03b1\u03c0\u03cc \u03c4\u03b1 178 \u03b1\u03b9\u03c4\u03ae\u03bc\u03b1\u03c4\u03b1, \u03b4\u03b7\u03bb\u03b1\u03b4\u03ae \u03c4\u03bf 70%, \u03c0\u03b5\u03c1\u03b9\u03b5\u03af\u03c7\u03b1\u03bd \u03c4\u03bf\u03c5\u03bb\u03ac\u03c7\u03b9\u03c3\u03c4\u03bf\u03bd \u03ad\u03bd\u03b1 attachment. \u039f\u03b9 \u03b4\u03cd\u03bf \u03c5\u03c0\u03cc \u03c3\u03c5\u03bd\u03b8\u03ae\u03ba\u03b7 \u03b4\u03b9\u03b1\u03c3\u03c4\u03ac\u03c3\u03b5\u03b9\u03c2 \u03c3\u03b7\u03bc\u03b5\u03b9\u03ce\u03bd\u03bf\u03bd\u03c4\u03b1\u03bd \u03c9\u03c2 N\/A \u03cc\u03c4\u03b1\u03bd \u03b4\u03b5\u03bd \u03c5\u03c0\u03ae\u03c1\u03c7\u03b1\u03bd \u03b4\u03b5\u03b4\u03bf\u03bc\u03ad\u03bd\u03b1 \u03ae \u03cc\u03c4\u03b1\u03bd \u03b7 \u03c0\u03c1\u03cc\u03b6\u03b1 \u03ae\u03c4\u03b1\u03bd \u03c4\u03bf \u03c6\u03c5\u03c3\u03b9\u03ba\u03cc \u03c0\u03b1\u03c1\u03b1\u03b4\u03bf\u03c4\u03ad\u03bf. \u0395\u03c0\u03bf\u03bc\u03ad\u03bd\u03c9\u03c2, \u03c4\u03bf \u03b5\u03cd\u03c1\u03b7\u03bc\u03b1 \u03cc\u03c4\u03b9 \u03ba\u03ac\u03c0\u03bf\u03b9\u03b1 runs \u03b4\u03b5\u03bd \u03c0\u03b1\u03c1\u03ae\u03b3\u03b1\u03b3\u03b1\u03bd \u03b1\u03c1\u03c7\u03b5\u03af\u03b1 \u03b4\u03b5\u03bd \u03c3\u03b7\u03bc\u03b1\u03af\u03bd\u03b5\u03b9 \u03cc\u03c4\u03b9 \u03ba\u03ac\u03b8\u03b5 \u03c4\u03ad\u03c4\u03bf\u03b9\u03bf run \u03b1\u03c0\u03ad\u03c4\u03c5\u03c7\u03b5\u00b7 \u03c7\u03c1\u03b5\u03b9\u03ac\u03b6\u03b5\u03c4\u03b1\u03b9 \u03bd\u03b1 \u03b4\u03b9\u03b1\u03b2\u03b1\u03c3\u03c4\u03b5\u03af \u03bc\u03b1\u03b6\u03af \u03bc\u03b5 \u03c4\u03bf \u03b1\u03af\u03c4\u03b7\u03bc\u03b1 \u03ba\u03b1\u03b9 \u03c4\u03bf rubric.<\/p>\n<h2 id=\"anoixti-koryfi\">\u0397 \u03ba\u03bf\u03c1\u03c5\u03c6\u03ae \u03c4\u03b7\u03c2 \u03ba\u03b1\u03c4\u03ac\u03c4\u03b1\u03be\u03b7\u03c2 \u03ad\u03bc\u03b5\u03b9\u03bd\u03b5 \u03b1\u03bd\u03bf\u03b9\u03c7\u03c4\u03ae<\/h2>\n<p>\u03a4\u03bf gpt-5.6-sol \u03b5\u03af\u03c7\u03b5 \u03c4\u03bf\u03bd \u03c5\u03c8\u03b7\u03bb\u03cc\u03c4\u03b5\u03c1\u03bf pooled \u03bc\u03ad\u03c3\u03bf \u03cc\u03c1\u03bf, 8,04, \u03b1\u03bb\u03bb\u03ac \u03c4\u03bf 95% bootstrap interval 7,80\u20138,23 \u03c0\u03b5\u03c1\u03bd\u03bf\u03cd\u03c3\u03b5 \u03b1\u03c0\u03cc \u03c4\u03bf \u03cc\u03c1\u03b9\u03bf \u03c4\u03bf\u03c5 8. \u0394\u03cd\u03bf \u03b1\u03c0\u03cc \u03c4\u03bf\u03c5\u03c2 \u03c4\u03c1\u03b5\u03b9\u03c2 \u03ba\u03c1\u03b9\u03c4\u03ad\u03c2 \u03ad\u03b2\u03b1\u03bb\u03b1\u03bd \u03c0\u03c1\u03ce\u03c4\u03bf \u03c4\u03bf claude-opus-5. \u039f\u03b9 \u03c3\u03c5\u03b3\u03b3\u03c1\u03b1\u03c6\u03b5\u03af\u03c2 \u03b5\u03c0\u03bf\u03bc\u03ad\u03bd\u03c9\u03c2 \u03b8\u03b5\u03c9\u03c1\u03bf\u03cd\u03bd \u03c0\u03b9\u03bf \u03b1\u03bd\u03b1\u03c0\u03b1\u03c1\u03b1\u03b3\u03ce\u03b3\u03b9\u03bc\u03b7 \u03c4\u03b7 \u03c3\u03c7\u03b5\u03c4\u03b9\u03ba\u03ae \u03c3\u03b5\u03b9\u03c1\u03ac \u03b1\u03c0\u03cc \u03c4\u03bf \u03b1\u03c0\u03cc\u03bb\u03c5\u03c4\u03bf \u03b5\u03c0\u03af\u03c0\u03b5\u03b4\u03bf \u03ba\u03b1\u03b9 \u03c7\u03b1\u03c1\u03b1\u03ba\u03c4\u03b7\u03c1\u03af\u03b6\u03bf\u03c5\u03bd \u03c4\u03b7\u03bd \u03ba\u03bf\u03c1\u03c5\u03c6\u03ae \u03bc\u03b7 \u03b5\u03c0\u03b9\u03bb\u03c5\u03bc\u03ad\u03bd\u03b7.<\/p>\n<p>\u0397 \u03b4\u03b9\u03ac\u03ba\u03c1\u03b9\u03c3\u03b7 \u03b5\u03af\u03bd\u03b1\u03b9 \u03bf\u03c5\u03c3\u03b9\u03ce\u03b4\u03b7\u03c2. \u039f\u03b9 \u03ba\u03c1\u03b9\u03c4\u03ad\u03c2 \u03c3\u03c5\u03bc\u03c6\u03c9\u03bd\u03bf\u03cd\u03c3\u03b1\u03bd \u03b1\u03c1\u03ba\u03b5\u03c4\u03ac \u03b3\u03b9\u03b1 \u03c4\u03b7 \u03c3\u03b5\u03b9\u03c1\u03ac, \u03b1\u03bb\u03bb\u03ac \u03cc\u03c7\u03b9 \u03b3\u03b9\u03b1 \u03c4\u03bf \u03c0\u03cc\u03c3\u03bf \u03c8\u03b7\u03bb\u03ac \u03b2\u03c1\u03b9\u03c3\u03ba\u03cc\u03c4\u03b1\u03bd \u03b7 \u03ba\u03bb\u03af\u03bc\u03b1\u03ba\u03b1. \u039f \u03b1\u03c5\u03c3\u03c4\u03b7\u03c1\u03cc\u03c4\u03b5\u03c1\u03bf\u03c2 \u03ad\u03ba\u03c1\u03b9\u03bd\u03b5 \u03c0\u03bb\u03ae\u03c1\u03c9\u03c2 \u03b5\u03c0\u03b9\u03c4\u03c5\u03c7\u03b7\u03bc\u03ad\u03bd\u03bf \u03c4\u03bf 22,2% \u03c4\u03c9\u03bd runs \u03ba\u03b1\u03b9 \u03bf \u03c0\u03b9\u03bf \u03b5\u03c0\u03b9\u03b5\u03b9\u03ba\u03ae\u03c2 \u03c4\u03bf 48,9%. \u0386\u03c1\u03b1 \u03ad\u03bd\u03b1 leaderboard \u03c7\u03c9\u03c1\u03af\u03c2 \u03b1\u03bd\u03ac\u03bb\u03c5\u03c3\u03b7 \u03c4\u03bf\u03c5 judge panel, \u03c4\u03bf\u03c5 rubric \u03ba\u03b1\u03b9 \u03c4\u03bf\u03c5 uncertainty \u03bc\u03c0\u03bf\u03c1\u03b5\u03af \u03bd\u03b1 \u03b4\u03b7\u03bc\u03b9\u03bf\u03c5\u03c1\u03b3\u03ae\u03c3\u03b5\u03b9 \u03c8\u03b5\u03c5\u03b4\u03b1\u03af\u03c3\u03b8\u03b7\u03c3\u03b7 \u03b1\u03ba\u03c1\u03af\u03b2\u03b5\u03b9\u03b1\u03c2.<\/p>\n<aside class=\"td-article-note\"><strong>\u03a4\u03bf 8\/10 \u03b5\u03af\u03bd\u03b1\u03b9 anchor \u03c4\u03bf\u03c5 rubric, \u03cc\u03c7\u03b9 \u03bc\u03b5\u03c4\u03c1\u03b7\u03bc\u03ad\u03bd\u03bf \u03b1\u03bd\u03b8\u03c1\u03ce\u03c0\u03b9\u03bd\u03bf baseline:<\/strong> \u03ba\u03b1\u03bd\u03ad\u03bd\u03b1\u03c2 domain scientist \u03b4\u03b5\u03bd \u03b2\u03b1\u03b8\u03bc\u03bf\u03bb\u03cc\u03b3\u03b7\u03c3\u03b5 \u03c4\u03b1 runs \u03ba\u03b1\u03b9 \u03b4\u03b5\u03bd \u03c5\u03c0\u03ac\u03c1\u03c7\u03b5\u03b9 expert performance \u03c3\u03c4\u03bf \u03af\u03b4\u03b9\u03bf task set. \u03a4\u03bf score \u03c0\u03b5\u03c1\u03b9\u03b3\u03c1\u03ac\u03c6\u03b5\u03b9 panel-assessed compliance, \u03cc\u03c7\u03b9 \u03b1\u03c0\u03cc\u03c3\u03c4\u03b1\u03c3\u03b7 \u03b1\u03c0\u03cc \u03c4\u03b7\u03bd \u03c0\u03c1\u03b1\u03b3\u03bc\u03b1\u03c4\u03b9\u03ba\u03ae \u03b1\u03bd\u03b8\u03c1\u03ce\u03c0\u03b9\u03bd\u03b7 \u03b5\u03c0\u03b9\u03c3\u03c4\u03b7\u03bc\u03bf\u03bd\u03b9\u03ba\u03ae \u03b5\u03c1\u03b3\u03b1\u03c3\u03af\u03b1.<\/aside>\n<p>\u0393\u03b9\u03b1 procurement, \u03c4\u03bf \u03c3\u03c9\u03c3\u03c4\u03cc \u03b5\u03c1\u03ce\u03c4\u03b7\u03bc\u03b1 \u03b4\u03b5\u03bd \u03b5\u03af\u03bd\u03b1\u03b9 \u03bc\u03cc\u03bd\u03bf \u00ab\u03c0\u03bf\u03b9\u03bf \u03bc\u03bf\u03bd\u03c4\u03ad\u03bb\u03bf \u03b2\u03b3\u03ae\u03ba\u03b5 \u03c0\u03c1\u03ce\u03c4\u03bf;\u00bb. \u03a7\u03c1\u03b5\u03b9\u03ac\u03b6\u03b5\u03c4\u03b1\u03b9 \u03bd\u03b1 \u03b5\u03be\u03b5\u03c4\u03b1\u03c3\u03c4\u03b5\u03af \u03c0\u03bf\u03b9\u03bf failure profile \u03ad\u03c7\u03b5\u03b9 \u03ba\u03ac\u03b8\u03b5 \u03c3\u03cd\u03c3\u03c4\u03b7\u03bc\u03b1, \u03bc\u03b5 \u03c0\u03bf\u03b9\u03bf\u03c5\u03c2 \u03ba\u03c1\u03b9\u03c4\u03ad\u03c2, \u03c3\u03b5 \u03c0\u03bf\u03b9\u03bf harness \u03ba\u03b1\u03b9 \u03c3\u03b5 \u03c0\u03bf\u03b9\u03b1 \u03b5\u03ba\u03b4\u03bf\u03c7\u03ae. \u03a4\u03b1 <a href=\"https:\/\/twodots.gr\/model-cards-open-weight-ai-governance\/\">model cards \u03ba\u03b1\u03b9 \u03b7 AI governance<\/a> \u03ad\u03c7\u03bf\u03c5\u03bd \u03b1\u03be\u03af\u03b1 \u03cc\u03c4\u03b1\u03bd \u03c3\u03c5\u03bd\u03b4\u03ad\u03bf\u03c5\u03bd \u03c4\u03b1 scores \u03bc\u03b5 protocol, \u03b7\u03bc\u03b5\u03c1\u03bf\u03bc\u03b7\u03bd\u03af\u03b1, thresholds, \u03b1\u03b2\u03b5\u03b2\u03b1\u03b9\u03cc\u03c4\u03b7\u03c4\u03b1 \u03ba\u03b1\u03b9 \u03b3\u03bd\u03c9\u03c3\u03c4\u03ac \u03ba\u03b5\u03bd\u03ac.<\/p>\n<h2 id=\"epikoinonia-akriveia\">\u0397 \u03b5\u03c0\u03b9\u03ba\u03bf\u03b9\u03bd\u03c9\u03bd\u03af\u03b1 \u03c0\u03c1\u03bf\u03b7\u03b3\u03b5\u03af\u03c4\u03b1\u03b9 \u03c4\u03b7\u03c2 \u03b5\u03c0\u03b9\u03c3\u03c4\u03b7\u03bc\u03bf\u03bd\u03b9\u03ba\u03ae\u03c2 \u03b1\u03ba\u03c1\u03af\u03b2\u03b5\u03b9\u03b1\u03c2<\/h2>\n<p>\u03a3\u03c4\u03b9\u03c2 39.934 \u03b5\u03c6\u03b1\u03c1\u03bc\u03bf\u03c3\u03bc\u03ad\u03bd\u03b5\u03c2 \u03b2\u03b1\u03b8\u03bc\u03bf\u03bb\u03bf\u03b3\u03ae\u03c3\u03b5\u03b9\u03c2, \u03c4\u03bf 47,6% \u03b2\u03c1\u03ad\u03b8\u03b7\u03ba\u03b5 \u03ba\u03ac\u03c4\u03c9 \u03b1\u03c0\u03cc \u03c4\u03bf \u03cc\u03c1\u03b9\u03bf 8. \u0397 \u03b5\u03c0\u03b9\u03c3\u03c4\u03b7\u03bc\u03bf\u03bd\u03b9\u03ba\u03ae \u03b1\u03ba\u03c1\u03af\u03b2\u03b5\u03b9\u03b1 \u03b5\u03af\u03c7\u03b5 \u03bc\u03ad\u03c3\u03bf 6,22 \u03ba\u03b1\u03b9 \u03b7 \u03b5\u03c0\u03b9\u03ba\u03bf\u03b9\u03bd\u03c9\u03bd\u03af\u03b1 7,33 \u03bc\u03b5 \u03af\u03b4\u03b9\u03bf \u03c0\u03b1\u03c1\u03bf\u03bd\u03bf\u03bc\u03b1\u03c3\u03c4\u03ae. \u03a4\u03bf \u03ba\u03b5\u03bd\u03cc \u03c4\u03c9\u03bd 1,11 \u03bc\u03bf\u03bd\u03ac\u03b4\u03c9\u03bd \u03b5\u03bc\u03c6\u03b1\u03bd\u03af\u03c3\u03c4\u03b7\u03ba\u03b5 \u03c3\u03c4\u03b7\u03bd \u03af\u03b4\u03b9\u03b1 \u03ba\u03b1\u03c4\u03b5\u03cd\u03b8\u03c5\u03bd\u03c3\u03b7 \u03ba\u03b1\u03b9 \u03c3\u03c4\u03b1 \u03b5\u03bd\u03bd\u03ad\u03b1 \u03bc\u03bf\u03bd\u03c4\u03ad\u03bb\u03b1.<\/p>\n<p>\u039f\u03b9 \u03c3\u03c5\u03b3\u03b3\u03c1\u03b1\u03c6\u03b5\u03af\u03c2 \u03bf\u03bc\u03b1\u03b4\u03bf\u03c0\u03bf\u03af\u03b7\u03c3\u03b1\u03bd \u03c7\u03c1\u03ae\u03c3\u03b7 \u03b5\u03c1\u03b3\u03b1\u03bb\u03b5\u03af\u03c9\u03bd, \u03b5\u03c0\u03b9\u03ba\u03bf\u03b9\u03bd\u03c9\u03bd\u03af\u03b1 \u03ba\u03b1\u03b9 reasoning \u03c9\u03c2 execution dimensions \u03bc\u03b5 \u03bc\u03ad\u03c3\u03bf 6,93. \u0397 \u03b5\u03c0\u03b9\u03c3\u03c4\u03b7\u03bc\u03bf\u03bd\u03b9\u03ba\u03ae \u03b1\u03ba\u03c1\u03af\u03b2\u03b5\u03b9\u03b1, \u03b7 \u03b5\u03b9\u03bb\u03b9\u03ba\u03c1\u03af\u03bd\u03b5\u03b9\u03b1 \u03ba\u03b1\u03b9 \u03b7 \u03c0\u03bf\u03b9\u03cc\u03c4\u03b7\u03c4\u03b1 artifacts \u03c9\u03c2 substance dimensions \u03b5\u03af\u03c7\u03b1\u03bd \u03bc\u03ad\u03c3\u03bf 6,34. \u0397 \u03bf\u03bc\u03b1\u03b4\u03bf\u03c0\u03bf\u03af\u03b7\u03c3\u03b7 \u03b5\u03af\u03bd\u03b1\u03b9 \u03b5\u03c1\u03bc\u03b7\u03bd\u03b5\u03c5\u03c4\u03b9\u03ba\u03ae \u03ba\u03b1\u03b9 \u03c4\u03bf \u03c7\u03ac\u03c3\u03bc\u03b1 0,59 \u03b5\u03be\u03b1\u03c1\u03c4\u03ac\u03c4\u03b1\u03b9 \u03b1\u03c0\u03cc \u03b1\u03c5\u03c4\u03ae\u03bd, \u03b5\u03bd\u03ce \u03b7 \u03b1\u03c0\u03b5\u03c5\u03b8\u03b5\u03af\u03b1\u03c2 \u03c3\u03cd\u03b3\u03ba\u03c1\u03b9\u03c3\u03b7 \u03b1\u03ba\u03c1\u03af\u03b2\u03b5\u03b9\u03b1\u03c2 \u03ba\u03b1\u03b9 \u03b5\u03c0\u03b9\u03ba\u03bf\u03b9\u03bd\u03c9\u03bd\u03af\u03b1\u03c2 \u03b5\u03af\u03bd\u03b1\u03b9 \u03b9\u03c3\u03c7\u03c5\u03c1\u03cc\u03c4\u03b5\u03c1\u03b7 \u03b5\u03c0\u03b5\u03b9\u03b4\u03ae \u03c7\u03c1\u03b7\u03c3\u03b9\u03bc\u03bf\u03c0\u03bf\u03b9\u03b5\u03af \u03b1\u03ba\u03c1\u03b9\u03b2\u03ce\u03c2 \u03c4\u03b1 \u03af\u03b4\u03b9\u03b1 4.806 assessments.<\/p>\n<p>\u0393\u03b9\u03b1 \u03b5\u03c0\u03b9\u03c7\u03b5\u03b9\u03c1\u03ae\u03c3\u03b5\u03b9\u03c2 \u03c0\u03bf\u03c5 \u03c7\u03c1\u03b7\u03c3\u03b9\u03bc\u03bf\u03c0\u03bf\u03b9\u03bf\u03cd\u03bd agents \u03c3\u03b5 analytics, research, reporting \u03ae \u03b1\u03c0\u03bf\u03c6\u03ac\u03c3\u03b5\u03b9\u03c2, \u03c4\u03bf \u03b5\u03cd\u03c1\u03b7\u03bc\u03b1 \u03bc\u03b5\u03c4\u03b1\u03ba\u03b9\u03bd\u03b5\u03af \u03c4\u03bf acceptance test. \u0397 \u03c6\u03c5\u03c3\u03b9\u03ba\u03ae \u03b3\u03bb\u03ce\u03c3\u03c3\u03b1 \u03b1\u03be\u03b9\u03bf\u03bb\u03bf\u03b3\u03b5\u03af\u03c4\u03b1\u03b9 \u03b3\u03b9\u03b1 \u03c3\u03b1\u03c6\u03ae\u03bd\u03b5\u03b9\u03b1, \u03b1\u03bb\u03bb\u03ac \u03ba\u03ac\u03b8\u03b5 claim \u03c0\u03c1\u03ad\u03c0\u03b5\u03b9 \u03bd\u03b1 \u03bc\u03c0\u03bf\u03c1\u03b5\u03af \u03bd\u03b1 \u03c3\u03c5\u03bd\u03b4\u03b5\u03b8\u03b5\u03af \u03bc\u03b5 \u03b4\u03b5\u03b4\u03bf\u03bc\u03ad\u03bd\u03b1, \u03c5\u03c0\u03bf\u03bb\u03bf\u03b3\u03b9\u03c3\u03bc\u03cc, \u03c0\u03b7\u03b3\u03ae \u03ae \u03b1\u03c1\u03c7\u03b5\u03af\u03bf. \u0397 \u03c0\u03b5\u03b9\u03c3\u03c4\u03b9\u03ba\u03ae \u03bc\u03bf\u03c1\u03c6\u03ae \u03b4\u03b5\u03bd \u03c0\u03c1\u03ad\u03c0\u03b5\u03b9 \u03bd\u03b1 \u03b1\u03bd\u03b5\u03b2\u03ac\u03b6\u03b5\u03b9 \u03bc\u03cc\u03bd\u03b7 \u03c4\u03b7\u03c2 \u03c4\u03bf confidence \u03b5\u03bd\u03cc\u03c2 reviewer.<\/p>\n<h2 id=\"artifacts\">\u03a4\u03b1 artifacts \u03b5\u03af\u03bd\u03b1\u03b9 \u03bc\u03ad\u03c1\u03bf\u03c2 \u03c4\u03b7\u03c2 \u03b1\u03c0\u03ac\u03bd\u03c4\u03b7\u03c3\u03b7\u03c2<\/h2>\n<p>\u0397 \u03c0\u03bf\u03b9\u03cc\u03c4\u03b7\u03c4\u03b1 artifacts \u03ae\u03c4\u03b1\u03bd \u03b7 \u03c7\u03b1\u03bc\u03b7\u03bb\u03cc\u03c4\u03b5\u03c1\u03b7 \u03b4\u03b9\u03ac\u03c3\u03c4\u03b1\u03c3\u03b7 \u03c3\u03c5\u03bd\u03bf\u03bb\u03b9\u03ba\u03ac, \u03bc\u03b5 \u03bc\u03ad\u03c3\u03bf 5,50, \u03ba\u03b1\u03b9 \u03b7 \u03b1\u03c3\u03b8\u03b5\u03bd\u03ad\u03c3\u03c4\u03b5\u03c1\u03b7 \u03b3\u03b9\u03b1 \u03b5\u03c0\u03c4\u03ac \u03b1\u03c0\u03cc \u03c4\u03b1 \u03b5\u03bd\u03bd\u03ad\u03b1 \u03bc\u03bf\u03bd\u03c4\u03ad\u03bb\u03b1. \u0395\u03c0\u03b9\u03c0\u03bb\u03ad\u03bf\u03bd, 767 \u03b1\u03c0\u03cc \u03c4\u03b1 1.602 runs, \u03b4\u03b7\u03bb\u03b1\u03b4\u03ae \u03c4\u03bf 47,9%, \u03c4\u03b5\u03bb\u03b5\u03af\u03c9\u03c3\u03b1\u03bd \u03c7\u03c9\u03c1\u03af\u03c2 \u03ba\u03b1\u03bd\u03ad\u03bd\u03b1 \u03b1\u03c1\u03c7\u03b5\u03af\u03bf \u03c3\u03c4\u03bf output tree. \u03a4\u03b1 runs \u03c7\u03c9\u03c1\u03af\u03c2 \u03b1\u03c1\u03c7\u03b5\u03af\u03bf \u03b5\u03af\u03c7\u03b1\u03bd \u03bc\u03ad\u03c3\u03bf overall score 5,26, \u03b5\u03bd\u03ce \u03cc\u03c3\u03b1 \u03ac\u03c6\u03b7\u03c3\u03b1\u03bd \u03ba\u03ac\u03c4\u03b9 \u03c0\u03af\u03c3\u03c9 \u03b5\u03af\u03c7\u03b1\u03bd 6,68.<\/p>\n<p>\u03a4\u03bf \u03b1\u03c0\u03bf\u03c4\u03ad\u03bb\u03b5\u03c3\u03bc\u03b1 \u03b4\u03b5\u03bd \u03b5\u03c0\u03b9\u03b2\u03c1\u03b1\u03b2\u03b5\u03cd\u03b5\u03b9 \u03b1\u03c0\u03bb\u03ce\u03c2 \u03c4\u03b7\u03bd \u03c0\u03bf\u03c3\u03cc\u03c4\u03b7\u03c4\u03b1. Runs \u03bc\u03b5 \u03ad\u03bd\u03b1 \u03ae \u03b4\u03cd\u03bf \u03b1\u03c1\u03c7\u03b5\u03af\u03b1 \u03b5\u03af\u03c7\u03b1\u03bd \u03bc\u03ad\u03c3\u03bf 6,87 \u03ba\u03b1\u03b9 runs \u03bc\u03b5 11 \u03ae \u03c0\u03b5\u03c1\u03b9\u03c3\u03c3\u03cc\u03c4\u03b5\u03c1\u03b1 6,64. \u0397 \u03bf\u03c5\u03c3\u03b9\u03b1\u03c3\u03c4\u03b9\u03ba\u03ae \u03b1\u03c3\u03c5\u03bd\u03ad\u03c7\u03b5\u03b9\u03b1 \u03b2\u03c1\u03b9\u03c3\u03ba\u03cc\u03c4\u03b1\u03bd \u03b1\u03bd\u03ac\u03bc\u03b5\u03c3\u03b1 \u03c3\u03c4\u03bf \u00ab\u03c4\u03af\u03c0\u03bf\u03c4\u03b1\u00bb \u03ba\u03b1\u03b9 \u03c3\u03c4\u03bf \u00ab\u03ba\u03ac\u03c4\u03b9\u00bb, \u03cc\u03c7\u03b9 \u03b1\u03bd\u03ac\u03bc\u03b5\u03c3\u03b1 \u03c3\u03b5 \u03bb\u03af\u03b3\u03b1 \u03ba\u03b1\u03b9 \u03c0\u03bf\u03bb\u03bb\u03ac \u03b1\u03c1\u03c7\u03b5\u03af\u03b1. \u0388\u03bd\u03b1 artifact \u03ad\u03c7\u03b5\u03b9 \u03b1\u03be\u03af\u03b1 \u03bc\u03cc\u03bd\u03bf \u03cc\u03c4\u03b1\u03bd \u03b5\u03af\u03bd\u03b1\u03b9 \u03c3\u03c9\u03c3\u03c4\u03cc, \u03c0\u03bb\u03ae\u03c1\u03b5\u03c2 \u03ba\u03b1\u03b9 \u03c7\u03c1\u03ae\u03c3\u03b9\u03bc\u03bf.<\/p>\n<p>\u0393\u03b9\u03b1 \u03ad\u03bd\u03b1 e-commerce \u03ae marketing team, \u03c4\u03bf \u03b1\u03bd\u03c4\u03af\u03c3\u03c4\u03bf\u03b9\u03c7\u03bf \u03c4\u03b5\u03c3\u03c4 \u03b5\u03af\u03bd\u03b1\u03b9 \u03c0\u03c1\u03b1\u03ba\u03c4\u03b9\u03ba\u03cc: \u03b1\u03bd\u03bf\u03af\u03b3\u03b5\u03b9 \u03c4\u03bf CSV; \u03ad\u03c7\u03b5\u03b9 \u03c4\u03b9\u03c2 \u03c3\u03c9\u03c3\u03c4\u03ad\u03c2 \u03c3\u03c4\u03ae\u03bb\u03b5\u03c2; \u03c0\u03b5\u03c1\u03b9\u03bb\u03b1\u03bc\u03b2\u03ac\u03bd\u03b5\u03b9 \u03cc\u03bb\u03bf \u03c4\u03bf \u03b6\u03b7\u03c4\u03bf\u03cd\u03bc\u03b5\u03bd\u03bf date range; \u03b5\u03ba\u03c4\u03b5\u03bb\u03b5\u03af\u03c4\u03b1\u03b9 \u03bf \u03ba\u03ce\u03b4\u03b9\u03ba\u03b1\u03c2; \u03c3\u03c5\u03bc\u03c6\u03c9\u03bd\u03bf\u03cd\u03bd \u03c4\u03b1 totals \u03ba\u03b1\u03b9 \u03c4\u03b1 \u03b3\u03c1\u03b1\u03c6\u03ae\u03bc\u03b1\u03c4\u03b1 \u03bc\u03b5 \u03c4\u03b1 raw data; \u0397 \u03b5\u03c5\u03b3\u03ad\u03bd\u03b5\u03b9\u03b1 \u03ba\u03b1\u03b9 \u03b7 \u03ba\u03b1\u03bb\u03ae \u03b4\u03bf\u03bc\u03ae \u03c4\u03bf\u03c5 \u03bc\u03b7\u03bd\u03cd\u03bc\u03b1\u03c4\u03bf\u03c2 \u03b4\u03b5\u03bd \u03b1\u03c0\u03bf\u03b4\u03b5\u03b9\u03ba\u03bd\u03cd\u03bf\u03c5\u03bd \u03ba\u03b1\u03bd\u03ad\u03bd\u03b1 \u03b1\u03c0\u03cc \u03b1\u03c5\u03c4\u03ac.<\/p>\n<p>\u03a4\u03bf \u03af\u03b4\u03b9\u03bf \u03bc\u03ac\u03b8\u03b7\u03bc\u03b1 \u03b5\u03bc\u03c6\u03b1\u03bd\u03af\u03b6\u03b5\u03c4\u03b1\u03b9 \u03ba\u03b1\u03b9 \u03c3\u03b5 benchmarks web agents, \u03cc\u03c0\u03bf\u03c5 \u03c4\u03bf \u03c4\u03b5\u03bb\u03b9\u03ba\u03cc state \u03bc\u03b5\u03c4\u03c1\u03ac \u03c0\u03b5\u03c1\u03b9\u03c3\u03c3\u03cc\u03c4\u03b5\u03c1\u03bf \u03b1\u03c0\u03cc \u03bc\u03b9\u03b1 \u03b5\u03cd\u03bb\u03bf\u03b3\u03b7 \u03c0\u03b5\u03c1\u03b9\u03b3\u03c1\u03b1\u03c6\u03ae \u03b5\u03bd\u03b5\u03c1\u03b3\u03b5\u03b9\u03ce\u03bd. \u03a4\u03bf <a href=\"https:\/\/twodots.gr\/mag-neo-benchmark-ai-agents-sto-web\/\">MAG \u03b3\u03b9\u03b1 AI agents \u03c3\u03c4\u03bf web<\/a> \u03b4\u03b5\u03af\u03c7\u03bd\u03b5\u03b9 \u03b3\u03b9\u03b1\u03c4\u03af \u03b7 rendered \u03ba\u03b1\u03c4\u03ac\u03c3\u03c4\u03b1\u03c3\u03b7, \u03c4\u03b1 side effects \u03ba\u03b1\u03b9 \u03c4\u03bf \u03c0\u03c1\u03b1\u03b3\u03bc\u03b1\u03c4\u03b9\u03ba\u03cc \u03b1\u03c0\u03bf\u03c4\u03ad\u03bb\u03b5\u03c3\u03bc\u03b1 \u03c0\u03c1\u03ad\u03c0\u03b5\u03b9 \u03bd\u03b1 \u03b5\u03bb\u03ad\u03b3\u03c7\u03bf\u03bd\u03c4\u03b1\u03b9 \u03bc\u03b1\u03b6\u03af \u03bc\u03b5 \u03c4\u03bf reasoning trace.<\/p>\n<h2 id=\"overclaiming\">\u03a4\u03bf \u03c3\u03c5\u03c7\u03bd\u03cc\u03c4\u03b5\u03c1\u03bf failure mode \u03ae\u03c4\u03b1\u03bd \u03c4\u03bf overclaiming<\/h2>\n<p>\u03a4\u03bf overclaiming \u03b5\u03c0\u03b9\u03c3\u03b7\u03bc\u03ac\u03bd\u03b8\u03b7\u03ba\u03b5 \u03c3\u03c4\u03bf 31,4% \u03c4\u03c9\u03bd assessments, \u03c0\u03b5\u03c1\u03b9\u03c3\u03c3\u03cc\u03c4\u03b5\u03c1\u03bf \u03b1\u03c0\u03cc \u03ba\u03ac\u03b8\u03b5 \u03ac\u03bb\u03bb\u03bf failure tag. \u0391\u03ba\u03bf\u03bb\u03bf\u03cd\u03b8\u03b7\u03c3\u03b1\u03bd missing artifacts \u03c3\u03c4\u03bf 22,6%, shallow analysis \u03c3\u03c4\u03bf 17,7%, truncated run \u03c3\u03c4\u03bf 16,4% \u03ba\u03b1\u03b9 premature completion \u03c3\u03c4\u03bf 12,5%. \u03a4\u03b1 tags \u03b4\u03b5\u03bd \u03ae\u03c4\u03b1\u03bd \u03b1\u03bc\u03bf\u03b9\u03b2\u03b1\u03af\u03b1 \u03b1\u03c0\u03bf\u03ba\u03bb\u03b5\u03b9\u03cc\u03bc\u03b5\u03bd\u03b1 \u03ba\u03b1\u03b9 \u03c0\u03c1\u03bf\u03ad\u03c1\u03c7\u03bf\u03bd\u03c4\u03b1\u03bd \u03b1\u03c0\u03cc \u03c4\u03bf\u03c5\u03c2 \u03af\u03b4\u03b9\u03bf\u03c5\u03c2 judges \u03c0\u03bf\u03c5 \u03ad\u03b2\u03b1\u03b6\u03b1\u03bd \u03c4\u03b1 scores.<\/p>\n<p>\u03a3\u03b5 \u03b5\u03c0\u03af\u03c0\u03b5\u03b4\u03bf run, \u03c4\u03bf\u03c5\u03bb\u03ac\u03c7\u03b9\u03c3\u03c4\u03bf\u03bd \u03ad\u03bd\u03b1\u03c2 \u03b1\u03c0\u03cc \u03c4\u03bf\u03c5\u03c2 \u03c4\u03c1\u03b5\u03b9\u03c2 \u03ba\u03c1\u03b9\u03c4\u03ad\u03c2 \u03b5\u03c0\u03b9\u03c3\u03ae\u03bc\u03b1\u03bd\u03b5 overclaiming \u03c3\u03c4\u03bf 54,4% \u03c4\u03c9\u03bd runs, \u03b4\u03cd\u03bf \u03ae \u03c0\u03b5\u03c1\u03b9\u03c3\u03c3\u03cc\u03c4\u03b5\u03c1\u03bf\u03b9 \u03c3\u03c4\u03bf 26,9% \u03ba\u03b1\u03b9 \u03cc\u03bb\u03bf\u03b9 \u03bf\u03b9 \u03ba\u03c1\u03b9\u03c4\u03ad\u03c2 \u03c3\u03c4\u03bf 12,9%. \u0391\u03c5\u03c4\u03ad\u03c2 \u03bf\u03b9 \u03c4\u03b9\u03bc\u03ad\u03c2 \u03b4\u03b5\u03bd \u03b5\u03af\u03bd\u03b1\u03b9 \u03b1\u03bd\u03c4\u03b9\u03c6\u03ac\u03c3\u03b5\u03b9\u03c2: \u03b1\u03c0\u03b1\u03bd\u03c4\u03bf\u03cd\u03bd \u03c3\u03b5 \u03b4\u03b9\u03b1\u03c6\u03bf\u03c1\u03b5\u03c4\u03b9\u03ba\u03cc aggregation rule. \u0395\u03af\u03bd\u03b1\u03b9 \u03c5\u03c0\u03b5\u03bd\u03b8\u03cd\u03bc\u03b9\u03c3\u03b7 \u03cc\u03c4\u03b9 \u03ba\u03ac\u03b8\u03b5 failure rate \u03c7\u03c1\u03b5\u03b9\u03ac\u03b6\u03b5\u03c4\u03b1\u03b9 \u03c3\u03b1\u03c6\u03ae \u03bc\u03bf\u03bd\u03ac\u03b4\u03b1 \u03b1\u03bd\u03ac\u03bb\u03c5\u03c3\u03b7\u03c2.<\/p>\n<p>\u03a4\u03bf \u03c0\u03c1\u03cc\u03b2\u03bb\u03b7\u03bc\u03b1 \u03b5\u03af\u03bd\u03b1\u03b9 \u03b5\u03c0\u03b9\u03ba\u03af\u03bd\u03b4\u03c5\u03bd\u03bf \u03b1\u03ba\u03c1\u03b9\u03b2\u03ce\u03c2 \u03b5\u03c0\u03b5\u03b9\u03b4\u03ae \u03bc\u03c0\u03bf\u03c1\u03b5\u03af \u03bd\u03b1 \u03c3\u03c5\u03bd\u03c5\u03c0\u03ac\u03c1\u03c7\u03b5\u03b9 \u03bc\u03b5 \u03ba\u03b1\u03bb\u03ae \u03b5\u03c0\u03b9\u03ba\u03bf\u03b9\u03bd\u03c9\u03bd\u03af\u03b1. \u0388\u03bd\u03b1 \u03c3\u03cd\u03c3\u03c4\u03b7\u03bc\u03b1 \u03bc\u03c0\u03bf\u03c1\u03b5\u03af \u03bd\u03b1 \u03c0\u03b5\u03c1\u03b9\u03b3\u03c1\u03ac\u03c8\u03b5\u03b9 \u03bc\u03b9\u03b1 \u03b1\u03bd\u03ac\u03bb\u03c5\u03c3\u03b7 \u03c9\u03c2 \u03bf\u03bb\u03bf\u03ba\u03bb\u03b7\u03c1\u03c9\u03bc\u03ad\u03bd\u03b7 \u03b5\u03bd\u03ce \u03bb\u03b5\u03af\u03c0\u03bf\u03c5\u03bd \u03b1\u03c1\u03c7\u03b5\u03af\u03b1, \u03ad\u03c7\u03bf\u03c5\u03bd \u03bc\u03b5\u03af\u03bd\u03b5\u03b9 unchecked assumptions \u03ae \u03b7 \u03bc\u03ad\u03b8\u03bf\u03b4\u03bf\u03c2 \u03b4\u03b5\u03bd \u03c3\u03c4\u03b7\u03c1\u03af\u03b6\u03b5\u03b9 \u03c4\u03bf \u03c3\u03c5\u03bc\u03c0\u03ad\u03c1\u03b1\u03c3\u03bc\u03b1. \u0397 \u03ba\u03b1\u03c4\u03ac\u03bb\u03bb\u03b7\u03bb\u03b7 \u03b1\u03bd\u03c4\u03af\u03b4\u03c1\u03b1\u03c3\u03b7 \u03b4\u03b5\u03bd \u03b5\u03af\u03bd\u03b1\u03b9 \u03b3\u03b5\u03bd\u03b9\u03ba\u03ae \u03b4\u03c5\u03c3\u03c0\u03b9\u03c3\u03c4\u03af\u03b1 \u03c0\u03c1\u03bf\u03c2 \u03c4\u03b7\u03bd AI, \u03b1\u03bb\u03bb\u03ac traceability: \u03c3\u03cd\u03bd\u03b4\u03b5\u03c3\u03b7 \u03ba\u03ac\u03b8\u03b5 claim \u03bc\u03b5 \u03c0\u03b7\u03b3\u03ae, \u03c5\u03c0\u03bf\u03bb\u03bf\u03b3\u03b9\u03c3\u03bc\u03cc, tool output \u03ae \u03c0\u03b1\u03c1\u03b1\u03b4\u03bf\u03c4\u03ad\u03bf \u03c0\u03bf\u03c5 \u03bc\u03c0\u03bf\u03c1\u03b5\u03af \u03bd\u03b1 \u03b5\u03c0\u03b1\u03bb\u03b7\u03b8\u03b5\u03c5\u03c4\u03b5\u03af.<\/p>\n<h2 id=\"attachments-dyskolia\">Attachments \u03ba\u03b1\u03b9 \u03bc\u03b5\u03b3\u03ac\u03bb\u03b1 prompts: \u03c0\u03c1\u03b1\u03b3\u03bc\u03b1\u03c4\u03b9\u03ba\u03ae \u03b4\u03c5\u03c3\u03ba\u03bf\u03bb\u03af\u03b1, \u03cc\u03c7\u03b9 \u03b1\u03c0\u03bb\u03ae \u03b1\u03b9\u03c4\u03af\u03b1<\/h2>\n<p>\u03a4\u03bf benchmark \u03c0\u03b5\u03c1\u03b9\u03bb\u03ac\u03bc\u03b2\u03b1\u03bd\u03b5 \u03c0\u03c1\u03b1\u03b3\u03bc\u03b1\u03c4\u03b9\u03ba\u03ac \u03c3\u03c5\u03bd\u03b7\u03bc\u03bc\u03ad\u03bd\u03b1, \u03bc\u03b5 \u03c4\u03bf .docx \u03bd\u03b1 \u03b5\u03af\u03bd\u03b1\u03b9 \u03bf \u03c3\u03c5\u03c7\u03bd\u03cc\u03c4\u03b5\u03c1\u03bf\u03c2 \u03c4\u03cd\u03c0\u03bf\u03c2. \u039f\u03b9 \u03ba\u03c1\u03b9\u03c4\u03ad\u03c2 \u03ad\u03bb\u03b5\u03b3\u03c7\u03b1\u03bd \u03b1\u03bd \u03c4\u03b1 \u03b1\u03c1\u03c7\u03b5\u03af\u03b1 \u03c6\u03bf\u03c1\u03c4\u03ce\u03b8\u03b7\u03ba\u03b1\u03bd, \u03b1\u03bd\u03b1\u03bb\u03cd\u03b8\u03b7\u03ba\u03b1\u03bd \u03c3\u03c9\u03c3\u03c4\u03ac, \u03b5\u03bb\u03ad\u03b3\u03c7\u03b8\u03b7\u03ba\u03b1\u03bd \u03b3\u03b9\u03b1 \u03c3\u03c7\u03ae\u03bc\u03b1\u03c4\u03b1, \u03b5\u03cd\u03c1\u03b7 \u03ba\u03b1\u03b9 \u03b5\u03bb\u03bb\u03b5\u03af\u03c8\u03b5\u03b9\u03c2 \u03ba\u03b1\u03b9 \u03b1\u03bd \u03b7 \u03c4\u03b5\u03bb\u03b9\u03ba\u03ae \u03b1\u03bd\u03ac\u03bb\u03c5\u03c3\u03b7 \u03ad\u03bc\u03b5\u03b9\u03bd\u03b5 \u03c0\u03b9\u03c3\u03c4\u03ae \u03c3\u03c4\u03b1 \u03b4\u03b5\u03b4\u03bf\u03bc\u03ad\u03bd\u03b1. \u0391\u03c5\u03c4\u03cc \u03b4\u03b9\u03b1\u03c7\u03c9\u03c1\u03af\u03b6\u03b5\u03b9 \u03c4\u03b7\u03bd \u03b1\u03c0\u03bb\u03ae \u03c0\u03b1\u03c1\u03b1\u03b3\u03c9\u03b3\u03ae \u03ba\u03b5\u03b9\u03bc\u03ad\u03bd\u03bf\u03c5 \u03b1\u03c0\u03cc \u03c0\u03c1\u03b1\u03b3\u03bc\u03b1\u03c4\u03b9\u03ba\u03ae agentic \u03b5\u03c1\u03b3\u03b1\u03c3\u03af\u03b1.<\/p>\n<p>\u03a4\u03bf overclaiming \u03ae\u03c4\u03b1\u03bd \u03c3\u03c5\u03c7\u03bd\u03cc\u03c4\u03b5\u03c1\u03bf \u03c3\u03c4\u03b1 tasks \u03bc\u03b5 attachments: 33,9% \u03c4\u03c9\u03bd assessments \u03ad\u03bd\u03b1\u03bd\u03c4\u03b9 25,4% \u03c7\u03c9\u03c1\u03af\u03c2 \u03b1\u03c1\u03c7\u03b5\u03af\u03b1. \u0397 \u03c3\u03c5\u03c3\u03c7\u03ad\u03c4\u03b9\u03c3\u03b7 \u03b4\u03b5\u03bd \u03b1\u03c0\u03bf\u03b4\u03b5\u03b9\u03ba\u03bd\u03cd\u03b5\u03b9 \u03cc\u03c4\u03b9 \u03c4\u03bf attachment \u03c0\u03c1\u03bf\u03ba\u03b1\u03bb\u03b5\u03af \u03c5\u03c0\u03b5\u03c1\u03b2\u03bf\u03bb\u03ae. \u03a4\u03b1 requests \u03bc\u03b5 \u03b1\u03c1\u03c7\u03b5\u03af\u03b1 \u03bc\u03c0\u03bf\u03c1\u03bf\u03cd\u03bd \u03bd\u03b1 \u03b5\u03af\u03bd\u03b1\u03b9 \u03c4\u03b1\u03c5\u03c4\u03cc\u03c7\u03c1\u03bf\u03bd\u03b1 \u03c0\u03b9\u03bf \u03c3\u03cd\u03bd\u03b8\u03b5\u03c4\u03b1, \u03c0\u03b9\u03bf \u03b1\u03c3\u03b1\u03c6\u03ae \u03ba\u03b1\u03b9 \u03c0\u03b9\u03bf \u03b1\u03c0\u03b1\u03b9\u03c4\u03b7\u03c4\u03b9\u03ba\u03ac \u03c9\u03c2 \u03c0\u03c1\u03bf\u03c2 \u03c4\u03bf domain.<\/p>\n<p>\u0397 \u03bc\u03b5\u03bb\u03ad\u03c4\u03b7 \u03b5\u03be\u03b5\u03c4\u03ac\u03b6\u03b5\u03b9 \u03b5\u03c0\u03af\u03c3\u03b7\u03c2 \u03bc\u03ae\u03ba\u03bf\u03c2 prompt \u03ba\u03b1\u03b9 \u03b4\u03c5\u03c3\u03ba\u03bf\u03bb\u03af\u03b1, \u03b1\u03bb\u03bb\u03ac \u03b1\u03c0\u03bf\u03c6\u03b5\u03cd\u03b3\u03b5\u03b9 \u03ad\u03bd\u03b1\u03bd \u03b5\u03cd\u03ba\u03bf\u03bb\u03bf \u03b1\u03b9\u03c4\u03b9\u03ce\u03b4\u03b7 \u03b9\u03c3\u03c7\u03c5\u03c1\u03b9\u03c3\u03bc\u03cc. \u0397 \u03b4\u03c5\u03c3\u03ba\u03bf\u03bb\u03af\u03b1 \u03b5\u03bd\u03cc\u03c2 \u03c0\u03c1\u03b1\u03b3\u03bc\u03b1\u03c4\u03b9\u03ba\u03bf\u03cd request \u03b4\u03b5\u03bd \u03c0\u03b5\u03c1\u03b9\u03b3\u03c1\u03ac\u03c6\u03b5\u03c4\u03b1\u03b9 \u03b1\u03c0\u03cc \u03ad\u03bd\u03b1\u03bd \u03bc\u03cc\u03bd\u03bf \u03b1\u03c1\u03b9\u03b8\u03bc\u03cc: \u03bc\u03c0\u03bf\u03c1\u03b5\u03af \u03bd\u03b1 \u03c0\u03c1\u03bf\u03ba\u03cd\u03c0\u03c4\u03b5\u03b9 \u03b1\u03c0\u03cc \u03b1\u03c3\u03ac\u03c6\u03b5\u03b9\u03b1, \u03b5\u03b9\u03b4\u03b9\u03ba\u03cc\u03c4\u03b7\u03c4\u03b1 \u03c0\u03b5\u03b4\u03af\u03bf\u03c5, \u03c0\u03bf\u03bb\u03bb\u03ac \u03b1\u03c1\u03c7\u03b5\u03af\u03b1, \u03ba\u03c1\u03c5\u03bc\u03bc\u03ad\u03bd\u03bf\u03c5\u03c2 \u03c0\u03b5\u03c1\u03b9\u03bf\u03c1\u03b9\u03c3\u03bc\u03bf\u03cd\u03c2 \u03ae \u03c4\u03b7\u03bd \u03b1\u03bd\u03ac\u03b3\u03ba\u03b7 \u03c0\u03b1\u03c1\u03b1\u03b3\u03c9\u03b3\u03ae\u03c2 \u03bb\u03b5\u03b9\u03c4\u03bf\u03c5\u03c1\u03b3\u03b9\u03ba\u03bf\u03cd \u03b1\u03c0\u03bf\u03c4\u03b5\u03bb\u03ad\u03c3\u03bc\u03b1\u03c4\u03bf\u03c2.<\/p>\n<p>\u03a3\u03b5 \u03bc\u03b9\u03b1 \u03b5\u03c0\u03b9\u03c7\u03b5\u03b9\u03c1\u03b7\u03c3\u03b9\u03b1\u03ba\u03ae \u03c1\u03bf\u03ae, \u03bf\u03b9 \u03b4\u03bf\u03ba\u03b9\u03bc\u03ad\u03c2 \u03c0\u03c1\u03ad\u03c0\u03b5\u03b9 \u03bd\u03b1 \u03ba\u03c1\u03b1\u03c4\u03bf\u03cd\u03bd \u03c4\u03b7\u03bd \u03c0\u03bf\u03bb\u03c5\u03c0\u03bb\u03bf\u03ba\u03cc\u03c4\u03b7\u03c4\u03b1 \u03c0\u03bf\u03c5 \u03ad\u03c7\u03b5\u03b9 \u03c3\u03b7\u03bc\u03b1\u03c3\u03af\u03b1: \u03b5\u03bb\u03bb\u03b9\u03c0\u03ae exports, \u03c0\u03bf\u03bb\u03bb\u03b1\u03c0\u03bb\u03ad\u03c2 \u03c0\u03b7\u03b3\u03ad\u03c2, \u03b1\u03bb\u03bb\u03b1\u03b3\u03ad\u03c2 schema, \u03b1\u03bd\u03c4\u03b9\u03c6\u03b1\u03c4\u03b9\u03ba\u03ad\u03c2 \u03bf\u03b4\u03b7\u03b3\u03af\u03b5\u03c2 \u03ba\u03b1\u03b9 \u03b1\u03bd\u03ac\u03b3\u03ba\u03b7 \u03b1\u03bd\u03b8\u03c1\u03ce\u03c0\u03b9\u03bd\u03bf\u03c5 handoff. \u0397 <a href=\"https:\/\/twodots.gr\/consilience-ai-agent-communication-control\/\">\u03b5\u03bb\u03b5\u03b3\u03c7\u03cc\u03bc\u03b5\u03bd\u03b7 \u03b5\u03c0\u03b9\u03ba\u03bf\u03b9\u03bd\u03c9\u03bd\u03af\u03b1 \u03bc\u03b5\u03c4\u03b1\u03be\u03cd AI agents<\/a> \u03b2\u03bf\u03b7\u03b8\u03ac \u03bc\u03cc\u03bd\u03bf \u03cc\u03c4\u03b1\u03bd \u03ba\u03c1\u03af\u03c3\u03b9\u03bc\u03bf evidence \u03b2\u03b3\u03b1\u03af\u03bd\u03b5\u03b9 \u03c3\u03c4\u03b7\u03bd \u03b5\u03c0\u03b9\u03c6\u03ac\u03bd\u03b5\u03b9\u03b1 \u03ba\u03b1\u03b9 \u03b5\u03c0\u03b1\u03bb\u03b7\u03b8\u03b5\u03cd\u03b5\u03c4\u03b1\u03b9\u00b7 \u03c0\u03b5\u03c1\u03b9\u03c3\u03c3\u03cc\u03c4\u03b5\u03c1\u03b1 \u03bc\u03b7\u03bd\u03cd\u03bc\u03b1\u03c4\u03b1 \u03b4\u03b5\u03bd \u03b9\u03c3\u03bf\u03b4\u03c5\u03bd\u03b1\u03bc\u03bf\u03cd\u03bd \u03bc\u03b5 \u03bc\u03b5\u03b3\u03b1\u03bb\u03cd\u03c4\u03b5\u03c1\u03b7 \u03b1\u03ba\u03c1\u03af\u03b2\u03b5\u03b9\u03b1.<\/p>\n<h2 id=\"kostos-prospatheia\">\u03a4\u03bf \u03ba\u03cc\u03c3\u03c4\u03bf\u03c2 \u03ba\u03b1\u03b9 \u03b7 \u03c0\u03c1\u03bf\u03c3\u03c0\u03ac\u03b8\u03b5\u03b9\u03b1 \u03b4\u03b5\u03bd \u03b5\u03b3\u03b3\u03c5\u03ce\u03bd\u03c4\u03b1\u03b9 \u03c0\u03bf\u03b9\u03cc\u03c4\u03b7\u03c4\u03b1<\/h2>\n<p>\u0397 \u03b5\u03ba\u03c4\u03ad\u03bb\u03b5\u03c3\u03b7 \u03c3\u03b5 sandbox \u03b5\u03c0\u03b9\u03c4\u03c1\u03ad\u03c0\u03b5\u03b9 \u03bd\u03b1 \u03bc\u03b5\u03c4\u03c1\u03b7\u03b8\u03bf\u03cd\u03bd \u03c7\u03c1\u03cc\u03bd\u03bf\u03c2, \u03c7\u03c1\u03ae\u03c3\u03b7 \u03b5\u03c1\u03b3\u03b1\u03bb\u03b5\u03af\u03c9\u03bd \u03ba\u03b1\u03b9 \u03ba\u03cc\u03c3\u03c4\u03bf\u03c2 \u03bc\u03b1\u03b6\u03af \u03bc\u03b5 \u03c4\u03b7\u03bd \u03c0\u03bf\u03b9\u03cc\u03c4\u03b7\u03c4\u03b1. \u0397 \u03c3\u03c5\u03c3\u03c7\u03ad\u03c4\u03b9\u03c3\u03b7 \u03b1\u03bd\u03ac\u03bc\u03b5\u03c3\u03b1 \u03c3\u03c4\u03b1 tool calls \u03ba\u03b1\u03b9 \u03c3\u03c4\u03bf score \u03ae\u03c4\u03b1\u03bd \u03b8\u03b5\u03c4\u03b9\u03ba\u03ae \u03b1\u03bb\u03bb\u03ac \u03b1\u03c3\u03b8\u03b5\u03bd\u03ae\u03c2, \u03bc\u03b5 Spearman \u03c1=0,27. \u0388\u03bd\u03b1\u03c2 agent \u03bc\u03c0\u03bf\u03c1\u03b5\u03af \u03bd\u03b1 \u03ba\u03ac\u03bd\u03b5\u03b9 \u03c0\u03bf\u03bb\u03bb\u03ad\u03c2 \u03ba\u03bb\u03ae\u03c3\u03b5\u03b9\u03c2, \u03bd\u03b1 \u03c0\u03b5\u03c1\u03b9\u03c6\u03ad\u03c1\u03b5\u03c4\u03b1\u03b9 \u03b1\u03bd\u03ac\u03bc\u03b5\u03c3\u03b1 \u03c3\u03b5 \u03b1\u03c0\u03bf\u03c4\u03c5\u03c7\u03b7\u03bc\u03ad\u03bd\u03b5\u03c2 \u03c0\u03c1\u03bf\u03c3\u03b5\u03b3\u03b3\u03af\u03c3\u03b5\u03b9\u03c2 \u03ae \u03bd\u03b1 \u03c0\u03b1\u03c1\u03b1\u03b4\u03ce\u03c3\u03b5\u03b9 \u03c0\u03bf\u03bb\u03bb\u03ac \u03b1\u03c1\u03c7\u03b5\u03af\u03b1 \u03c7\u03c9\u03c1\u03af\u03c2 \u03bd\u03b1 \u03bb\u03cd\u03c3\u03b5\u03b9 \u03c3\u03c9\u03c3\u03c4\u03ac \u03c4\u03bf \u03c0\u03c1\u03cc\u03b2\u03bb\u03b7\u03bc\u03b1.<\/p>\n<p>\u0391\u03bd\u03c4\u03af\u03b8\u03b5\u03c4\u03b1, \u03b7 \u03b1\u03c5\u03c4\u03bf\u03b5\u03c0\u03b1\u03bb\u03ae\u03b8\u03b5\u03c5\u03c3\u03b7 \u03ad\u03b4\u03c9\u03c3\u03b5 \u03ad\u03bd\u03b1 \u03c0\u03b9\u03bf \u03ac\u03bc\u03b5\u03c3\u03bf, deterministic \u03c3\u03ae\u03bc\u03b1. \u03a4\u03bf claude-opus-5 \u03c0\u03c1\u03b1\u03b3\u03bc\u03b1\u03c4\u03bf\u03c0\u03bf\u03af\u03b7\u03c3\u03b5 \u03ba\u03b1\u03c4\u03ac \u03bc\u03ad\u03c3\u03bf \u03cc\u03c1\u03bf 23,9 verification actions \u03b1\u03bd\u03ac run, \u03ad\u03bd\u03b1\u03bd\u03c4\u03b9 0,58 \u03b3\u03b9\u03b1 \u03c4\u03bf gemma-4-31b-it. \u0397 \u03c3\u03b5\u03b9\u03c1\u03ac \u03c4\u03c9\u03bd \u03bc\u03bf\u03bd\u03c4\u03ad\u03bb\u03c9\u03bd \u03c9\u03c2 \u03c0\u03c1\u03bf\u03c2 verification frequency \u03b1\u03ba\u03bf\u03bb\u03bf\u03cd\u03b8\u03b7\u03c3\u03b5 \u03c3\u03c4\u03b5\u03bd\u03ac \u03c4\u03b7 \u03c3\u03b5\u03b9\u03c1\u03ac \u03c4\u03bf\u03c5 score, \u03c7\u03c9\u03c1\u03af\u03c2 \u03bd\u03b1 \u03c7\u03c1\u03b5\u03b9\u03ac\u03b6\u03b5\u03c4\u03b1\u03b9 LLM judge \u03b3\u03b9\u03b1 \u03c4\u03b7\u03bd \u03ba\u03b1\u03c4\u03b1\u03bc\u03ad\u03c4\u03c1\u03b7\u03c3\u03b7.<\/p>\n<p>\u0393\u03b9\u03b1 procurement \u03ba\u03b1\u03b9 operations, \u03c4\u03b1 KPIs \u03b4\u03b5\u03bd \u03c0\u03c1\u03ad\u03c0\u03b5\u03b9 \u03bd\u03b1 \u03b1\u03bd\u03c4\u03b1\u03bc\u03b5\u03af\u03b2\u03bf\u03c5\u03bd \u03b1\u03c0\u03bb\u03ce\u03c2 token spend, \u03b4\u03b9\u03ac\u03c1\u03ba\u03b5\u03b9\u03b1 \u03ae \u03b1\u03c1\u03b9\u03b8\u03bc\u03cc \u03b5\u03bd\u03b5\u03c1\u03b3\u03b5\u03b9\u03ce\u03bd. \u03a7\u03c1\u03b5\u03b9\u03ac\u03b6\u03bf\u03bd\u03c4\u03b1\u03b9 outcome checks: \u03b1\u03bd \u03b5\u03ba\u03c0\u03bb\u03b7\u03c1\u03ce\u03b8\u03b7\u03ba\u03b5 \u03c4\u03bf \u03b1\u03af\u03c4\u03b7\u03bc\u03b1, \u03b1\u03bd \u03b7 \u03bc\u03ad\u03b8\u03bf\u03b4\u03bf\u03c2 \u03b5\u03af\u03bd\u03b1\u03b9 \u03c3\u03c9\u03c3\u03c4\u03ae, \u03b1\u03bd \u03c4\u03b1 \u03b4\u03b5\u03b4\u03bf\u03bc\u03ad\u03bd\u03b1 \u03c7\u03c1\u03b7\u03c3\u03b9\u03bc\u03bf\u03c0\u03bf\u03b9\u03ae\u03b8\u03b7\u03ba\u03b1\u03bd \u03c0\u03b9\u03c3\u03c4\u03ac, \u03b1\u03bd \u03c4\u03bf \u03c0\u03b1\u03c1\u03b1\u03b4\u03bf\u03c4\u03ad\u03bf \u03b1\u03bd\u03bf\u03af\u03b3\u03b5\u03b9 \u03ba\u03b1\u03b9 \u03b1\u03bd \u03bf agent \u03be\u03b1\u03bd\u03b1\u03ad\u03bb\u03b5\u03b3\u03be\u03b5 \u03ba\u03c1\u03af\u03c3\u03b9\u03bc\u03b1 outputs.<\/p>\n<h2 id=\"krities-organo\">\u039f\u03b9 \u03ba\u03c1\u03b9\u03c4\u03ad\u03c2 \u03b5\u03af\u03bd\u03b1\u03b9 \u03ba\u03b1\u03b9 \u03b1\u03c5\u03c4\u03bf\u03af \u03bc\u03ad\u03c1\u03bf\u03c2 \u03c4\u03bf\u03c5 \u03c3\u03c5\u03c3\u03c4\u03ae\u03bc\u03b1\u03c4\u03bf\u03c2<\/h2>\n<p>\u03a4\u03bf K-Bench \u03b4\u03b5\u03bd \u03b1\u03bd\u03c4\u03b9\u03bc\u03b5\u03c4\u03c9\u03c0\u03af\u03b6\u03b5\u03b9 \u03c4\u03bf\u03bd LLM judge \u03c9\u03c2 \u03b1\u03cc\u03c1\u03b1\u03c4\u03b7 \u03b1\u03bb\u03ae\u03b8\u03b5\u03b9\u03b1. \u039c\u03b5\u03bb\u03b5\u03c4\u03ac \u03c4\u03b7 \u03c3\u03c5\u03bc\u03c6\u03c9\u03bd\u03af\u03b1, \u03c4\u03b7 \u03bc\u03b5\u03c4\u03b1\u03c4\u03cc\u03c0\u03b9\u03c3\u03b7 \u03ba\u03bb\u03af\u03bc\u03b1\u03ba\u03b1\u03c2, \u03c4\u03b9\u03c2 split decisions \u03ba\u03b1\u03b9 \u03c4\u03b7\u03bd \u03c0\u03b5\u03c1\u03af\u03c0\u03c4\u03c9\u03c3\u03b7 \u03cc\u03c0\u03bf\u03c5 \u03ba\u03c1\u03b9\u03c4\u03ad\u03c2 \u03b2\u03b1\u03b8\u03bc\u03bf\u03bb\u03bf\u03b3\u03bf\u03cd\u03bd runs \u03c4\u03b7\u03c2 \u03b4\u03b9\u03ba\u03ae\u03c2 \u03c4\u03bf\u03c5\u03c2 model family. \u0397 \u03bf\u03c5\u03c3\u03b9\u03b1\u03c3\u03c4\u03b9\u03ba\u03ae \u03b4\u03b9\u03b1\u03c0\u03af\u03c3\u03c4\u03c9\u03c3\u03b7 \u03b5\u03af\u03bd\u03b1\u03b9 \u03cc\u03c4\u03b9 \u03b7 \u03c3\u03b5\u03b9\u03c1\u03ac \u03b1\u03bd\u03b1\u03c0\u03b1\u03c1\u03ac\u03b3\u03b5\u03c4\u03b1\u03b9 \u03ba\u03b1\u03bb\u03cd\u03c4\u03b5\u03c1\u03b1 \u03b1\u03c0\u03cc \u03c4\u03bf \u03b1\u03c0\u03cc\u03bb\u03c5\u03c4\u03bf score.<\/p>\n<p>\u0394\u03cd\u03bf \u03b1\u03c0\u03cc \u03c4\u03bf\u03c5\u03c2 \u03c4\u03c1\u03b5\u03b9\u03c2 \u03ba\u03c1\u03b9\u03c4\u03ad\u03c2 \u03ae\u03c4\u03b1\u03bd \u03ba\u03b1\u03b9 \u03b4\u03b9\u03b1\u03b3\u03c9\u03bd\u03b9\u03b6\u03cc\u03bc\u03b5\u03bd\u03b1 \u03bc\u03bf\u03bd\u03c4\u03ad\u03bb\u03b1. \u039c\u03b5\u03c4\u03ac \u03b1\u03c0\u03cc calibration adjustment, \u03c4\u03bf gpt-5.6-sol \u03ad\u03b4\u03b9\u03bd\u03b5 \u03c3\u03c4\u03b1 \u03b4\u03b9\u03ba\u03ac \u03c4\u03bf\u03c5 runs \u03c0\u03bb\u03b5\u03bf\u03bd\u03ad\u03ba\u03c4\u03b7\u03bc\u03b1 +0,83 \u03bc\u03bf\u03bd\u03ac\u03b4\u03c9\u03bd \u03c3\u03b5 \u03c3\u03c7\u03ad\u03c3\u03b7 \u03bc\u03b5 \u03c4\u03bf \u03c0\u03ce\u03c2 \u03c4\u03bf\u03c0\u03bf\u03b8\u03b5\u03c4\u03bf\u03cd\u03c3\u03b1\u03bd \u03c4\u03b7\u03bd \u03af\u03b4\u03b9\u03b1 \u03b4\u03b9\u03b1\u03c6\u03bf\u03c1\u03ac \u03bf\u03b9 \u03ac\u03bb\u03bb\u03bf\u03b9 judges. \u03a4\u03bf qwen3.8-max, \u03bf \u03bc\u03cc\u03bd\u03bf\u03c2 judge \u03c0\u03bf\u03c5 \u03b4\u03b5\u03bd \u03ae\u03c4\u03b1\u03bd contestant, \u03ad\u03b2\u03b1\u03bb\u03b5 \u03c0\u03c1\u03ce\u03c4\u03bf \u03c4\u03bf claude-opus-5.<\/p>\n<p>\u0391\u03c5\u03c4\u03ae \u03b7 \u03b1\u03c1\u03c7\u03ae \u03b1\u03c6\u03bf\u03c1\u03ac \u03ba\u03ac\u03b8\u03b5 \u03b5\u03c4\u03b1\u03b9\u03c1\u03b9\u03ba\u03cc evaluation. \u0391\u03bd \u03b1\u03bb\u03bb\u03ac\u03be\u03b5\u03b9 \u03bf judge, \u03c4\u03bf prompt \u03ae \u03c4\u03bf rubric, \u03bc\u03c0\u03bf\u03c1\u03b5\u03af \u03bd\u03b1 \u03b1\u03bb\u03bb\u03ac\u03be\u03b5\u03b9 \u03c4\u03bf \u03b5\u03c0\u03af\u03c0\u03b5\u03b4\u03bf. \u0397 \u03b5\u03c0\u03b9\u03c7\u03b5\u03af\u03c1\u03b7\u03c3\u03b7 \u03c7\u03c1\u03b5\u03b9\u03ac\u03b6\u03b5\u03c4\u03b1\u03b9 \u03c3\u03c4\u03b1\u03b8\u03b5\u03c1\u03ad\u03c2 \u03ac\u03b3\u03ba\u03c5\u03c1\u03b5\u03c2, human review \u03c3\u03b5 \u03b4\u03b5\u03af\u03b3\u03bc\u03b1, versioning, inter-rater checks \u03ba\u03b1\u03b9 \u03b1\u03bd\u03b1\u03c6\u03bf\u03c1\u03ad\u03c2 \u03b1\u03b2\u03b5\u03b2\u03b1\u03b9\u03cc\u03c4\u03b7\u03c4\u03b1\u03c2. \u0394\u03b9\u03b1\u03c6\u03bf\u03c1\u03b5\u03c4\u03b9\u03ba\u03ac, \u03bc\u03b9\u03b1 \u03bc\u03b5\u03c4\u03b1\u03b2\u03bf\u03bb\u03ae score \u03bc\u03c0\u03bf\u03c1\u03b5\u03af \u03bd\u03b1 \u03b5\u03af\u03bd\u03b1\u03b9 \u03b1\u03bb\u03bb\u03b1\u03b3\u03ae \u03c4\u03bf\u03c5 \u03bf\u03c1\u03b3\u03ac\u03bd\u03bf\u03c5 \u03ba\u03b1\u03b9 \u03cc\u03c7\u03b9 \u03c4\u03bf\u03c5 \u03bc\u03bf\u03bd\u03c4\u03ad\u03bb\u03bf\u03c5.<\/p>\n<h2 id=\"oria-privacy\">\u038c\u03c1\u03b9\u03b1, privacy \u03ba\u03b1\u03b9 \u03b4\u03c5\u03bd\u03b1\u03c4\u03cc\u03c4\u03b7\u03c4\u03b1 \u03b1\u03bd\u03b1\u03c0\u03b1\u03c1\u03b1\u03b3\u03c9\u03b3\u03ae\u03c2<\/h2>\n<p>\u0394\u03b5\u03bd \u03c5\u03c0\u03ac\u03c1\u03c7\u03b5\u03b9 calibration \u03bc\u03b5 \u03b1\u03bd\u03b8\u03c1\u03ce\u03c0\u03bf\u03c5\u03c2 \u03bf\u03cd\u03c4\u03b5 expert baseline, \u03ac\u03c1\u03b1 \u03c4\u03bf 8 \u03b4\u03b5\u03bd \u03b1\u03c0\u03bf\u03b4\u03b5\u03b9\u03ba\u03bd\u03cd\u03b5\u03b9 \u03cc\u03c4\u03b9 \u03ad\u03bd\u03b1\u03c2 domain scientist \u03b8\u03b1 \u03ad\u03b4\u03b9\u03bd\u03b5 \u03c4\u03b7\u03bd \u03af\u03b4\u03b9\u03b1 \u03b2\u03b1\u03b8\u03bc\u03bf\u03bb\u03bf\u03b3\u03af\u03b1. \u039a\u03ac\u03b8\u03b5 run \u03b5\u03ba\u03c4\u03b5\u03bb\u03ad\u03c3\u03c4\u03b7\u03ba\u03b5 \u03bc\u03af\u03b1 \u03c6\u03bf\u03c1\u03ac \u03ba\u03b1\u03b9 \u03c4\u03b1 bootstrap intervals \u03ba\u03b1\u03bb\u03cd\u03c0\u03c4\u03bf\u03c5\u03bd \u03b1\u03b2\u03b5\u03b2\u03b1\u03b9\u03cc\u03c4\u03b7\u03c4\u03b1 \u03c4\u03bf\u03c5 \u03b4\u03b5\u03af\u03b3\u03bc\u03b1\u03c4\u03bf\u03c2 \u03c4\u03c9\u03bd 178 sessions, \u03cc\u03c7\u03b9 run-to-run variation, \u03b4\u03b9\u03b1\u03c6\u03bf\u03c1\u03ac \u03ba\u03c1\u03b9\u03c4\u03ce\u03bd \u03ae \u03ac\u03bb\u03bb\u03b7 \u03c3\u03cd\u03bd\u03b8\u03b5\u03c3\u03b7 panel. \u03a4\u03b1 \u03b1\u03c0\u03bf\u03c4\u03b5\u03bb\u03ad\u03c3\u03bc\u03b1\u03c4\u03b1 \u03c0\u03b5\u03c1\u03b9\u03b3\u03c1\u03ac\u03c6\u03bf\u03c5\u03bd \u03b1\u03c5\u03c4\u03ac \u03c4\u03b1 models, \u03c3\u03b5 \u03b1\u03c5\u03c4\u03ac \u03c4\u03b1 tasks, \u03c3\u03c4\u03bf stock pi 0.84.0 harness \u03ba\u03b1\u03b9 \u03c3\u03c4\u03b9\u03c2 \u03b7\u03bc\u03b5\u03c1\u03bf\u03bc\u03b7\u03bd\u03af\u03b5\u03c2 \u03c4\u03b7\u03c2 \u03ba\u03b1\u03bc\u03c0\u03ac\u03bd\u03b9\u03b1\u03c2.<\/p>\n<p>\u03a4\u03bf \u03b4\u03b5\u03af\u03b3\u03bc\u03b1 \u03b4\u03b5\u03bd \u03b5\u03af\u03bd\u03b1\u03b9 \u03c4\u03c5\u03c7\u03b1\u03af\u03bf \u03b4\u03b5\u03af\u03b3\u03bc\u03b1 \u03cc\u03bb\u03b7\u03c2 \u03c4\u03b7\u03c2 \u03b5\u03c0\u03b9\u03c3\u03c4\u03ae\u03bc\u03b7\u03c2. \u03a0\u03c1\u03bf\u03ad\u03c1\u03c7\u03b5\u03c4\u03b1\u03b9 \u03b1\u03c0\u03cc traffic \u03b5\u03bd\u03cc\u03c2 \u03c0\u03c1\u03bf\u03ca\u03cc\u03bd\u03c4\u03bf\u03c2, \u03bc\u03b5 \u03ac\u03bd\u03b9\u03c3\u03b7 \u03b5\u03ba\u03c0\u03c1\u03bf\u03c3\u03ce\u03c0\u03b7\u03c3\u03b7 \u03c4\u03b5\u03c3\u03c3\u03ac\u03c1\u03c9\u03bd broad domains. \u0395\u03af\u03ba\u03bf\u03c3\u03b9 \u03b4\u03cd\u03bf sessions \u03c3\u03c4\u03b1 \u03bf\u03c0\u03bf\u03af\u03b1 \u03c4\u03bf\u03c5\u03bb\u03ac\u03c7\u03b9\u03c3\u03c4\u03bf\u03bd \u03ad\u03bd\u03b1 \u03bc\u03bf\u03bd\u03c4\u03ad\u03bb\u03bf \u03b1\u03c1\u03bd\u03ae\u03b8\u03b7\u03ba\u03b5 \u03b1\u03c0\u03bf\u03ba\u03bb\u03b5\u03af\u03c3\u03c4\u03b7\u03ba\u03b1\u03bd \u03b3\u03b9\u03b1 \u03bd\u03b1 \u03b4\u03b9\u03b1\u03c4\u03b7\u03c1\u03b7\u03b8\u03b5\u03af complete-case \u03c3\u03cd\u03b3\u03ba\u03c1\u03b9\u03c3\u03b7. \u0388\u03c4\u03c3\u03b9 \u03c4\u03bf benchmark \u03b4\u03b5\u03bd \u03ba\u03b1\u03bb\u03cd\u03c0\u03c4\u03b5\u03b9 \u03ba\u03b1\u03bb\u03ac \u03b1\u03b9\u03c4\u03ae\u03bc\u03b1\u03c4\u03b1 \u03ba\u03bf\u03bd\u03c4\u03ac \u03c3\u03c4\u03b1 \u03cc\u03c1\u03b9\u03b1 \u03b1\u03c3\u03c6\u03ac\u03bb\u03b5\u03b9\u03b1\u03c2 \u03c4\u03c9\u03bd \u03c0\u03b1\u03c1\u03cc\u03c7\u03c9\u03bd.<\/p>\n<p>\u039f\u03b9 \u03c3\u03c5\u03bd\u03b5\u03b4\u03c1\u03af\u03b5\u03c2 \u03c7\u03c1\u03b7\u03c3\u03b9\u03bc\u03bf\u03c0\u03bf\u03b9\u03ae\u03b8\u03b7\u03ba\u03b1\u03bd \u03b2\u03ac\u03c3\u03b5\u03b9 \u03c4\u03c9\u03bd \u03cc\u03c1\u03c9\u03bd \u03c4\u03b7\u03c2 K-Dense, \u03c7\u03c9\u03c1\u03af\u03c2 \u03be\u03b5\u03c7\u03c9\u03c1\u03b9\u03c3\u03c4\u03ae \u03b5\u03bd\u03b7\u03bc\u03ad\u03c1\u03c9\u03c3\u03b7 \u03ae \u03b1\u03c4\u03bf\u03bc\u03b9\u03ba\u03cc opt-in \u03b3\u03b9\u03b1 \u03c4\u03b7\u03bd \u03ba\u03b1\u03bc\u03c0\u03ac\u03bd\u03b9\u03b1. \u03a4\u03b1 \u03b1\u03b9\u03c4\u03ae\u03bc\u03b1\u03c4\u03b1 \u03ba\u03b1\u03b9 \u03c4\u03b1 attachments \u03c3\u03c4\u03ac\u03bb\u03b8\u03b7\u03ba\u03b1\u03bd \u03cc\u03c0\u03c9\u03c2 \u03c5\u03c0\u03bf\u03b2\u03bb\u03ae\u03b8\u03b7\u03ba\u03b1\u03bd \u03c3\u03b5 \u03b5\u03bd\u03bd\u03ad\u03b1 \u03c4\u03c1\u03af\u03c4\u03bf\u03c5\u03c2 \u03c0\u03b1\u03c1\u03cc\u03c7\u03bf\u03c5\u03c2, \u03c7\u03c9\u03c1\u03af\u03c2 content-level de-identification \u03ae audit identifiers. \u039f\u03b9 endpoints \u03ae\u03c4\u03b1\u03bd \u03c1\u03c5\u03b8\u03bc\u03b9\u03c3\u03bc\u03ad\u03bd\u03bf\u03b9 \u03b3\u03b9\u03b1 zero retention, \u03b1\u03bb\u03bb\u03ac \u03b1\u03c5\u03c4\u03cc \u03b4\u03b5\u03bd \u03b1\u03bd\u03c4\u03b9\u03ba\u03b1\u03b8\u03b9\u03c3\u03c4\u03ac data minimization, redaction \u03ba\u03b1\u03b9 \u03c1\u03b7\u03c4\u03cc governance \u03c3\u03b5 \u03bc\u03b9\u03b1 \u03b5\u03c4\u03b1\u03b9\u03c1\u03b9\u03ba\u03ae \u03c5\u03bb\u03bf\u03c0\u03bf\u03af\u03b7\u03c3\u03b7.<\/p>\n<aside class=\"td-article-note\"><strong>\u039c\u03b5\u03c1\u03b9\u03ba\u03ae \u03b1\u03bd\u03b1\u03c0\u03b1\u03c1\u03b1\u03b3\u03c9\u03b3\u03b9\u03bc\u03cc\u03c4\u03b7\u03c4\u03b1, \u03cc\u03c7\u03b9 \u03c0\u03bb\u03ae\u03c1\u03b7\u03c2 \u03b5\u03c0\u03b1\u03bd\u03ac\u03bb\u03b7\u03c8\u03b7 \u03c4\u03bf\u03c5 benchmark:<\/strong> \u03b8\u03b1 \u03b4\u03b7\u03bc\u03bf\u03c3\u03b9\u03b5\u03c5\u03b8\u03bf\u03cd\u03bd de-identified scores, analysis code \u03ba\u03b1\u03b9 \u03c4\u03bf \u03c0\u03bb\u03ae\u03c1\u03b5\u03c2 rubric, \u03cc\u03bc\u03c9\u03c2 prompts, attachments, transcripts \u03ba\u03b1\u03b9 artifacts \u03c0\u03b1\u03c1\u03b1\u03bc\u03ad\u03bd\u03bf\u03c5\u03bd \u03b9\u03b4\u03b9\u03c9\u03c4\u03b9\u03ba\u03ac. \u0395\u03be\u03c9\u03c4\u03b5\u03c1\u03b9\u03ba\u03ae \u03bf\u03bc\u03ac\u03b4\u03b1 \u03bc\u03c0\u03bf\u03c1\u03b5\u03af \u03bd\u03b1 \u03b5\u03bb\u03ad\u03b3\u03be\u03b5\u03b9 \u03c4\u03bf\u03c5\u03c2 \u03c5\u03c0\u03bf\u03bb\u03bf\u03b3\u03b9\u03c3\u03bc\u03bf\u03cd\u03c2, \u03cc\u03c7\u03b9 \u03bd\u03b1 \u03c4\u03c1\u03ad\u03be\u03b5\u03b9 \u03bd\u03ad\u03bf \u03bc\u03bf\u03bd\u03c4\u03ad\u03bb\u03bf \u03c3\u03c4\u03b1 \u03af\u03b4\u03b9\u03b1 tasks.<\/aside>\n<p>\u038c\u03bb\u03bf\u03b9 \u03bf\u03b9 \u03c3\u03c5\u03b3\u03b3\u03c1\u03b1\u03c6\u03b5\u03af\u03c2 \u03b5\u03c1\u03b3\u03ac\u03b6\u03bf\u03bd\u03c4\u03b1\u03b9 \u03c3\u03c4\u03b7\u03bd K-Dense, \u03c4\u03b7\u03c2 \u03bf\u03c0\u03bf\u03af\u03b1\u03c2 \u03c4\u03bf \u03c0\u03c1\u03bf\u03ca\u03cc\u03bd \u03c0\u03b1\u03c1\u03b5\u03af\u03c7\u03b5 \u03c4\u03bf corpus \u03ba\u03b1\u03b9 \u03c4\u03b7\u03c2 \u03bf\u03c0\u03bf\u03af\u03b1\u03c2 \u03b5\u03af\u03bd\u03b1\u03b9 \u03c4\u03bf harness \u03ba\u03b1\u03b9 \u03c4\u03bf rubric. \u0391\u03c5\u03c4\u03cc \u03b4\u03b5\u03bd \u03b1\u03ba\u03c5\u03c1\u03ce\u03bd\u03b5\u03b9 \u03c4\u03b1 \u03b5\u03c5\u03c1\u03ae\u03bc\u03b1\u03c4\u03b1, \u03b1\u03bb\u03bb\u03ac \u03b1\u03c0\u03bf\u03c4\u03b5\u03bb\u03b5\u03af \u03c3\u03c7\u03b5\u03c4\u03b9\u03ba\u03cc competing interest. \u03a3\u03b5 \u03c0\u03b1\u03c1\u03b1\u03b3\u03c9\u03b3\u03b9\u03ba\u03ae \u03b5\u03c0\u03b9\u03bb\u03bf\u03b3\u03ae \u03bc\u03bf\u03bd\u03c4\u03ad\u03bb\u03bf\u03c5, \u03ad\u03bd\u03b1 \u03b5\u03c3\u03c9\u03c4\u03b5\u03c1\u03b9\u03ba\u03cc benchmark \u03c7\u03c1\u03b5\u03b9\u03ac\u03b6\u03b5\u03c4\u03b1\u03b9 \u03b1\u03bd\u03b5\u03be\u03ac\u03c1\u03c4\u03b7\u03c4\u03b1 acceptance tests \u03c0\u03ac\u03bd\u03c9 \u03c3\u03c4\u03b1 \u03b4\u03b9\u03ba\u03ac \u03c3\u03b1\u03c2 \u03b4\u03b5\u03b4\u03bf\u03bc\u03ad\u03bd\u03b1 \u03ba\u03b1\u03b9 failure costs.<\/p>\n<div class=\"td-decision-band\">\n<p class=\"td-decision-kicker\">Go\/no-go \u03b3\u03b9\u03b1 \u03b5\u03c0\u03b9\u03c7\u03b5\u03b9\u03c1\u03b7\u03c3\u03b9\u03b1\u03ba\u03cc AI agent<\/p>\n<p class=\"td-decision-title\">\u0391\u03c0\u03bf\u03b4\u03b5\u03c7\u03c4\u03b5\u03af\u03c4\u03b5 \u03b5\u03c1\u03b3\u03b1\u03c3\u03af\u03b1 \u03bc\u03cc\u03bd\u03bf \u03cc\u03c4\u03b1\u03bd \u03b7 \u03b1\u03c0\u03ac\u03bd\u03c4\u03b7\u03c3\u03b7 \u03ba\u03b1\u03b9 \u03c4\u03b1 artifacts \u03c3\u03c5\u03bc\u03c6\u03c9\u03bd\u03bf\u03cd\u03bd<\/p>\n<p>Go \u03c3\u03b7\u03bc\u03b1\u03af\u03bd\u03b5\u03b9 \u03c0\u03bb\u03ae\u03c1\u03b5\u03c2 \u03c0\u03b1\u03c1\u03b1\u03b4\u03bf\u03c4\u03ad\u03bf, traceable claims, \u03c0\u03b5\u03c1\u03b1\u03c3\u03bc\u03ad\u03bd\u03b1 deterministic checks, \u03c3\u03c9\u03c3\u03c4\u03cc \u03c7\u03b5\u03b9\u03c1\u03b9\u03c3\u03bc\u03cc \u03b4\u03b5\u03b4\u03bf\u03bc\u03ad\u03bd\u03c9\u03bd, \u03b5\u03c0\u03b1\u03c1\u03ba\u03ae uncertainty \u03ba\u03b1\u03b9 \u03b1\u03bd\u03b8\u03c1\u03ce\u03c0\u03b9\u03bd\u03b7 \u03ad\u03b3\u03ba\u03c1\u03b9\u03c3\u03b7 \u03cc\u03c0\u03bf\u03c5 \u03c4\u03bf \u03ba\u03cc\u03c3\u03c4\u03bf\u03c2 \u03bb\u03ac\u03b8\u03bf\u03c5\u03c2 \u03b5\u03af\u03bd\u03b1\u03b9 \u03c5\u03c8\u03b7\u03bb\u03cc. No-go \u03c3\u03b7\u03bc\u03b1\u03af\u03bd\u03b5\u03b9 \u03ac\u03b4\u03b5\u03b9\u03bf \u03ae \u03c7\u03b1\u03bb\u03b1\u03c3\u03bc\u03ad\u03bd\u03bf artifact, unverifiable claim, missing range, \u03b1\u03c0\u03bf\u03c4\u03c5\u03c7\u03b7\u03bc\u03ad\u03bd\u03bf formula, \u03bc\u03b7 \u03b5\u03bb\u03b5\u03b3\u03bc\u03ad\u03bd\u03b7 \u03c0\u03b7\u03b3\u03ae \u03ae \u03c4\u03b5\u03bb\u03b9\u03ba\u03ae \u03b1\u03c0\u03ac\u03bd\u03c4\u03b7\u03c3\u03b7 \u03c0\u03bf\u03c5 \u03b4\u03b7\u03bb\u03ce\u03bd\u03b5\u03b9 \u03c0\u03b5\u03c1\u03b9\u03c3\u03c3\u03cc\u03c4\u03b5\u03c1\u03b1 \u03b1\u03c0\u03cc \u03cc\u03c3\u03b1 \u03ad\u03b3\u03b9\u03bd\u03b1\u03bd.<\/p>\n<\/div>\n<h2 id=\"evaluation-stack\">\u0388\u03bd\u03b1 \u03c0\u03c1\u03b1\u03ba\u03c4\u03b9\u03ba\u03cc evaluation stack \u03b3\u03b9\u03b1 \u03b5\u03c0\u03b9\u03c7\u03b5\u03b9\u03c1\u03ae\u03c3\u03b5\u03b9\u03c2<\/h2>\n<p>\u0397 \u03b5\u03c0\u03b9\u03c7\u03b5\u03b9\u03c1\u03b7\u03c3\u03b9\u03b1\u03ba\u03ae \u03b5\u03c6\u03b1\u03c1\u03bc\u03bf\u03b3\u03ae \u03c4\u03c9\u03bd \u03b5\u03c5\u03c1\u03b7\u03bc\u03ac\u03c4\u03c9\u03bd \u03b5\u03af\u03bd\u03b1\u03b9 \u03b1\u03bd\u03b1\u03bb\u03c5\u03c4\u03b9\u03ba\u03ae \u03c0\u03c1\u03bf\u03ad\u03ba\u03c4\u03b1\u03c3\u03b7 \u03ba\u03b1\u03b9 \u03cc\u03c7\u03b9 \u03b4\u03bf\u03ba\u03b9\u03bc\u03ae \u03c4\u03bf\u03c5 paper \u03c3\u03b5 e-commerce. \u0388\u03bd\u03b1 \u03c3\u03bf\u03b2\u03b1\u03c1\u03cc stack \u03be\u03b5\u03ba\u03b9\u03bd\u03ac \u03b1\u03c0\u03cc \u03c0\u03c1\u03b1\u03b3\u03bc\u03b1\u03c4\u03b9\u03ba\u03ac tasks \u03c4\u03b7\u03c2 \u03b5\u03c4\u03b1\u03b9\u03c1\u03b5\u03af\u03b1\u03c2, frozen test set, \u03c3\u03b1\u03c6\u03ad\u03c2 rubric \u03ba\u03b1\u03b9 acceptance criteria \u03b3\u03b9\u03b1 \u03c4\u03bf \u03c4\u03b5\u03bb\u03b9\u03ba\u03cc state. \u039a\u03ac\u03b8\u03b5 \u03b1\u03c0\u03bf\u03c4\u03ad\u03bb\u03b5\u03c3\u03bc\u03b1 \u03c0\u03c1\u03ad\u03c0\u03b5\u03b9 \u03bd\u03b1 \u03c3\u03c5\u03bd\u03bf\u03b4\u03b5\u03cd\u03b5\u03c4\u03b1\u03b9 \u03b1\u03c0\u03cc model version, harness version, permissions, judge setup \u03ba\u03b1\u03b9 \u03c0\u03b1\u03c1\u03b1\u03b4\u03b5\u03af\u03b3\u03bc\u03b1\u03c4\u03b1 failures.<\/p>\n<div class=\"td-step-list\">\n<p class=\"td-step-list-title\">\u0395\u03c0\u03c4\u03ac \u03b2\u03ae\u03bc\u03b1\u03c4\u03b1 \u03b3\u03b9\u03b1 evaluation \u03c0\u03bf\u03c5 \u03bc\u03b5\u03c4\u03c1\u03ac \u03c4\u03b7 \u03b4\u03bf\u03c5\u03bb\u03b5\u03b9\u03ac<\/p>\n<ol>\n<li><span class=\"td-step-kicker\">Step 1<\/span><strong>\u03a3\u03c5\u03bb\u03bb\u03ad\u03be\u03c4\u03b5 \u03b1\u03bd\u03c4\u03b9\u03c0\u03c1\u03bf\u03c3\u03c9\u03c0\u03b5\u03c5\u03c4\u03b9\u03ba\u03ac \u03c0\u03c1\u03b1\u03b3\u03bc\u03b1\u03c4\u03b9\u03ba\u03ac tasks<\/strong>\n<p>\u0394\u03b9\u03b1\u03bb\u03ad\u03be\u03c4\u03b5 \u03b5\u03c1\u03b3\u03b1\u03c3\u03af\u03b5\u03c2 \u03bc\u03b5 \u03c4\u03bf \u03c0\u03c1\u03b1\u03b3\u03bc\u03b1\u03c4\u03b9\u03ba\u03cc \u03bc\u03b5\u03af\u03b3\u03bc\u03b1 \u03b1\u03c3\u03ac\u03c6\u03b5\u03b9\u03b1\u03c2, attachments, \u03b5\u03be\u03b1\u03b9\u03c1\u03ad\u03c3\u03b5\u03c9\u03bd \u03ba\u03b1\u03b9 \u03b5\u03c0\u03b9\u03c7\u03b5\u03b9\u03c1\u03b7\u03c3\u03b9\u03b1\u03ba\u03bf\u03cd \u03ba\u03cc\u03c3\u03c4\u03bf\u03c5\u03c2, \u03b1\u03c6\u03bf\u03cd \u03c0\u03c1\u03ce\u03c4\u03b1 \u03b1\u03c6\u03b1\u03b9\u03c1\u03ad\u03c3\u03b5\u03c4\u03b5 \u03c0\u03b5\u03c1\u03b9\u03c4\u03c4\u03ac \u03c0\u03c1\u03bf\u03c3\u03c9\u03c0\u03b9\u03ba\u03ac \u03ae \u03b5\u03bc\u03c0\u03b9\u03c3\u03c4\u03b5\u03c5\u03c4\u03b9\u03ba\u03ac \u03b4\u03b5\u03b4\u03bf\u03bc\u03ad\u03bd\u03b1.<\/p>\n<\/li>\n<li><span class=\"td-step-kicker\">Step 2<\/span><strong>\u03a0\u03b1\u03b3\u03ce\u03c3\u03c4\u03b5 inputs, environment \u03ba\u03b1\u03b9 version<\/strong>\n<p>\u039a\u03c1\u03b1\u03c4\u03ae\u03c3\u03c4\u03b5 \u03c3\u03c4\u03b1\u03b8\u03b5\u03c1\u03ac model identifier, system prompt, tools, permissions, libraries, network access \u03ba\u03b1\u03b9 test data \u03ce\u03c3\u03c4\u03b5 \u03bc\u03b9\u03b1 \u03b1\u03bb\u03bb\u03b1\u03b3\u03ae score \u03bd\u03b1 \u03b5\u03af\u03bd\u03b1\u03b9 \u03b5\u03c1\u03bc\u03b7\u03bd\u03b5\u03cd\u03c3\u03b9\u03bc\u03b7.<\/p>\n<\/li>\n<li><span class=\"td-step-kicker\">Step 3<\/span><strong>\u039f\u03c1\u03af\u03c3\u03c4\u03b5 \u03c7\u03c9\u03c1\u03b9\u03c3\u03c4\u03ac communication \u03ba\u03b1\u03b9 substance<\/strong>\n<p>\u0392\u03b1\u03b8\u03bc\u03bf\u03bb\u03bf\u03b3\u03ae\u03c3\u03c4\u03b5 \u03c3\u03b1\u03c6\u03ae\u03bd\u03b5\u03b9\u03b1, \u03b1\u03ba\u03c1\u03af\u03b2\u03b5\u03b9\u03b1, \u03ba\u03ac\u03bb\u03c5\u03c8\u03b7, data handling, \u03b5\u03b9\u03bb\u03b9\u03ba\u03c1\u03af\u03bd\u03b5\u03b9\u03b1 \u03ba\u03b1\u03b9 artifact quality \u03c9\u03c2 \u03b4\u03b9\u03b1\u03c6\u03bf\u03c1\u03b5\u03c4\u03b9\u03ba\u03bf\u03cd\u03c2 \u03ac\u03be\u03bf\u03bd\u03b5\u03c2, \u03cc\u03c7\u03b9 \u03c9\u03c2 \u03ad\u03bd\u03b1 \u03b1\u03b4\u03b9\u03b1\u03c6\u03b1\u03bd\u03ad\u03c2 \u03c3\u03c5\u03bd\u03bf\u03bb\u03b9\u03ba\u03cc score.<\/p>\n<\/li>\n<li><span class=\"td-step-kicker\">Step 4<\/span><strong>\u0391\u03bd\u03bf\u03af\u03be\u03c4\u03b5 \u03ba\u03b1\u03b9 \u03b5\u03bb\u03ad\u03b3\u03be\u03c4\u03b5 \u03ba\u03ac\u03b8\u03b5 artifact<\/strong>\n<p>\u039a\u03ac\u03bd\u03c4\u03b5 parse \u03b1\u03c1\u03c7\u03b5\u03af\u03b1, \u03c4\u03c1\u03ad\u03be\u03c4\u03b5 formulas \u03ba\u03b1\u03b9 code, \u03b5\u03bb\u03ad\u03b3\u03be\u03c4\u03b5 schemas, totals, ranges, links \u03ba\u03b1\u03b9 \u03b1\u03c0\u03b1\u03b9\u03c4\u03bf\u03cd\u03bc\u03b5\u03bd\u03b5\u03c2 \u03bc\u03bf\u03c1\u03c6\u03ad\u03c2. \u0397 \u03cd\u03c0\u03b1\u03c1\u03be\u03b7 \u03b1\u03c1\u03c7\u03b5\u03af\u03bf\u03c5 \u03b4\u03b5\u03bd \u03b1\u03c0\u03bf\u03b4\u03b5\u03b9\u03ba\u03bd\u03cd\u03b5\u03b9 \u03c7\u03c1\u03b7\u03c3\u03b9\u03bc\u03cc\u03c4\u03b7\u03c4\u03b1.<\/p>\n<\/li>\n<li><span class=\"td-step-kicker\">Step 5<\/span><strong>\u03a3\u03c5\u03bd\u03b4\u03ad\u03c3\u03c4\u03b5 claims \u03bc\u03b5 evidence<\/strong>\n<p>\u0396\u03b7\u03c4\u03ae\u03c3\u03c4\u03b5 \u03c0\u03b1\u03c1\u03b1\u03c0\u03bf\u03bc\u03c0\u03ae \u03c3\u03b5 \u03c0\u03b7\u03b3\u03ae, \u03c5\u03c0\u03bf\u03bb\u03bf\u03b3\u03b9\u03c3\u03bc\u03cc, tool output \u03ae \u03b3\u03c1\u03b1\u03bc\u03bc\u03ae \u03b4\u03b5\u03b4\u03bf\u03bc\u03ad\u03bd\u03c9\u03bd \u03ba\u03b1\u03b9 \u03b1\u03bd\u03b9\u03c7\u03bd\u03b5\u03cd\u03c3\u03c4\u03b5 overclaiming, fabricated results \u03ba\u03b1\u03b9 premature completion.<\/p>\n<\/li>\n<li><span class=\"td-step-kicker\">Step 6<\/span><strong>\u0392\u03b1\u03b8\u03bc\u03bf\u03bd\u03bf\u03bc\u03ae\u03c3\u03c4\u03b5 judges \u03bc\u03b5 \u03b1\u03bd\u03b8\u03c1\u03ce\u03c0\u03b9\u03bd\u03bf \u03b4\u03b5\u03af\u03b3\u03bc\u03b1<\/strong>\n<p>\u039c\u03b5\u03c4\u03c1\u03ae\u03c3\u03c4\u03b5 \u03c3\u03c5\u03bc\u03c6\u03c9\u03bd\u03af\u03b1, \u03b1\u03c5\u03c3\u03c4\u03b7\u03c1\u03cc\u03c4\u03b7\u03c4\u03b1 \u03ba\u03b1\u03b9 \u03bf\u03b9\u03ba\u03bf\u03b3\u03b5\u03bd\u03b5\u03b9\u03b1\u03ba\u03ae \u03c0\u03c1\u03bf\u03ba\u03b1\u03c4\u03ac\u03bb\u03b7\u03c8\u03b7 \u03c4\u03c9\u03bd LLM judges \u03b1\u03c0\u03ad\u03bd\u03b1\u03bd\u03c4\u03b9 \u03c3\u03b5 \u03b5\u03b9\u03b4\u03b9\u03ba\u03bf\u03cd\u03c2 \u03c4\u03bf\u03c5 domain \u03ba\u03b1\u03b9 \u03ba\u03c1\u03b1\u03c4\u03ae\u03c3\u03c4\u03b5 split decisions \u03bf\u03c1\u03b1\u03c4\u03ad\u03c2.<\/p>\n<\/li>\n<li><span class=\"td-step-kicker\">Step 7<\/span><strong>\u0392\u03ac\u03bb\u03c4\u03b5 \u03cc\u03c1\u03b9\u03b1, handoff \u03ba\u03b1\u03b9 rollback<\/strong>\n<p>\u039f\u03c1\u03af\u03c3\u03c4\u03b5 \u03c0\u03bf\u03b9\u03b1 failures \u03bc\u03c0\u03bb\u03bf\u03ba\u03ac\u03c1\u03bf\u03c5\u03bd \u03c4\u03bf output, \u03c0\u03cc\u03c4\u03b5 \u03b1\u03c0\u03b1\u03b9\u03c4\u03b5\u03af\u03c4\u03b1\u03b9 \u03b1\u03bd\u03b8\u03c1\u03ce\u03c0\u03b9\u03bd\u03b7 \u03b5\u03c0\u03b9\u03b2\u03b5\u03b2\u03b1\u03af\u03c9\u03c3\u03b7 \u03ba\u03b1\u03b9 \u03c0\u03ce\u03c2 \u03b5\u03c0\u03b1\u03bd\u03b1\u03c6\u03ad\u03c1\u03b5\u03c4\u03b1\u03b9 \u03b7 \u03c0\u03c1\u03bf\u03b7\u03b3\u03bf\u03cd\u03bc\u03b5\u03bd\u03b7 \u03ad\u03ba\u03b4\u03bf\u03c3\u03b7 \u03cc\u03c4\u03b1\u03bd \u03bd\u03ad\u03bf \u03bc\u03bf\u03bd\u03c4\u03ad\u03bb\u03bf \u03ae harness \u03c5\u03c0\u03bf\u03b2\u03b1\u03b8\u03bc\u03af\u03b6\u03b5\u03b9 \u03ba\u03c1\u03af\u03c3\u03b9\u03bc\u03b1 cases.<\/p>\n<\/li>\n<\/ol>\n<\/div>\n<p>\u0393\u03b9\u03b1 marketing, \u03c0\u03b5\u03c1\u03b9\u03b5\u03c7\u03cc\u03bc\u03b5\u03bd\u03bf \u03ba\u03b1\u03b9 analytics, \u03bf\u03b9 \u03ad\u03bb\u03b5\u03b3\u03c7\u03bf\u03b9 \u03bc\u03c0\u03bf\u03c1\u03bf\u03cd\u03bd \u03bd\u03b1 \u03c0\u03b5\u03c1\u03b9\u03bb\u03b1\u03bc\u03b2\u03ac\u03bd\u03bf\u03c5\u03bd \u03b5\u03b3\u03ba\u03c5\u03c1\u03cc\u03c4\u03b7\u03c4\u03b1 URLs, \u03b1\u03bd\u03c4\u03b9\u03c3\u03c4\u03bf\u03af\u03c7\u03b9\u03c3\u03b7 claims \u03bc\u03b5 \u03c0\u03b7\u03b3\u03ad\u03c2, \u03c3\u03c5\u03bc\u03c6\u03c9\u03bd\u03af\u03b1 totals \u03bc\u03b5 raw data, \u03b5\u03ba\u03c4\u03ad\u03bb\u03b5\u03c3\u03b7 formulas, \u03c3\u03c9\u03c3\u03c4\u03ae \u03b3\u03bb\u03ce\u03c3\u03c3\u03b1 \u03ba\u03b1\u03b9 \u03c0\u03bb\u03ae\u03c1\u03b7 \u03c0\u03b1\u03c1\u03ac\u03b4\u03bf\u03c3\u03b7 \u03b1\u03c1\u03c7\u03b5\u03af\u03c9\u03bd. \u0393\u03b9\u03b1 customer-facing agents \u03c7\u03c1\u03b5\u03b9\u03ac\u03b6\u03bf\u03bd\u03c4\u03b1\u03b9 \u03b5\u03c0\u03b9\u03c0\u03bb\u03ad\u03bf\u03bd tests \u03ba\u03bb\u03b9\u03bc\u03ac\u03ba\u03c9\u03c3\u03b7\u03c2, privacy, refusal behavior \u03ba\u03b1\u03b9 human handoff.<\/p>\n<p>\u03a4\u03b1 agentic tests \u03b4\u03b5\u03bd \u03b1\u03bd\u03c4\u03b9\u03ba\u03b1\u03b8\u03b9\u03c3\u03c4\u03bf\u03cd\u03bd \u03c4\u03bf\u03c5\u03c2 deterministic \u03b5\u03bb\u03ad\u03b3\u03c7\u03bf\u03c5\u03c2. \u038c\u03c0\u03c9\u03c2 \u03ba\u03b1\u03b9 \u03c3\u03b5 \u03ad\u03bd\u03b1 \u03b1\u03c3\u03c6\u03b1\u03bb\u03ad\u03c2 <a href=\"https:\/\/twodots.gr\/webmcp-ai-agents-asfales-website\/\">website \u03c0\u03bf\u03c5 \u03b5\u03ba\u03b8\u03ad\u03c4\u03b5\u03b9 \u03bb\u03b5\u03b9\u03c4\u03bf\u03c5\u03c1\u03b3\u03af\u03b5\u03c2 \u03c3\u03b5 AI agents<\/a>, schema validation, business rules, permissions, idempotency \u03ba\u03b1\u03b9 \u03c4\u03b5\u03bb\u03b9\u03ba\u03cc state \u03c0\u03c1\u03ad\u03c0\u03b5\u03b9 \u03bd\u03b1 \u03b5\u03bb\u03ad\u03b3\u03c7\u03bf\u03bd\u03c4\u03b1\u03b9 \u03bc\u03b5 \u03ba\u03b1\u03bd\u03bf\u03bd\u03b9\u03ba\u03ac unit, integration \u03ba\u03b1\u03b9 end-to-end tests.<\/p>\n<h2 id=\"symperasma\">\u03a4\u03bf \u03c3\u03c9\u03c3\u03c4\u03cc \u03c3\u03c5\u03bc\u03c0\u03ad\u03c1\u03b1\u03c3\u03bc\u03b1 \u03b4\u03b5\u03bd \u03b5\u03af\u03bd\u03b1\u03b9 \u03ad\u03bd\u03b1\u03c2 \u03bd\u03b9\u03ba\u03b7\u03c4\u03ae\u03c2<\/h2>\n<p>\u03a4\u03bf K-Bench \u03b4\u03b5\u03af\u03c7\u03bd\u03b5\u03b9 \u03cc\u03c4\u03b9 \u03c0\u03c1\u03b1\u03b3\u03bc\u03b1\u03c4\u03b9\u03ba\u03ac \u03b5\u03c0\u03b9\u03c3\u03c4\u03b7\u03bc\u03bf\u03bd\u03b9\u03ba\u03ac \u03b1\u03b9\u03c4\u03ae\u03bc\u03b1\u03c4\u03b1 \u03c0\u03b1\u03c1\u03b1\u03bc\u03ad\u03bd\u03bf\u03c5\u03bd \u03b4\u03cd\u03c3\u03ba\u03bf\u03bb\u03b1 \u03b1\u03ba\u03cc\u03bc\u03b7 \u03ba\u03b1\u03b9 \u03b3\u03b9\u03b1 \u03b9\u03c3\u03c7\u03c5\u03c1\u03ac \u03bc\u03bf\u03bd\u03c4\u03ad\u03bb\u03b1. \u039a\u03b1\u03bd\u03ad\u03bd\u03b1 \u03c3\u03cd\u03c3\u03c4\u03b7\u03bc\u03b1 \u03b4\u03b5\u03bd \u03be\u03b5\u03c0\u03b5\u03c1\u03bd\u03ac \u03b1\u03b4\u03b9\u03b1\u03bc\u03c6\u03b9\u03c3\u03b2\u03ae\u03c4\u03b7\u03c4\u03b1 \u03c4\u03bf acceptable threshold \u03ba\u03b1\u03b9 \u03c3\u03c4\u03bf\u03c5\u03c2 \u03c4\u03c1\u03b5\u03b9\u03c2 \u03ba\u03c1\u03b9\u03c4\u03ad\u03c2. \u0397 \u03c3\u03c7\u03b5\u03c4\u03b9\u03ba\u03ae \u03ba\u03b1\u03c4\u03ac\u03c4\u03b1\u03be\u03b7 \u03b5\u03af\u03bd\u03b1\u03b9 \u03c3\u03c4\u03b1\u03b8\u03b5\u03c1\u03cc\u03c4\u03b5\u03c1\u03b7 \u03b1\u03c0\u03cc \u03c4\u03bf \u03b1\u03c0\u03cc\u03bb\u03c5\u03c4\u03bf \u03b5\u03c0\u03af\u03c0\u03b5\u03b4\u03bf, \u03b5\u03bd\u03ce \u03b7 \u03b5\u03c0\u03b9\u03c3\u03c4\u03b7\u03bc\u03bf\u03bd\u03b9\u03ba\u03ae \u03b1\u03ba\u03c1\u03af\u03b2\u03b5\u03b9\u03b1, \u03b7 \u03c0\u03bf\u03b9\u03cc\u03c4\u03b7\u03c4\u03b1 artifacts \u03ba\u03b1\u03b9 \u03c4\u03bf overclaiming \u03b1\u03c0\u03bf\u03ba\u03b1\u03bb\u03cd\u03c0\u03c4\u03bf\u03c5\u03bd \u03b1\u03b4\u03c5\u03bd\u03b1\u03bc\u03af\u03b5\u03c2 \u03c0\u03bf\u03c5 \u03bc\u03b9\u03b1 \u03b1\u03c0\u03bb\u03ae \u03b1\u03bd\u03ac\u03b3\u03bd\u03c9\u03c3\u03b7 \u03c4\u03bf\u03c5 \u03c4\u03b5\u03bb\u03b9\u03ba\u03bf\u03cd \u03ba\u03b5\u03b9\u03bc\u03ad\u03bd\u03bf\u03c5 \u03b8\u03b1 \u03ad\u03c7\u03b1\u03bd\u03b5.<\/p>\n<p>\u03a4\u03bf benchmark \u03ad\u03c7\u03b5\u03b9 \u03ba\u03b1\u03b9 \u03bf\u03c5\u03c3\u03b9\u03b1\u03c3\u03c4\u03b9\u03ba\u03bf\u03cd\u03c2 \u03c0\u03b5\u03c1\u03b9\u03bf\u03c1\u03b9\u03c3\u03bc\u03bf\u03cd\u03c2: \u03b5\u03c3\u03c9\u03c4\u03b5\u03c1\u03b9\u03ba\u03cc corpus, LLM-only judging, \u03bc\u03af\u03b1 \u03c0\u03c1\u03bf\u03c3\u03c0\u03ac\u03b8\u03b5\u03b9\u03b1 \u03b1\u03bd\u03ac run, \u03b9\u03b4\u03b9\u03c9\u03c4\u03b9\u03ba\u03ac tasks, complete-case exclusion \u03ba\u03b1\u03b9 \u03bc\u03b7 de-identified replay \u03c0\u03c1\u03bf\u03c2 \u03c4\u03c1\u03af\u03c4\u03bf\u03c5\u03c2 \u03c0\u03b1\u03c1\u03cc\u03c7\u03bf\u03c5\u03c2. \u039f\u03b9 \u03c0\u03b5\u03c1\u03b9\u03bf\u03c1\u03b9\u03c3\u03bc\u03bf\u03af \u03b4\u03b5\u03bd \u03b5\u03af\u03bd\u03b1\u03b9 \u03c5\u03c0\u03bf\u03c3\u03b7\u03bc\u03b5\u03af\u03c9\u03c3\u03b7\u00b7 \u03ba\u03b1\u03b8\u03bf\u03c1\u03af\u03b6\u03bf\u03c5\u03bd \u03c0\u03bf\u03cd \u03bc\u03c0\u03bf\u03c1\u03b5\u03af \u03ba\u03b1\u03b9 \u03c0\u03bf\u03cd \u03b4\u03b5\u03bd \u03bc\u03c0\u03bf\u03c1\u03b5\u03af \u03bd\u03b1 \u03bc\u03b5\u03c4\u03b1\u03c6\u03b5\u03c1\u03b8\u03b5\u03af \u03c4\u03bf \u03b1\u03c0\u03bf\u03c4\u03ad\u03bb\u03b5\u03c3\u03bc\u03b1.<\/p>\n<p>\u0393\u03b9\u03b1 \u03c4\u03bf\u03c5\u03c2 \u03c5\u03c0\u03b5\u03cd\u03b8\u03c5\u03bd\u03bf\u03c5\u03c2 AI, \u03c4\u03bf \u03ba\u03b5\u03bd\u03c4\u03c1\u03b9\u03ba\u03cc \u03bc\u03ac\u03b8\u03b7\u03bc\u03b1 \u03b5\u03af\u03bd\u03b1\u03b9 \u03b1\u03c0\u03bb\u03cc: \u03b1\u03be\u03b9\u03bf\u03bb\u03bf\u03b3\u03ae\u03c3\u03c4\u03b5 \u03c4\u03b7 \u03b4\u03bf\u03c5\u03bb\u03b5\u03b9\u03ac, \u03cc\u03c7\u03b9 \u03bc\u03cc\u03bd\u03bf \u03c4\u03b7\u03bd \u03b1\u03c0\u03ac\u03bd\u03c4\u03b7\u03c3\u03b7. \u0391\u03bd\u03bf\u03af\u03be\u03c4\u03b5 \u03c4\u03b1 \u03b1\u03c1\u03c7\u03b5\u03af\u03b1, \u03b5\u03ba\u03c4\u03b5\u03bb\u03ad\u03c3\u03c4\u03b5 \u03c4\u03bf\u03bd \u03ba\u03ce\u03b4\u03b9\u03ba\u03b1, \u03b5\u03bb\u03ad\u03b3\u03be\u03c4\u03b5 \u03c4\u03b1 \u03b4\u03b5\u03b4\u03bf\u03bc\u03ad\u03bd\u03b1, \u03b6\u03b7\u03c4\u03ae\u03c3\u03c4\u03b5 \u03c4\u03b5\u03ba\u03bc\u03ae\u03c1\u03b9\u03b1 \u03b3\u03b9\u03b1 \u03c4\u03bf\u03c5\u03c2 \u03b9\u03c3\u03c7\u03c5\u03c1\u03b9\u03c3\u03bc\u03bf\u03cd\u03c2 \u03ba\u03b1\u03b9 \u03bc\u03b5\u03c4\u03c1\u03ae\u03c3\u03c4\u03b5 \u03b1\u03bd \u03c4\u03bf \u03b1\u03c0\u03bf\u03c4\u03ad\u03bb\u03b5\u03c3\u03bc\u03b1 \u03b9\u03ba\u03b1\u03bd\u03bf\u03c0\u03bf\u03b9\u03b5\u03af \u03c4\u03bf \u03c0\u03c1\u03b1\u03b3\u03bc\u03b1\u03c4\u03b9\u03ba\u03cc \u03b1\u03af\u03c4\u03b7\u03bc\u03b1. \u0397 \u03b5\u03c5\u03c6\u03c1\u03ac\u03b4\u03b5\u03b9\u03b1 \u03b5\u03af\u03bd\u03b1\u03b9 \u03c7\u03c1\u03ae\u03c3\u03b9\u03bc\u03b7\u00b7 \u03b4\u03b5\u03bd \u03b5\u03af\u03bd\u03b1\u03b9 \u03b1\u03c0\u03cc\u03b4\u03b5\u03b9\u03be\u03b7 \u03b1\u03ba\u03c1\u03af\u03b2\u03b5\u03b9\u03b1\u03c2.<\/p>\n<section class=\"td-service-cta\">\n<div class=\"td-service-cta-content\">\n<p class=\"td-service-cta-kicker\">\u0391\u03c0\u03cc \u03c4\u03bf demo \u03c3\u03b5 \u03bc\u03b5\u03c4\u03c1\u03ae\u03c3\u03b9\u03bc\u03b7 \u03b1\u03be\u03b9\u03bf\u03c0\u03b9\u03c3\u03c4\u03af\u03b1<\/p>\n<p class=\"td-service-cta-title\">\u03a3\u03c7\u03b5\u03b4\u03b9\u03ac\u03c3\u03c4\u03b5 AI \u03b1\u03c5\u03c4\u03bf\u03bc\u03b1\u03c4\u03b9\u03c3\u03bc\u03bf\u03cd\u03c2 \u03bc\u03b5 acceptance tests \u03b3\u03b9\u03b1 \u03c4\u03bf \u03c0\u03c1\u03b1\u03b3\u03bc\u03b1\u03c4\u03b9\u03ba\u03cc \u03b1\u03c0\u03bf\u03c4\u03ad\u03bb\u03b5\u03c3\u03bc\u03b1<\/p>\n<p>\u0397 TWO DOTS \u03c7\u03b1\u03c1\u03c4\u03bf\u03b3\u03c1\u03b1\u03c6\u03b5\u03af \u03c4\u03b1 \u03ba\u03c1\u03af\u03c3\u03b9\u03bc\u03b1 tasks, \u03bf\u03c1\u03af\u03b6\u03b5\u03b9 rubrics \u03ba\u03b1\u03b9 deterministic checks, \u03b5\u03bb\u03ad\u03b3\u03c7\u03b5\u03b9 artifacts, permissions \u03ba\u03b1\u03b9 handoffs \u03ba\u03b1\u03b9 \u03c3\u03c5\u03bd\u03b4\u03ad\u03b5\u03b9 \u03ba\u03ac\u03b8\u03b5 AI workflow \u03bc\u03b5 monitoring, \u03b1\u03bd\u03b8\u03c1\u03ce\u03c0\u03b9\u03bd\u03b7 \u03b5\u03c0\u03bf\u03c0\u03c4\u03b5\u03af\u03b1 \u03ba\u03b1\u03b9 rollback.<\/p>\n<div class=\"td-service-cta-actions\"><a class=\"td-service-cta-button\" href=\"https:\/\/twodots.gr\/aftomatismoi-epicheiriseon-ai\/\">\u03a3\u03c5\u03b6\u03b7\u03c4\u03ae\u03c3\u03c4\u03b5 \u03ad\u03bd\u03b1\u03bd \u03b5\u03bb\u03b5\u03b3\u03c7\u03cc\u03bc\u03b5\u03bd\u03bf AI \u03b1\u03c5\u03c4\u03bf\u03bc\u03b1\u03c4\u03b9\u03c3\u03bc\u03cc<\/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 K-Bench 01;<\/summary>\n<div class=\"td-faq-content\">\n<p>\u0395\u03af\u03bd\u03b1\u03b9 benchmark \u03bc\u03b5 178 \u03c0\u03c1\u03ce\u03c4\u03b1 \u03b1\u03b9\u03c4\u03ae\u03bc\u03b1\u03c4\u03b1 \u03b1\u03c0\u03cc \u03c0\u03c1\u03b1\u03b3\u03bc\u03b1\u03c4\u03b9\u03ba\u03ad\u03c2 \u03c3\u03c5\u03bd\u03b5\u03b4\u03c1\u03af\u03b5\u03c2 \u03b5\u03c0\u03b9\u03c3\u03c4\u03b7\u03bc\u03bf\u03bd\u03b9\u03ba\u03ce\u03bd \u03c7\u03c1\u03b7\u03c3\u03c4\u03ce\u03bd. \u0395\u03bd\u03bd\u03ad\u03b1 \u03bc\u03bf\u03bd\u03c4\u03ad\u03bb\u03b1 \u03c4\u03b1 \u03b5\u03ba\u03c4\u03ad\u03bb\u03b5\u03c3\u03b1\u03bd \u03c3\u03b5 \u03af\u03b4\u03b9\u03b1 sandboxes \u03ba\u03b1\u03b9 \u03c4\u03c1\u03b5\u03b9\u03c2 LLM judges \u03b1\u03be\u03b9\u03bf\u03bb\u03cc\u03b3\u03b7\u03c3\u03b1\u03bd \u03c4\u03bf transcript \u03ba\u03b1\u03b9 \u03c4\u03b1 \u03c0\u03b1\u03c1\u03b1\u03b3\u03cc\u03bc\u03b5\u03bd\u03b1 \u03b1\u03c1\u03c7\u03b5\u03af\u03b1.<\/p>\n<\/div>\n<\/details>\n<details class=\"td-faq-item\">\n<summary class=\"td-faq-title\">\u03a0\u03cc\u03c3\u03b1 runs \u03ba\u03b1\u03b9 assessments \u03c0\u03b5\u03c1\u03b9\u03bb\u03b1\u03bc\u03b2\u03ac\u03bd\u03b5\u03b9;<\/summary>\n<div class=\"td-faq-content\">\n<p>\u039f\u03bb\u03bf\u03ba\u03bb\u03b7\u03c1\u03ce\u03b8\u03b7\u03ba\u03b1\u03bd 1.602 runs. \u03a4\u03c1\u03b5\u03b9\u03c2 \u03ba\u03c1\u03b9\u03c4\u03ad\u03c2 \u03c0\u03b1\u03c1\u03ae\u03b3\u03b1\u03b3\u03b1\u03bd 4.806 assessments \u03ba\u03b1\u03b9 \u03c3\u03c5\u03bd\u03bf\u03bb\u03b9\u03ba\u03ac 39.934 \u03b5\u03c6\u03b1\u03c1\u03bc\u03bf\u03c3\u03bc\u03ad\u03bd\u03b5\u03c2 \u03b2\u03b1\u03b8\u03bc\u03bf\u03bb\u03bf\u03b3\u03ae\u03c3\u03b5\u03b9\u03c2 \u03b4\u03b9\u03b1\u03c3\u03c4\u03ac\u03c3\u03b5\u03c9\u03bd \u03ae \u03c3\u03c5\u03bd\u03bf\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\u03bf\u03b9\u03bf \u03bc\u03bf\u03bd\u03c4\u03ad\u03bb\u03bf \u03ae\u03c4\u03b1\u03bd \u03c0\u03c1\u03ce\u03c4\u03bf;<\/summary>\n<div class=\"td-faq-content\">\n<p>\u03a4\u03bf gpt-5.6-sol \u03b5\u03af\u03c7\u03b5 \u03c4\u03bf\u03bd \u03c5\u03c8\u03b7\u03bb\u03cc\u03c4\u03b5\u03c1\u03bf pooled \u03bc\u03ad\u03c3\u03bf \u03cc\u03c1\u03bf, 8,04, \u03b1\u03bb\u03bb\u03ac \u03c4\u03bf 95% interval \u03c0\u03b5\u03c1\u03bd\u03bf\u03cd\u03c3\u03b5 \u03c4\u03bf \u03cc\u03c1\u03b9\u03bf 8 \u03ba\u03b1\u03b9 \u03b4\u03cd\u03bf \u03b1\u03c0\u03cc \u03c4\u03bf\u03c5\u03c2 \u03c4\u03c1\u03b5\u03b9\u03c2 \u03ba\u03c1\u03b9\u03c4\u03ad\u03c2 \u03ad\u03b2\u03b1\u03bb\u03b1\u03bd \u03c0\u03c1\u03ce\u03c4\u03bf \u03c4\u03bf claude-opus-5. \u0397 \u03ba\u03bf\u03c1\u03c5\u03c6\u03ae \u03c0\u03b1\u03c1\u03b1\u03bc\u03ad\u03bd\u03b5\u03b9 \u03bc\u03b7 \u03b5\u03c0\u03b9\u03bb\u03c5\u03bc\u03ad\u03bd\u03b7.<\/p>\n<\/div>\n<\/details>\n<details class=\"td-faq-item\">\n<summary class=\"td-faq-title\">\u03a0\u03bf\u03b9\u03bf \u03ae\u03c4\u03b1\u03bd \u03c4\u03bf \u03b2\u03b1\u03c3\u03b9\u03ba\u03cc \u03ba\u03b5\u03bd\u03cc \u03c0\u03bf\u03b9\u03cc\u03c4\u03b7\u03c4\u03b1\u03c2;<\/summary>\n<div class=\"td-faq-content\">\n<p>\u0397 \u03b5\u03c0\u03b9\u03ba\u03bf\u03b9\u03bd\u03c9\u03bd\u03af\u03b1 \u03b5\u03af\u03c7\u03b5 \u03bc\u03ad\u03c3\u03bf 7,33 \u03ba\u03b1\u03b9 \u03b7 \u03b5\u03c0\u03b9\u03c3\u03c4\u03b7\u03bc\u03bf\u03bd\u03b9\u03ba\u03ae \u03b1\u03ba\u03c1\u03af\u03b2\u03b5\u03b9\u03b1 6,22, \u03bc\u03b5 \u03c4\u03b7\u03bd \u03b5\u03c0\u03b9\u03ba\u03bf\u03b9\u03bd\u03c9\u03bd\u03af\u03b1 \u03c5\u03c8\u03b7\u03bb\u03cc\u03c4\u03b5\u03c1\u03b7 \u03ba\u03b1\u03b9 \u03c3\u03c4\u03b1 \u03b5\u03bd\u03bd\u03ad\u03b1 \u03bc\u03bf\u03bd\u03c4\u03ad\u03bb\u03b1. \u0397 \u03c0\u03b5\u03b9\u03c3\u03c4\u03b9\u03ba\u03ae \u03b4\u03b9\u03b1\u03c4\u03cd\u03c0\u03c9\u03c3\u03b7 \u03b5\u03c0\u03bf\u03bc\u03ad\u03bd\u03c9\u03c2 \u03b4\u03b5\u03bd \u03b1\u03c1\u03ba\u03b5\u03af \u03c9\u03c2 \u03ad\u03bd\u03b4\u03b5\u03b9\u03be\u03b7 \u03bf\u03c1\u03b8\u03cc\u03c4\u03b7\u03c4\u03b1\u03c2.<\/p>\n<\/div>\n<\/details>\n<details class=\"td-faq-item\">\n<summary class=\"td-faq-title\">\u03a0\u03bf\u03b9\u03bf failure tag \u03ae\u03c4\u03b1\u03bd \u03c3\u03c5\u03c7\u03bd\u03cc\u03c4\u03b5\u03c1\u03bf;<\/summary>\n<div class=\"td-faq-content\">\n<p>\u03a4\u03bf overclaiming, \u03c3\u03c4\u03bf 31,4% \u03c4\u03c9\u03bd assessments. \u0395\u03c0\u03b5\u03b9\u03b4\u03ae \u03c4\u03b1 failure tags \u03ae\u03c4\u03b1\u03bd judge-assigned \u03ba\u03b1\u03b9 \u03cc\u03c7\u03b9 \u03b1\u03bc\u03bf\u03b9\u03b2\u03b1\u03af\u03b1 \u03b1\u03c0\u03bf\u03ba\u03bb\u03b5\u03b9\u03cc\u03bc\u03b5\u03bd\u03b1, \u03ba\u03ac\u03b8\u03b5 \u03c0\u03bf\u03c3\u03bf\u03c3\u03c4\u03cc \u03c0\u03c1\u03ad\u03c0\u03b5\u03b9 \u03bd\u03b1 \u03b4\u03b9\u03b1\u03b2\u03ac\u03b6\u03b5\u03c4\u03b1\u03b9 \u03bc\u03b1\u03b6\u03af \u03bc\u03b5 \u03c4\u03b7 \u03bc\u03bf\u03bd\u03ac\u03b4\u03b1 aggregation.<\/p>\n<\/div>\n<\/details>\n<details class=\"td-faq-item\">\n<summary class=\"td-faq-title\">\u0393\u03b9\u03b1\u03c4\u03af \u03ad\u03c7\u03bf\u03c5\u03bd \u03c3\u03b7\u03bc\u03b1\u03c3\u03af\u03b1 \u03c4\u03b1 artifacts;<\/summary>\n<div class=\"td-faq-content\">\n<p>\u0395\u03c0\u03b5\u03b9\u03b4\u03ae \u03ad\u03bd\u03b1\u03c2 agent \u03bc\u03c0\u03bf\u03c1\u03b5\u03af \u03bd\u03b1 \u03c0\u03b5\u03c1\u03b9\u03b3\u03c1\u03ac\u03c8\u03b5\u03b9 \u03ba\u03b1\u03bb\u03ac \u03bc\u03b9\u03b1 \u03b5\u03c1\u03b3\u03b1\u03c3\u03af\u03b1 \u03c7\u03c9\u03c1\u03af\u03c2 \u03bd\u03b1 \u03ad\u03c7\u03b5\u03b9 \u03c0\u03b1\u03c1\u03b1\u03b4\u03ce\u03c3\u03b5\u03b9 \u03bb\u03b5\u03b9\u03c4\u03bf\u03c5\u03c1\u03b3\u03b9\u03ba\u03cc \u03ba\u03ce\u03b4\u03b9\u03ba\u03b1, \u03c3\u03c9\u03c3\u03c4\u03ac \u03b4\u03b5\u03b4\u03bf\u03bc\u03ad\u03bd\u03b1 \u03ae \u03c4\u03b1 \u03b6\u03b7\u03c4\u03bf\u03cd\u03bc\u03b5\u03bd\u03b1 \u03b1\u03c1\u03c7\u03b5\u03af\u03b1. \u03a4\u03bf 47,9% \u03c4\u03c9\u03bd runs \u03c4\u03b5\u03bb\u03b5\u03af\u03c9\u03c3\u03b5 \u03c7\u03c9\u03c1\u03af\u03c2 \u03b1\u03c1\u03c7\u03b5\u03af\u03bf \u03c3\u03c4\u03bf output tree, \u03b1\u03bd \u03ba\u03b1\u03b9 \u03cc\u03c7\u03b9 \u03ba\u03ac\u03b8\u03b5 task \u03b1\u03c0\u03b1\u03b9\u03c4\u03bf\u03cd\u03c3\u03b5 artifact.<\/p>\n<\/div>\n<\/details>\n<details class=\"td-faq-item\">\n<summary class=\"td-faq-title\">\u0389\u03c4\u03b1\u03bd \u03bf\u03b9 \u03ba\u03c1\u03b9\u03c4\u03ad\u03c2 \u03ac\u03bd\u03b8\u03c1\u03c9\u03c0\u03bf\u03b9;<\/summary>\n<div class=\"td-faq-content\">\n<p>\u038c\u03c7\u03b9. \u0389\u03c4\u03b1\u03bd \u03c4\u03c1\u03af\u03b1 language models. \u0394\u03b5\u03bd \u03c5\u03c0\u03ae\u03c1\u03c7\u03b5 human calibration \u03ae expert baseline, \u03b5\u03c0\u03bf\u03bc\u03ad\u03bd\u03c9\u03c2 \u03c4\u03bf \u03b1\u03c0\u03cc\u03bb\u03c5\u03c4\u03bf \u03b5\u03c0\u03af\u03c0\u03b5\u03b4\u03bf \u03c4\u03c9\u03bd scores \u03b5\u03be\u03b1\u03c1\u03c4\u03ac\u03c4\u03b1\u03b9 \u03b1\u03c0\u03cc \u03c4\u03bf judge panel \u03ba\u03b1\u03b9 \u03c4\u03bf rubric.<\/p>\n<\/div>\n<\/details>\n<details class=\"td-faq-item\">\n<summary class=\"td-faq-title\">\u039c\u03c0\u03bf\u03c1\u03b5\u03af \u03c4\u03bf benchmark \u03bd\u03b1 \u03b1\u03bd\u03b1\u03c0\u03b1\u03c1\u03b1\u03c7\u03b8\u03b5\u03af \u03c0\u03bb\u03ae\u03c1\u03c9\u03c2;<\/summary>\n<div class=\"td-faq-content\">\n<p>\u038c\u03c7\u03b9 \u03bc\u03b5 \u03bd\u03ad\u03b1 \u03bc\u03bf\u03bd\u03c4\u03ad\u03bb\u03b1 \u03c3\u03c4\u03b1 \u03af\u03b4\u03b9\u03b1 tasks, \u03b5\u03c0\u03b5\u03b9\u03b4\u03ae prompts, attachments, transcripts \u03ba\u03b1\u03b9 artifacts \u03b4\u03b5\u03bd \u03b4\u03b7\u03bc\u03bf\u03c3\u03b9\u03b5\u03cd\u03bf\u03bd\u03c4\u03b1\u03b9. \u03a4\u03b1 de-identified scores, \u03c4\u03bf analysis code \u03ba\u03b1\u03b9 \u03c4\u03bf rubric \u03b5\u03c0\u03b9\u03c4\u03c1\u03ad\u03c0\u03bf\u03c5\u03bd \u03ad\u03bb\u03b5\u03b3\u03c7\u03bf \u03c4\u03c9\u03bd \u03c5\u03c0\u03bf\u03bb\u03bf\u03b3\u03b9\u03c3\u03bc\u03ce\u03bd, \u03cc\u03c7\u03b9 \u03c0\u03bb\u03ae\u03c1\u03b7 \u03b5\u03c0\u03b1\u03bd\u03ac\u03bb\u03b7\u03c8\u03b7 \u03c4\u03b7\u03c2 \u03ba\u03b1\u03bc\u03c0\u03ac\u03bd\u03b9\u03b1\u03c2.<\/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\/2608.21601\" target=\"_blank\" rel=\"noopener\">arXiv \u2014 K-Bench: measuring model performance on real scientific agent requests<\/a><\/li>\n<li><a href=\"https:\/\/arxiv.org\/html\/2608.21601\" target=\"_blank\" rel=\"noopener\">arXiv HTML \u2014 \u03c0\u03bb\u03ae\u03c1\u03b7\u03c2 \u03bc\u03b5\u03b8\u03bf\u03b4\u03bf\u03bb\u03bf\u03b3\u03af\u03b1, \u03c0\u03af\u03bd\u03b1\u03ba\u03b5\u03c2, rubric \u03ba\u03b1\u03b9 \u03c0\u03b5\u03c1\u03b9\u03bf\u03c1\u03b9\u03c3\u03bc\u03bf\u03af \u03c4\u03bf\u03c5 K-Bench 01<\/a><\/li>\n<li><a href=\"https:\/\/www.k-dense.ai\/blog\/introducing-k-bench-01-internal-benchmark\" target=\"_blank\" rel=\"noopener\">K-Dense \u2014 \u03c0\u03b1\u03c1\u03bf\u03c5\u03c3\u03af\u03b1\u03c3\u03b7 \u03c4\u03bf\u03c5 \u03b5\u03c3\u03c9\u03c4\u03b5\u03c1\u03b9\u03ba\u03bf\u03cd benchmark \u03ba\u03b1\u03b9 \u03c4\u03c9\u03bd \u03c0\u03c1\u03b1\u03ba\u03c4\u03b9\u03ba\u03ce\u03bd \u03b5\u03c5\u03c1\u03b7\u03bc\u03ac\u03c4\u03c9\u03bd<\/a><\/li>\n<li><a href=\"https:\/\/arxiv.org\/abs\/2410.05080\" target=\"_blank\" rel=\"noopener\">arXiv \u2014 ScienceAgentBench \u03b3\u03b9\u03b1 \u03b1\u03be\u03b9\u03bf\u03bb\u03cc\u03b3\u03b7\u03c3\u03b7 agents \u03c3\u03b5 data-driven \u03b5\u03c0\u03b9\u03c3\u03c4\u03b7\u03bc\u03bf\u03bd\u03b9\u03ba\u03ad\u03c2 \u03b5\u03c1\u03b3\u03b1\u03c3\u03af\u03b5\u03c2<\/a><\/li>\n<li><a href=\"https:\/\/www.nist.gov\/itl\/ai-risk-management-framework\" target=\"_blank\" rel=\"noopener\">NIST \u2014 AI Risk Management Framework<\/a><\/li>\n<\/ul>\n<\/div>","protected":false},"excerpt":{"rendered":"<p>\u03a4\u03bf K-Bench \u03b4\u03b5\u03af\u03c7\u03bd\u03b5\u03b9 \u03b3\u03b9\u03b1\u03c4\u03af \u03b7 \u03b1\u03be\u03b9\u03bf\u03bb\u03cc\u03b3\u03b7\u03c3\u03b7 AI agents \u03c0\u03c1\u03ad\u03c0\u03b5\u03b9 \u03bd\u03b1 \u03b5\u03bb\u03ad\u03b3\u03c7\u03b5\u03b9 \u03b1\u03ba\u03c1\u03af\u03b2\u03b5\u03b9\u03b1, artifacts \u03ba\u03b1\u03b9 overclaiming \u2014 \u03cc\u03c7\u03b9 \u03bc\u03cc\u03bd\u03bf \u03bc\u03b9\u03b1 \u03c0\u03b5\u03b9\u03c3\u03c4\u03b9\u03ba\u03ae \u03b1\u03c0\u03ac\u03bd\u03c4\u03b7\u03c3\u03b7.<\/p>","protected":false},"author":1,"featured_media":98390,"comment_status":"open","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"_gspb_post_css":"","content-type":"","footnotes":""},"categories":[199],"tags":[20615,6506,17818,9260,19919],"class_list":["post-97801","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-techniti-noimosyni","tag-ai-accuracy","tag-ai-agents","tag-ai-benchmarks","tag-ai-governance","tag-scientific-ai"],"blocksy_meta":{"styles_descriptor":{"styles":{"desktop":"","tablet":"","mobile":""},"google_fonts":[],"version":8}},"_links":{"self":[{"href":"https:\/\/twodots.gr\/en\/wp-json\/wp\/v2\/posts\/97801","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=97801"}],"version-history":[{"count":1,"href":"https:\/\/twodots.gr\/en\/wp-json\/wp\/v2\/posts\/97801\/revisions"}],"predecessor-version":[{"id":98391,"href":"https:\/\/twodots.gr\/en\/wp-json\/wp\/v2\/posts\/97801\/revisions\/98391"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/twodots.gr\/en\/wp-json\/wp\/v2\/media\/98390"}],"wp:attachment":[{"href":"https:\/\/twodots.gr\/en\/wp-json\/wp\/v2\/media?parent=97801"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/twodots.gr\/en\/wp-json\/wp\/v2\/categories?post=97801"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/twodots.gr\/en\/wp-json\/wp\/v2\/tags?post=97801"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}