{"id":90929,"date":"2026-09-03T11:53:58","date_gmt":"2026-09-03T08:53:58","guid":{"rendered":"https:\/\/twodots.gr\/?p=90929"},"modified":"2026-09-03T11:54:00","modified_gmt":"2026-09-03T08:54:00","slug":"sparc-ai-provlepsi-anthropinis-kinisis-avevaiotita","status":"publish","type":"post","link":"https:\/\/twodots.gr\/bg\/sparc-ai-provlepsi-anthropinis-kinisis-avevaiotita\/","title":{"rendered":"SPARC: \u03c0\u03ce\u03c2 \u03b7 AI \u03c0\u03c1\u03bf\u03b2\u03bb\u03ad\u03c0\u03b5\u03b9 \u03c4\u03b7\u03bd \u03b1\u03bd\u03b8\u03c1\u03ce\u03c0\u03b9\u03bd\u03b7 \u03ba\u03af\u03bd\u03b7\u03c3\u03b7 \u03bc\u03b1\u03b6\u03af \u03bc\u03b5 \u03c4\u03b7\u03bd \u03b1\u03b2\u03b5\u03b2\u03b1\u03b9\u03cc\u03c4\u03b7\u03c4\u03ac \u03c4\u03b7\u03c2"},"content":{"rendered":"<div class=\"td-article-lede\">\n<p><strong>\u03a4\u03bf SPARC \u03b4\u03b5\u03bd \u03c0\u03c1\u03bf\u03c3\u03c0\u03b1\u03b8\u03b5\u03af \u03b1\u03c0\u03bb\u03ce\u03c2 \u03bd\u03b1 \u03bc\u03b1\u03bd\u03c4\u03ad\u03c8\u03b5\u03b9 \u03c4\u03b7\u03bd \u03b5\u03c0\u03cc\u03bc\u03b5\u03bd\u03b7 \u03ba\u03af\u03bd\u03b7\u03c3\u03b7 \u03b5\u03bd\u03cc\u03c2 \u03b1\u03bd\u03b8\u03c1\u03ce\u03c0\u03bf\u03c5\u0387 \u03c0\u03c1\u03bf\u03c3\u03c0\u03b1\u03b8\u03b5\u03af \u03bd\u03b1 \u03b4\u03b5\u03af\u03be\u03b5\u03b9 \u03c0\u03cc\u03c3\u03bf \u03c7\u03ce\u03c1\u03bf \u03c0\u03c1\u03ad\u03c0\u03b5\u03b9 \u03bd\u03b1 \u03b1\u03c6\u03ae\u03c3\u03b5\u03b9 \u03ad\u03bd\u03b1 \u03c3\u03cd\u03c3\u03c4\u03b7\u03bc\u03b1 \u03b3\u03b9\u03b1 \u03c4\u03bf \u03b5\u03bd\u03b4\u03b5\u03c7\u03cc\u03bc\u03b5\u03bd\u03bf \u03bd\u03b1 \u03ba\u03ac\u03bd\u03b5\u03b9 \u03bb\u03ac\u03b8\u03bf\u03c2.<\/strong> \u03a3\u03c5\u03bd\u03b4\u03c5\u03ac\u03b6\u03b5\u03b9 deterministic \u03c0\u03c1\u03cc\u03b2\u03bb\u03b5\u03c8\u03b7, \u03b4\u03bf\u03bc\u03b7\u03bc\u03ad\u03bd\u03b7 aleatoric covariance, \u03b1\u03bd\u03b1\u03bb\u03c5\u03c4\u03b9\u03ba\u03ae epistemic \u03ba\u03bb\u03b9\u03bc\u03ac\u03ba\u03c9\u03c3\u03b7 \u03ba\u03b1\u03b9 split conformal calibration \u03c3\u03b5 \u03ad\u03bd\u03b1 forward pass. \u0397 \u03c0\u03c1\u03b1\u03ba\u03c4\u03b9\u03ba\u03ae \u03c4\u03bf\u03c5 \u03b1\u03be\u03af\u03b1 \u03b5\u03af\u03bd\u03b1\u03b9 \u03ad\u03bd\u03b1 \u03b5\u03bb\u03ad\u03b3\u03be\u03b9\u03bc\u03bf risk interface \u03b3\u03b9\u03b1 \u03c1\u03bf\u03bc\u03c0\u03cc\u03c4 \u03ba\u03b1\u03b9 safety-aware AI, \u03cc\u03c7\u03b9 \u03ad\u03bd\u03b1\u03c2 \u03b3\u03b5\u03bd\u03b9\u03ba\u03cc\u03c2 \u00ab\u03b4\u03b5\u03af\u03ba\u03c4\u03b7\u03c2 \u03b5\u03bc\u03c0\u03b9\u03c3\u03c4\u03bf\u03c3\u03cd\u03bd\u03b7\u03c2\u00bb.<\/p>\n<\/div>\n<div class=\"td-article-toc\">\n<div class=\"td-toc-title\">\u03a0\u03b5\u03c1\u03b9\u03b5\u03c7\u03cc\u03bc\u03b5\u03bd\u03b1<\/div>\n<ul>\n<li><a href=\"#point-forecast-den-arkei\">\u0393\u03b9\u03b1\u03c4\u03af \u03ad\u03bd\u03b1 \u03c3\u03c9\u03c3\u03c4\u03cc point forecast \u03b4\u03b5\u03bd \u03b5\u03af\u03bd\u03b1\u03b9 \u03b1\u03c1\u03ba\u03b5\u03c4\u03cc<\/a><\/li>\n<li><a href=\"#dyo-typoi-avevaiotitas\">Aleatoric \u03ba\u03b1\u03b9 epistemic \u03b1\u03b2\u03b5\u03b2\u03b1\u03b9\u03cc\u03c4\u03b7\u03c4\u03b1 \u03b4\u03b5\u03bd \u03b5\u03af\u03bd\u03b1\u03b9 \u03c4\u03bf \u03af\u03b4\u03b9\u03bf<\/a><\/li>\n<li><a href=\"#pos-leitourgei-sparc\">\u03a0\u03ce\u03c2 \u03bb\u03b5\u03b9\u03c4\u03bf\u03c5\u03c1\u03b3\u03b5\u03af \u03c4\u03bf SPARC \u03c3\u03b5 \u03c4\u03c1\u03af\u03b1 \u03b5\u03c0\u03af\u03c0\u03b5\u03b4\u03b1<\/a><\/li>\n<li><a href=\"#ti-metrai-kappa\">\u03a4\u03b9 \u03bc\u03b5\u03c4\u03c1\u03ac \u03c0\u03c1\u03b1\u03b3\u03bc\u03b1\u03c4\u03b9\u03ba\u03ac \u03c4\u03bf \u03ba<\/a><\/li>\n<li><a href=\"#prediction-tubes\">Prediction tubes \u03b1\u03bd\u03c4\u03af \u03b3\u03b9\u03b1 \u03b1\u03c3\u03b1\u03c6\u03ad\u03c2 confidence score<\/a><\/li>\n<li><a href=\"#peiramatiko-protokollo\">\u03a4\u03bf \u03c0\u03b5\u03b9\u03c1\u03b1\u03bc\u03b1\u03c4\u03b9\u03ba\u03cc \u03c0\u03c1\u03c9\u03c4\u03cc\u03ba\u03bf\u03bb\u03bb\u03bf \u03ba\u03b1\u03b9 \u03c4\u03b1 \u03b4\u03b5\u03b4\u03bf\u03bc\u03ad\u03bd\u03b1<\/a><\/li>\n<li><a href=\"#vasika-apotelesmata\">\u03a4\u03b1 \u03b2\u03b1\u03c3\u03b9\u03ba\u03ac \u03b1\u03c0\u03bf\u03c4\u03b5\u03bb\u03ad\u03c3\u03bc\u03b1\u03c4\u03b1 \u03c7\u03c9\u03c1\u03af\u03c2 \u03c5\u03c0\u03b5\u03c1\u03b2\u03bf\u03bb\u03ad\u03c2<\/a><\/li>\n<li><a href=\"#trade-offs-ablation\">\u03a4\u03b9 \u03b1\u03c0\u03bf\u03ba\u03ac\u03bb\u03c5\u03c8\u03b5 \u03c4\u03bf ablation \u03b3\u03b9\u03b1 \u03c4\u03b1 trade-offs<\/a><\/li>\n<li><a href=\"#single-pass-compute\">\u0388\u03bd\u03b1 forward pass \u03b4\u03b5\u03bd \u03c3\u03b7\u03bc\u03b1\u03af\u03bd\u03b5\u03b9 \u03bc\u03b7\u03b4\u03b5\u03bd\u03b9\u03ba\u03cc \u03ba\u03cc\u03c3\u03c4\u03bf\u03c2<\/a><\/li>\n<li><a href=\"#exchangeability-periorismoi\">\u0397 \u03b5\u03b3\u03b3\u03cd\u03b7\u03c3\u03b7 \u03c3\u03c4\u03b1\u03bc\u03b1\u03c4\u03ac \u03b5\u03ba\u03b5\u03af \u03cc\u03c0\u03bf\u03c5 \u03c3\u03c0\u03ac\u03b5\u03b9 \u03b7 exchangeability<\/a><\/li>\n<li><a href=\"#risk-interface-epicheiriseis\">\u0391\u03c0\u03cc \u03c4\u03b7 \u03c1\u03bf\u03bc\u03c0\u03bf\u03c4\u03b9\u03ba\u03ae \u03c3\u03b5 \u03ad\u03bd\u03b1 \u03b3\u03b5\u03bd\u03b9\u03ba\u03cc risk interface<\/a><\/li>\n<li><a href=\"#deployment-checklist\">\u0395\u03c0\u03c4\u03ac \u03ad\u03bb\u03b5\u03b3\u03c7\u03bf\u03b9 \u03c0\u03c1\u03b9\u03bd \u03b1\u03c0\u03cc deployment<\/a><\/li>\n<li><a href=\"#symperasma-sparc\">\u03a4\u03bf \u03c3\u03c5\u03bc\u03c0\u03ad\u03c1\u03b1\u03c3\u03bc\u03b1: \u03b3\u03c1\u03ae\u03b3\u03bf\u03c1\u03b7 \u03c0\u03c1\u03cc\u03b2\u03bb\u03b5\u03c8\u03b7 \u03bc\u03b5 \u03b5\u03b9\u03bb\u03b9\u03ba\u03c1\u03b9\u03bd\u03ae \u03cc\u03c1\u03b9\u03b1<\/a><\/li>\n<\/ul>\n<\/div>\n<p>\u0397 \u03b5\u03c1\u03b3\u03b1\u03c3\u03af\u03b1 <em>SPARC: Single-Pass Scaling for Motion Forecasting with Conformal Bayesian Last Layers<\/em> \u03c4\u03c9\u03bd Sakif Hossain, Julian Teusch \u03ba\u03b1\u03b9 J\u00f6rg P. M\u00fcller \u03b5\u03be\u03b5\u03c4\u03ac\u03b6\u03b5\u03b9 \u03ad\u03bd\u03b1 \u03b4\u03cd\u03c3\u03ba\u03bf\u03bb\u03bf \u03ba\u03b5\u03bd\u03cc \u03c4\u03bf\u03c5 human motion forecasting. \u0388\u03bd\u03b1 \u03bc\u03bf\u03bd\u03c4\u03ad\u03bb\u03bf \u03bc\u03c0\u03bf\u03c1\u03b5\u03af \u03bd\u03b1 \u03b4\u03af\u03bd\u03b5\u03b9 \u03b1\u03bd\u03c4\u03b1\u03b3\u03c9\u03bd\u03b9\u03c3\u03c4\u03b9\u03ba\u03cc \u03bc\u03ad\u03c3\u03bf \u03c3\u03c6\u03ac\u03bb\u03bc\u03b1 \u03ba\u03b1\u03b9 \u03c0\u03b1\u03c1\u2019 \u03cc\u03bb\u03b1 \u03b1\u03c5\u03c4\u03ac \u03bd\u03b1 \u03b5\u03af\u03bd\u03b1\u03b9 \u03b5\u03c0\u03b9\u03ba\u03af\u03bd\u03b4\u03c5\u03bd\u03b1 \u03b2\u03ad\u03b2\u03b1\u03b9\u03bf \u03cc\u03c4\u03b1\u03bd \u03c3\u03c5\u03bd\u03b1\u03bd\u03c4\u03ae\u03c3\u03b5\u03b9 \u03ba\u03af\u03bd\u03b7\u03c3\u03b7 \u03bc\u03b1\u03ba\u03c1\u03b9\u03ac \u03b1\u03c0\u03cc \u03c4\u03bf training support. \u0393\u03b9\u03b1 \u03ad\u03bd\u03b1 cobot, \u03ad\u03bd\u03b1 \u03c3\u03cd\u03c3\u03c4\u03b7\u03bc\u03b1 human-aware navigation \u03ae \u03ad\u03bd\u03b1 collision-avoidance layer, \u03b7 \u03b1\u03b2\u03b5\u03b2\u03b1\u03b9\u03cc\u03c4\u03b7\u03c4\u03b1 \u03c0\u03c1\u03ad\u03c0\u03b5\u03b9 \u03bd\u03b1 \u03c3\u03c5\u03bd\u03b4\u03ad\u03b5\u03c4\u03b1\u03b9 \u03bc\u03b5 \u03c4\u03b7 \u03bc\u03b5\u03bb\u03bb\u03bf\u03bd\u03c4\u03b9\u03ba\u03ae \u03c4\u03c1\u03bf\u03c7\u03b9\u03ac \u03ba\u03b1\u03b9 \u03bc\u03b5 \u03c3\u03c5\u03b3\u03ba\u03b5\u03ba\u03c1\u03b9\u03bc\u03ad\u03bd\u03b7 \u03c0\u03bf\u03bb\u03b9\u03c4\u03b9\u03ba\u03ae \u03b4\u03c1\u03ac\u03c3\u03b7\u03c2.<\/p>\n<h2 id=\"point-forecast-den-arkei\">\u0393\u03b9\u03b1\u03c4\u03af \u03ad\u03bd\u03b1 \u03c3\u03c9\u03c3\u03c4\u03cc point forecast \u03b4\u03b5\u03bd \u03b5\u03af\u03bd\u03b1\u03b9 \u03b1\u03c1\u03ba\u03b5\u03c4\u03cc<\/h2>\n<p>\u03a3\u03c4\u03bf human motion forecasting, \u03c4\u03bf \u03bc\u03bf\u03bd\u03c4\u03ad\u03bb\u03bf \u03c0\u03b1\u03c1\u03b1\u03c4\u03b7\u03c1\u03b5\u03af \u03ad\u03bd\u03b1 \u03c3\u03cd\u03bd\u03c4\u03bf\u03bc\u03bf \u03b9\u03c3\u03c4\u03bf\u03c1\u03b9\u03ba\u03cc \u03b1\u03c0\u03cc 3D \u03c3\u03c4\u03ac\u03c3\u03b5\u03b9\u03c2 \u03ba\u03b1\u03b9 \u03c0\u03c1\u03bf\u03b2\u03bb\u03ad\u03c0\u03b5\u03b9 \u03c4\u03b9\u03c2 \u03b5\u03c0\u03cc\u03bc\u03b5\u03bd\u03b5\u03c2 \u03b8\u03ad\u03c3\u03b5\u03b9\u03c2 \u03c4\u03c9\u03bd \u03b1\u03c1\u03b8\u03c1\u03ce\u03c3\u03b5\u03c9\u03bd. \u03a4\u03bf MPJPE \u03bc\u03b5\u03c4\u03c1\u03ac \u03c4\u03b7 \u03bc\u03ad\u03c3\u03b7 \u03b1\u03c0\u03cc\u03c3\u03c4\u03b1\u03c3\u03b7 \u03c4\u03b7\u03c2 \u03c0\u03c1\u03cc\u03b2\u03bb\u03b5\u03c8\u03b7\u03c2 \u03b1\u03c0\u03cc \u03c4\u03b7\u03bd \u03c0\u03c1\u03b1\u03b3\u03bc\u03b1\u03c4\u03b9\u03ba\u03ae \u03b8\u03ad\u03c3\u03b7. \u0395\u03af\u03bd\u03b1\u03b9 \u03c7\u03c1\u03ae\u03c3\u03b9\u03bc\u03bf metric \u03b1\u03ba\u03c1\u03af\u03b2\u03b5\u03b9\u03b1\u03c2, \u03b1\u03bb\u03bb\u03ac \u03b4\u03b5\u03bd \u03b1\u03c0\u03b1\u03bd\u03c4\u03ac \u03b1\u03bd \u03c4\u03bf \u03c3\u03cd\u03c3\u03c4\u03b7\u03bc\u03b1 \u03b3\u03bd\u03c9\u03c1\u03af\u03b6\u03b5\u03b9 \u03c0\u03cc\u03c4\u03b5 \u03b7 \u03c0\u03c1\u03cc\u03b2\u03bb\u03b5\u03c8\u03ae \u03c4\u03bf\u03c5 \u03b5\u03af\u03bd\u03b1\u03b9 \u03b1\u03c3\u03c4\u03b1\u03b8\u03ae\u03c2 \u03bf\u03cd\u03c4\u03b5 \u03c0\u03cc\u03c3\u03bf \u03bc\u03b5\u03b3\u03ac\u03bb\u03bf safety margin \u03c7\u03c1\u03b5\u03b9\u03ac\u03b6\u03b5\u03c4\u03b1\u03b9 \u03b3\u03cd\u03c1\u03c9 \u03c4\u03b7\u03c2.<\/p>\n<p>\u0397 \u03b4\u03b9\u03b1\u03c6\u03bf\u03c1\u03ac \u03b5\u03af\u03bd\u03b1\u03b9 \u03ba\u03c1\u03af\u03c3\u03b9\u03bc\u03b7 \u03c3\u03b5 shared workspaces. \u039c\u03b9\u03b1 \u03b1\u03c0\u03cc\u03ba\u03bb\u03b9\u03c3\u03b7 \u03bb\u03af\u03b3\u03c9\u03bd \u03b5\u03ba\u03b1\u03c4\u03bf\u03c3\u03c4\u03ce\u03bd \u03c3\u03b5 \u03c0\u03c1\u03bf\u03b2\u03bb\u03ad\u03c8\u03b9\u03bc\u03b7 \u03ba\u03af\u03bd\u03b7\u03c3\u03b7 \u03b4\u03b5\u03bd \u03ad\u03c7\u03b5\u03b9 \u03c4\u03bf \u03af\u03b4\u03b9\u03bf operational risk \u03bc\u03b5 \u03c4\u03b7\u03bd \u03af\u03b4\u03b9\u03b1 \u03b1\u03c0\u03cc\u03ba\u03bb\u03b9\u03c3\u03b7 \u03c3\u03b5 \u03ba\u03af\u03bd\u03b7\u03c3\u03b7 \u03c0\u03bf\u03c5 \u03c4\u03bf \u03bc\u03bf\u03bd\u03c4\u03ad\u03bb\u03bf \u03b4\u03b5\u03bd \u03ad\u03c7\u03b5\u03b9 \u03ba\u03b1\u03bb\u03ac \u03b5\u03ba\u03c0\u03c1\u03bf\u03c3\u03c9\u03c0\u03ae\u03c3\u03b5\u03b9. \u0397 \u03c0\u03c1\u03b1\u03ba\u03c4\u03b9\u03ba\u03ae \u03c3\u03cd\u03bd\u03b4\u03b5\u03c3\u03b7 \u03bc\u03b5 <a href=\"https:\/\/twodots.gr\/automation-robotics-ai-business\/\">\u03c1\u03bf\u03bc\u03c0\u03bf\u03c4\u03b9\u03ba\u03ae \u03ba\u03b1\u03b9 AI \u03c9\u03c2 \u03b5\u03c0\u03b9\u03c7\u03b5\u03b9\u03c1\u03b7\u03c3\u03b9\u03b1\u03ba\u03cc \u03b5\u03c1\u03b3\u03b1\u03bb\u03b5\u03af\u03bf<\/a> \u03b1\u03c1\u03c7\u03af\u03b6\u03b5\u03b9 \u03b1\u03ba\u03c1\u03b9\u03b2\u03ce\u03c2 \u03b5\u03b4\u03ce: \u03c0\u03c1\u03b9\u03bd \u03c3\u03c5\u03bd\u03b4\u03b5\u03b8\u03b5\u03af \u03bc\u03b9\u03b1 \u03c0\u03c1\u03cc\u03b2\u03bb\u03b5\u03c8\u03b7 \u03bc\u03b5 actuator, \u03b5\u03b9\u03b4\u03bf\u03c0\u03bf\u03af\u03b7\u03c3\u03b7 \u03ae fallback, \u03b7 \u03bf\u03bc\u03ac\u03b4\u03b1 \u03c7\u03c1\u03b5\u03b9\u03ac\u03b6\u03b5\u03c4\u03b1\u03b9 \u03ba\u03b1\u03b9 \u03ad\u03bd\u03b1 \u03b1\u03be\u03b9\u03cc\u03c0\u03b9\u03c3\u03c4\u03bf \u03c0\u03b5\u03c1\u03af\u03b3\u03c1\u03b1\u03bc\u03bc\u03b1 \u03b1\u03b2\u03b5\u03b2\u03b1\u03b9\u03cc\u03c4\u03b7\u03c4\u03b1\u03c2.<\/p>\n<aside class=\"td-article-note\">\n<p><strong>\u03a4\u03bf MPJPE \u03b4\u03b5\u03bd \u03b5\u03af\u03bd\u03b1\u03b9 safety certificate.<\/strong> \u03a3\u03c5\u03bd\u03bf\u03c8\u03af\u03b6\u03b5\u03b9 \u03c4\u03bf \u03bc\u03ad\u03c3\u03bf point error, \u03b1\u03bb\u03bb\u03ac \u03b4\u03b5\u03bd \u03b1\u03c0\u03bf\u03c4\u03c5\u03c0\u03ce\u03bd\u03b5\u03b9 \u03bc\u03cc\u03bd\u03bf \u03c4\u03bf\u03c5 \u03c4\u03b7\u03bd \u03ba\u03ac\u03bb\u03c5\u03c8\u03b7 \u03c4\u03c9\u03bd intervals, \u03c4\u03b7 \u03c3\u03c5\u03bc\u03c0\u03b5\u03c1\u03b9\u03c6\u03bf\u03c1\u03ac \u03c5\u03c0\u03cc distribution shift \u03ae \u03c4\u03bf \u03ba\u03cc\u03c3\u03c4\u03bf\u03c2 \u03bc\u03b9\u03b1\u03c2 \u03c5\u03c0\u03b5\u03c1\u03b2\u03bf\u03bb\u03b9\u03ba\u03ac \u03c3\u03c4\u03b5\u03bd\u03ae\u03c2 prediction tube. \u0393\u03b9\u03b1 deployment \u03c7\u03c1\u03b5\u03b9\u03ac\u03b6\u03bf\u03bd\u03c4\u03b1\u03b9 \u03be\u03b5\u03c7\u03c9\u03c1\u03b9\u03c3\u03c4\u03ad\u03c2 \u03bc\u03b5\u03c4\u03c1\u03b9\u03ba\u03ad\u03c2 \u03b3\u03b9\u03b1 accuracy, calibration, width \u03ba\u03b1\u03b9 failures \u03c5\u03c8\u03b7\u03bb\u03bf\u03cd \u03c1\u03af\u03c3\u03ba\u03bf\u03c5.<\/p>\n<\/aside>\n<h2 id=\"dyo-typoi-avevaiotitas\">Aleatoric \u03ba\u03b1\u03b9 epistemic \u03b1\u03b2\u03b5\u03b2\u03b1\u03b9\u03cc\u03c4\u03b7\u03c4\u03b1 \u03b4\u03b5\u03bd \u03b5\u03af\u03bd\u03b1\u03b9 \u03c4\u03bf \u03af\u03b4\u03b9\u03bf<\/h2>\n<p>\u0397 aleatoric \u03b1\u03b2\u03b5\u03b2\u03b1\u03b9\u03cc\u03c4\u03b7\u03c4\u03b1 \u03b1\u03c6\u03bf\u03c1\u03ac \u03c4\u03b7\u03bd \u03b5\u03b3\u03b3\u03b5\u03bd\u03ae \u03b1\u03c3\u03ac\u03c6\u03b5\u03b9\u03b1 \u03c4\u03bf\u03c5 \u03bc\u03ad\u03bb\u03bb\u03bf\u03bd\u03c4\u03bf\u03c2. \u0391\u03c0\u03cc \u03c4\u03bf \u03af\u03b4\u03b9\u03bf observed prefix \u03bc\u03c0\u03bf\u03c1\u03b5\u03af \u03bd\u03b1 \u03b1\u03ba\u03bf\u03bb\u03bf\u03c5\u03b8\u03bf\u03cd\u03bd \u03c0\u03bf\u03bb\u03bb\u03ad\u03c2 \u03b5\u03cd\u03bb\u03bf\u03b3\u03b5\u03c2 \u03ba\u03b9\u03bd\u03ae\u03c3\u03b5\u03b9\u03c2: \u03ad\u03bd\u03b1\u03c2 \u03ac\u03bd\u03b8\u03c1\u03c9\u03c0\u03bf\u03c2 \u03af\u03c3\u03c9\u03c2 \u03c3\u03c5\u03bd\u03b5\u03c7\u03af\u03c3\u03b5\u03b9 \u03b5\u03c5\u03b8\u03b5\u03af\u03b1, \u03c3\u03c4\u03c1\u03af\u03c8\u03b5\u03b9 \u03ae \u03c3\u03c4\u03b1\u03bc\u03b1\u03c4\u03ae\u03c3\u03b5\u03b9. \u0391\u03c5\u03c4\u03ae \u03b7 \u03c0\u03bf\u03bb\u03bb\u03b1\u03c0\u03bb\u03cc\u03c4\u03b7\u03c4\u03b1 \u03b4\u03b5\u03bd \u03b5\u03be\u03b1\u03c6\u03b1\u03bd\u03af\u03b6\u03b5\u03c4\u03b1\u03b9 \u03b1\u03ba\u03cc\u03bc\u03b7 \u03ba\u03b1\u03b9 \u03bc\u03b5 \u03c0\u03b5\u03c1\u03b9\u03c3\u03c3\u03cc\u03c4\u03b5\u03c1\u03b1 \u03b4\u03b5\u03b4\u03bf\u03bc\u03ad\u03bd\u03b1. \u03a4\u03bf \u03bc\u03bf\u03bd\u03c4\u03ad\u03bb\u03bf \u03c0\u03c1\u03ad\u03c0\u03b5\u03b9 \u03bd\u03b1 \u03b4\u03b9\u03b1\u03c4\u03b7\u03c1\u03b5\u03af \u03c4\u03b9\u03c2 \u03c7\u03c1\u03bf\u03bd\u03b9\u03ba\u03ad\u03c2 \u03ba\u03b1\u03b9 \u03c3\u03ba\u03b5\u03bb\u03b5\u03c4\u03b9\u03ba\u03ad\u03c2 \u03c3\u03c5\u03c3\u03c7\u03b5\u03c4\u03af\u03c3\u03b5\u03b9\u03c2, \u03b5\u03c0\u03b5\u03b9\u03b4\u03ae \u03bf\u03b9 \u03b1\u03c1\u03b8\u03c1\u03ce\u03c3\u03b5\u03b9\u03c2 \u03b4\u03b5\u03bd \u03ba\u03b9\u03bd\u03bf\u03cd\u03bd\u03c4\u03b1\u03b9 \u03c9\u03c2 \u03b1\u03bd\u03b5\u03be\u03ac\u03c1\u03c4\u03b7\u03c4\u03bf\u03b9 \u03b1\u03c1\u03b9\u03b8\u03bc\u03bf\u03af.<\/p>\n<p>\u0397 epistemic \u03b1\u03b2\u03b5\u03b2\u03b1\u03b9\u03cc\u03c4\u03b7\u03c4\u03b1 \u03b1\u03c6\u03bf\u03c1\u03ac \u03cc\u03c3\u03b1 \u03c4\u03bf \u03bc\u03bf\u03bd\u03c4\u03ad\u03bb\u03bf \u03b4\u03b5\u03bd \u03ad\u03c7\u03b5\u03b9 \u03bc\u03ac\u03b8\u03b5\u03b9 \u03ba\u03b1\u03bb\u03ac. \u0391\u03c5\u03be\u03ac\u03bd\u03b5\u03c4\u03b1\u03b9 \u03cc\u03c4\u03b1\u03bd \u03c4\u03b1 learned features \u03bc\u03b9\u03b1\u03c2 \u03b5\u03b9\u03c3\u03cc\u03b4\u03bf\u03c5 \u03b1\u03c0\u03ad\u03c7\u03bf\u03c5\u03bd \u03b1\u03c0\u03cc \u03c4\u03b7 \u03c3\u03c4\u03ae\u03c1\u03b9\u03be\u03b7 \u03c4\u03c9\u03bd training data. \u0397 \u03b4\u03b9\u03ac\u03ba\u03c1\u03b9\u03c3\u03b7 \u03b8\u03c5\u03bc\u03af\u03b6\u03b5\u03b9 \u03c4\u03bf \u03b3\u03b9\u03b1\u03c4\u03af \u03ad\u03bd\u03b1 <a href=\"https:\/\/twodots.gr\/transition-complexity-profile-ai-world-models\/\">benchmark \u03c7\u03c1\u03b5\u03b9\u03ac\u03b6\u03b5\u03c4\u03b1\u03b9 \u03bd\u03b1 \u03bc\u03b5\u03c4\u03c1\u03ac \u03c4\u03b7 \u03b4\u03c5\u03c3\u03ba\u03bf\u03bb\u03af\u03b1 \u03c4\u03bf\u03c5 \u03c0\u03b5\u03c1\u03b9\u03b2\u03ac\u03bb\u03bb\u03bf\u03bd\u03c4\u03bf\u03c2<\/a> \u03ba\u03b1\u03b9 \u03cc\u03c7\u03b9 \u03bc\u03cc\u03bd\u03bf \u03c4\u03bf \u03c4\u03b5\u03bb\u03b9\u03ba\u03cc success rate: \u03b4\u03cd\u03bf \u03c0\u03b1\u03c1\u03cc\u03bc\u03bf\u03b9\u03b1 outputs \u03bc\u03c0\u03bf\u03c1\u03b5\u03af \u03bd\u03b1 \u03c0\u03c1\u03bf\u03ad\u03c1\u03c7\u03bf\u03bd\u03c4\u03b1\u03b9 \u03b1\u03c0\u03cc \u03b5\u03bd\u03c4\u03b5\u03bb\u03ce\u03c2 \u03b4\u03b9\u03b1\u03c6\u03bf\u03c1\u03b5\u03c4\u03b9\u03ba\u03ae \u03b2\u03b5\u03b2\u03b1\u03b9\u03cc\u03c4\u03b7\u03c4\u03b1 \u03b3\u03bd\u03ce\u03c3\u03b7\u03c2.<\/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\">Aleatoric \u03b1\u03b2\u03b5\u03b2\u03b1\u03b9\u03cc\u03c4\u03b7\u03c4\u03b1<\/p>\n<p>\u03a0\u03b5\u03c1\u03b9\u03b3\u03c1\u03ac\u03c6\u03b5\u03b9 \u03c4\u03b9\u03c2 \u03c0\u03bf\u03bb\u03bb\u03ad\u03c2 \u03b5\u03cd\u03bb\u03bf\u03b3\u03b5\u03c2 \u03bc\u03b5\u03bb\u03bb\u03bf\u03bd\u03c4\u03b9\u03ba\u03ad\u03c2 \u03ba\u03b9\u03bd\u03ae\u03c3\u03b5\u03b9\u03c2 \u03ba\u03b1\u03b9 \u03c4\u03b7 \u03b4\u03bf\u03bc\u03b7\u03bc\u03ad\u03bd\u03b7 \u03bc\u03b5\u03c4\u03b1\u03b2\u03bb\u03b7\u03c4\u03cc\u03c4\u03b7\u03c4\u03b1 \u03bc\u03b5\u03c4\u03b1\u03be\u03cd \u03c7\u03c1\u03cc\u03bd\u03bf\u03c5, \u03b1\u03c1\u03b8\u03c1\u03ce\u03c3\u03b5\u03c9\u03bd \u03ba\u03b1\u03b9 \u03c3\u03c5\u03bd\u03c4\u03b5\u03c4\u03b1\u03b3\u03bc\u03ad\u03bd\u03c9\u03bd.<\/p>\n<div class=\"td-badge-row\"><span class=\"td-badge\">\u0391\u03c3\u03ac\u03c6\u03b5\u03b9\u03b1 \u03b4\u03b5\u03b4\u03bf\u03bc\u03ad\u03bd\u03c9\u03bd<\/span><span class=\"td-badge\">Covariance shape<\/span><\/div>\n<\/div>\n<div class=\"td-platform-card td-platform-card--navy\">\n<p class=\"td-comparison-title\">Epistemic \u03b1\u03b2\u03b5\u03b2\u03b1\u03b9\u03cc\u03c4\u03b7\u03c4\u03b1<\/p>\n<p>\u03a3\u03c4\u03bf SPARC \u03b5\u03ba\u03c6\u03c1\u03ac\u03b6\u03b5\u03c4\u03b1\u03b9 \u03c9\u03c2 feature-space leverage \u03ba \u03c0\u03bf\u03c5 \u03b4\u03b9\u03bf\u03b3\u03ba\u03ce\u03bd\u03b5\u03b9 \u03c4\u03bf \u03bc\u03ad\u03b3\u03b5\u03b8\u03bf\u03c2 \u03c4\u03b7\u03c2 covariance \u03cc\u03c4\u03b1\u03bd \u03b7 \u03b5\u03af\u03c3\u03bf\u03b4\u03bf\u03c2 \u03ad\u03c7\u03b5\u03b9 \u03b1\u03c3\u03b8\u03b5\u03bd\u03ad\u03c3\u03c4\u03b5\u03c1\u03b7 training support.<\/p>\n<div class=\"td-badge-row\"><span class=\"td-badge\">Model knowledge<\/span><span class=\"td-badge\">Scale inflation<\/span><\/div>\n<\/div>\n<div class=\"td-platform-card\">\n<p class=\"td-comparison-title\">Conformal calibration<\/p>\n<p>\u039c\u03b5\u03c4\u03b1\u03c4\u03c1\u03ad\u03c0\u03b5\u03b9 \u03c4\u03b1 standardized residuals \u03b5\u03bd\u03cc\u03c2 held-out calibration set \u03c3\u03b5 quantiles \u03b3\u03b9\u03b1 95% marginal prediction tubes \u03c5\u03c0\u03cc exchangeability.<\/p>\n<div class=\"td-badge-row\"><span class=\"td-badge\">Coverage target<\/span><span class=\"td-badge\">Post-hoc layer<\/span><\/div>\n<\/div>\n<\/div>\n<\/div>\n<h2 id=\"pos-leitourgei-sparc\">\u03a0\u03ce\u03c2 \u03bb\u03b5\u03b9\u03c4\u03bf\u03c5\u03c1\u03b3\u03b5\u03af \u03c4\u03bf SPARC \u03c3\u03b5 \u03c4\u03c1\u03af\u03b1 \u03b5\u03c0\u03af\u03c0\u03b5\u03b4\u03b1<\/h2>\n<p>\u03a0\u03c1\u03ce\u03c4\u03b1, \u03ad\u03bd\u03b1 siMLPe\/DCT-style MLP \u03c0\u03c1\u03bf\u03b2\u03bb\u03ad\u03c0\u03b5\u03b9 \u03c4\u03b7 \u03bc\u03ad\u03c3\u03b7 \u03bc\u03b5\u03bb\u03bb\u03bf\u03bd\u03c4\u03b9\u03ba\u03ae \u03ba\u03af\u03bd\u03b7\u03c3\u03b7. \u039f\u03b9 \u03b5\u03af\u03c3\u03bf\u03b4\u03bf\u03b9 \u03bc\u03b5\u03c4\u03b1\u03c3\u03c7\u03b7\u03bc\u03b1\u03c4\u03af\u03b6\u03bf\u03bd\u03c4\u03b1\u03b9 \u03c3\u03b5 \u03c3\u03c5\u03bd\u03c4\u03b5\u03bb\u03b5\u03c3\u03c4\u03ad\u03c2 discrete cosine transform \u03ba\u03b1\u03b9 \u03b5\u03c0\u03b9\u03c3\u03c4\u03c1\u03ad\u03c6\u03bf\u03c5\u03bd \u03c3\u03c4\u03bf\u03bd \u03c7\u03c1\u03cc\u03bd\u03bf \u03bc\u03ad\u03c3\u03c9 inverse DCT. \u039f feature extractor \u03c0\u03b1\u03c1\u03b1\u03bc\u03ad\u03bd\u03b5\u03b9 deterministic, \u03ce\u03c3\u03c4\u03b5 \u03b7 \u03b2\u03b1\u03c3\u03b9\u03ba\u03ae \u03c0\u03c1\u03cc\u03b2\u03bb\u03b5\u03c8\u03b7 \u03bd\u03b1 \u03bc\u03b7\u03bd \u03b1\u03c0\u03b1\u03b9\u03c4\u03b5\u03af ensemble \u03ae \u03b5\u03c0\u03b1\u03bd\u03b1\u03bb\u03b1\u03bc\u03b2\u03b1\u03bd\u03cc\u03bc\u03b5\u03bd\u03b1 stochastic passes.<\/p>\n<p>\u0394\u03b5\u03cd\u03c4\u03b5\u03c1\u03bf\u03bd, \u03ad\u03bd\u03b1 Gaussian head \u03bc\u03bf\u03bd\u03c4\u03b5\u03bb\u03bf\u03c0\u03bf\u03b9\u03b5\u03af \u03c4\u03b7 \u03b4\u03bf\u03bc\u03b7\u03bc\u03ad\u03bd\u03b7 aleatoric covariance. \u0397 matrix-normal \u03c0\u03b1\u03c1\u03ac\u03bc\u03b5\u03c4\u03c1\u03bf\u03c2 \u03b1\u03c0\u03bf\u03c4\u03c5\u03c0\u03ce\u03bd\u03b5\u03b9 temporal correlations, \u03b5\u03bd\u03ce \u03bc\u03b9\u03b1 graph\/GMRF \u03ba\u03b1\u03c4\u03b1\u03c3\u03ba\u03b5\u03c5\u03ae \u03c3\u03c5\u03bd\u03b4\u03ad\u03b5\u03b9 \u03c4\u03b9\u03c2 \u03b1\u03c1\u03b8\u03c1\u03ce\u03c3\u03b5\u03b9\u03c2 \u03bc\u03ad\u03c3\u03b1 \u03b1\u03c0\u03cc sparse precision. \u03a4\u03c1\u03af\u03c4\u03bf\u03bd, \u03bc\u03b9\u03b1 conjugate Bayesian last layer \u03c5\u03c0\u03bf\u03bb\u03bf\u03b3\u03af\u03b6\u03b5\u03b9 \u03b1\u03bd\u03b1\u03bb\u03c5\u03c4\u03b9\u03ba\u03ac horizon-wise \u03ba \u03ba\u03b1\u03b9 \u03b4\u03b9\u03bf\u03b3\u03ba\u03ce\u03bd\u03b5\u03b9 \u03c4\u03b7\u03bd covariance \u03c7\u03c9\u03c1\u03af\u03c2 \u03bd\u03b1 \u03b1\u03bb\u03bb\u03ac\u03b6\u03b5\u03b9 \u03c4\u03bf correlation pattern \u03c4\u03b7\u03c2. \u03a3\u03c4\u03bf \u03c4\u03ad\u03bb\u03bf\u03c2, \u03c4\u03bf split conformal \u03b2\u03ae\u03bc\u03b1 \u03b2\u03b1\u03b8\u03bc\u03bf\u03bd\u03bf\u03bc\u03b5\u03af \u03c4\u03b1 prediction tubes.<\/p>\n<div class=\"td-decision-band\">\n<p class=\"td-decision-band-kicker\">\u0397 \u03b1\u03c1\u03c7\u03b9\u03c4\u03b5\u03ba\u03c4\u03bf\u03bd\u03b9\u03ba\u03ae \u03b9\u03b4\u03ad\u03b1<\/p>\n<h3>\u0386\u03bb\u03bb\u03bf \u03c4\u03bf \u03c3\u03c7\u03ae\u03bc\u03b1 \u03c4\u03b7\u03c2 \u03b1\u03b2\u03b5\u03b2\u03b1\u03b9\u03cc\u03c4\u03b7\u03c4\u03b1\u03c2 \u03ba\u03b1\u03b9 \u03ac\u03bb\u03bb\u03bf \u03c4\u03bf \u03bc\u03ad\u03b3\u03b5\u03b8\u03cc\u03c2 \u03c4\u03b7\u03c2<\/h3>\n<p>\u0397 structured covariance \u03c0\u03b5\u03c1\u03b9\u03b3\u03c1\u03ac\u03c6\u03b5\u03b9 \u03c0\u03ce\u03c2 \u03c3\u03c5\u03bd\u03b4\u03ad\u03bf\u03bd\u03c4\u03b1\u03b9 \u03c7\u03c1\u03cc\u03bd\u03bf\u03c2 \u03ba\u03b1\u03b9 \u03c3\u03ba\u03b5\u03bb\u03b5\u03c4\u03cc\u03c2. \u03a4\u03bf \u03ba \u03c0\u03b5\u03c1\u03b9\u03b3\u03c1\u03ac\u03c6\u03b5\u03b9 \u03c0\u03cc\u03c3\u03bf \u03c0\u03c1\u03ad\u03c0\u03b5\u03b9 \u03bd\u03b1 \u03b4\u03b9\u03bf\u03b3\u03ba\u03c9\u03b8\u03b5\u03af \u03b1\u03c5\u03c4\u03ae \u03b7 \u03b4\u03bf\u03bc\u03ae \u03b3\u03b9\u03b1 \u03bc\u03b9\u03b1 \u03c3\u03c5\u03b3\u03ba\u03b5\u03ba\u03c1\u03b9\u03bc\u03ad\u03bd\u03b7 \u03b5\u03af\u03c3\u03bf\u03b4\u03bf. \u03a4\u03bf conformal quantile \u03b5\u03bb\u03ad\u03b3\u03c7\u03b5\u03b9 \u03c4\u03b7\u03bd \u03b5\u03bc\u03c0\u03b5\u03b9\u03c1\u03b9\u03ba\u03ae \u03ba\u03ac\u03bb\u03c5\u03c8\u03b7. \u0391\u03bd \u03bc\u03b9\u03b1 \u03c5\u03bb\u03bf\u03c0\u03bf\u03af\u03b7\u03c3\u03b7 \u03c4\u03b1 \u03c3\u03c5\u03bc\u03c0\u03c4\u03cd\u03be\u03b5\u03b9 \u03c3\u03b5 \u03ad\u03bd\u03b1 \u03b1\u03b4\u03b9\u03b1\u03c6\u03b1\u03bd\u03ad\u03c2 confidence score, \u03c7\u03ac\u03bd\u03b5\u03b9 \u03c4\u03bf \u03b2\u03b1\u03c3\u03b9\u03ba\u03cc deployment interface \u03c4\u03bf\u03c5 SPARC.<\/p>\n<\/div>\n<h2 id=\"ti-metrai-kappa\">\u03a4\u03b9 \u03bc\u03b5\u03c4\u03c1\u03ac \u03c0\u03c1\u03b1\u03b3\u03bc\u03b1\u03c4\u03b9\u03ba\u03ac \u03c4\u03bf \u03ba<\/h2>\n<p>\u03a4\u03bf \u03ba \u03c5\u03c0\u03bf\u03bb\u03bf\u03b3\u03af\u03b6\u03b5\u03c4\u03b1\u03b9 \u03b1\u03c0\u03cc quadratic form \u03c0\u03ac\u03bd\u03c9 \u03c3\u03c4\u03b1 learned features \u03ba\u03b1\u03b9 \u03c3\u03b5 inverse sufficient-statistics matrix \u03c0\u03bf\u03c5 \u03ad\u03c7\u03b5\u03b9 \u03c0\u03c1\u03bf\u03c3\u03b1\u03c1\u03bc\u03bf\u03c3\u03c4\u03b5\u03af \u03bc\u03af\u03b1 \u03c6\u03bf\u03c1\u03ac offline. \u039c\u03b5 \u03b1\u03c0\u03bb\u03ac \u03bb\u03cc\u03b3\u03b9\u03b1, \u03b1\u03c5\u03be\u03ac\u03bd\u03b5\u03c4\u03b1\u03b9 \u03cc\u03c4\u03b1\u03bd \u03c4\u03bf feature vector \u03ad\u03c7\u03b5\u03b9 \u03bc\u03b5\u03b3\u03ac\u03bb\u03b7 leverage \u03c3\u03b5 \u03c3\u03c7\u03ad\u03c3\u03b7 \u03bc\u03b5 \u03c4\u03bf training design. \u03a3\u03c4\u03bf inference, \u03c4\u03bf \u03c3\u03cd\u03c3\u03c4\u03b7\u03bc\u03b1 \u03c7\u03c1\u03b5\u03b9\u03ac\u03b6\u03b5\u03c4\u03b1\u03b9 \u03c4\u03bf forward pass \u03c4\u03bf\u03c5 \u03b4\u03b9\u03ba\u03c4\u03cd\u03bf\u03c5 \u03ba\u03b1\u03b9 \u03ba\u03bb\u03b5\u03b9\u03c3\u03c4\u03bf\u03cd \u03c4\u03cd\u03c0\u03bf\u03c5 horizon-wise \u03c5\u03c0\u03bf\u03bb\u03bf\u03b3\u03b9\u03c3\u03bc\u03bf\u03cd\u03c2, \u03cc\u03c7\u03b9 posterior sampling.<\/p>\n<p>\u0391\u03c5\u03c4\u03cc \u03b4\u03b5\u03bd \u03bc\u03b5\u03c4\u03b1\u03c4\u03c1\u03ad\u03c0\u03b5\u03b9 \u03c4\u03bf \u03ba \u03c3\u03b5 \u03ba\u03b1\u03b8\u03bf\u03bb\u03b9\u03ba\u03cc out-of-distribution detector. \u0397 \u03c0\u03bf\u03b9\u03cc\u03c4\u03b7\u03c4\u03ac \u03c4\u03bf\u03c5 \u03b5\u03be\u03b1\u03c1\u03c4\u03ac\u03c4\u03b1\u03b9 \u03b1\u03c0\u03cc \u03c4\u03b7 learned representation: \u03b1\u03bd \u03c4\u03b1 features \u03b4\u03b5\u03bd \u03b1\u03c0\u03bf\u03c4\u03c5\u03c0\u03ce\u03bd\u03bf\u03c5\u03bd \u03c4\u03b9 \u03b5\u03af\u03bd\u03b1\u03b9 \u03c0\u03c1\u03b1\u03b3\u03bc\u03b1\u03c4\u03b9\u03ba\u03ac \u00ab\u03ac\u03b3\u03bd\u03c9\u03c3\u03c4\u03bf\u00bb, \u03c4\u03bf signal \u03bc\u03c0\u03bf\u03c1\u03b5\u03af \u03bd\u03b1 \u03b5\u03af\u03bd\u03b1\u03b9 \u03b1\u03b4\u03cd\u03bd\u03b1\u03bc\u03bf. \u03a4\u03bf paper \u03c7\u03c1\u03b7\u03c3\u03b9\u03bc\u03bf\u03c0\u03bf\u03b9\u03b5\u03af \u03b5\u03c0\u03af\u03c3\u03b7\u03c2 timewise shrinkage \u03c0\u03c1\u03bf\u03c2 global sequence-level scale \u03ba\u03b1\u03b9 monotone temperature \u03b3\u03b9\u03b1 \u03c4\u03bf trade-off NLL \u03ba\u03b1\u03b9 interval width. \u039f\u03b9 \u03c0\u03b1\u03c1\u03ac\u03bc\u03b5\u03c4\u03c1\u03bf\u03b9 \u03b5\u03c0\u03b9\u03bb\u03ad\u03b3\u03bf\u03bd\u03c4\u03b1\u03b9 \u03c3\u03b5 held-out training windows \u03ba\u03b1\u03b9 \u03c3\u03c4\u03b1\u03b8\u03b5\u03c1\u03bf\u03c0\u03bf\u03b9\u03bf\u03cd\u03bd\u03c4\u03b1\u03b9 \u03c0\u03c1\u03b9\u03bd \u03b1\u03c0\u03cc calibration \u03ba\u03b1\u03b9 evaluation.<\/p>\n<h2 id=\"prediction-tubes\">Prediction tubes \u03b1\u03bd\u03c4\u03af \u03b3\u03b9\u03b1 \u03b1\u03c3\u03b1\u03c6\u03ad\u03c2 confidence score<\/h2>\n<p>\u0397 \u03ba\u03cd\u03c1\u03b9\u03b1 \u03ad\u03be\u03bf\u03b4\u03bf\u03c2 \u03b4\u03b5\u03bd \u03b5\u03af\u03bd\u03b1\u03b9 \u03ad\u03bd\u03b1\u03c2 \u03bc\u03bf\u03bd\u03b1\u03b4\u03b9\u03ba\u03cc\u03c2 \u03b1\u03c1\u03b9\u03b8\u03bc\u03cc\u03c2 confidence. \u0393\u03b9\u03b1 \u03ba\u03ac\u03b8\u03b5 \u03bc\u03b5\u03bb\u03bb\u03bf\u03bd\u03c4\u03b9\u03ba\u03cc \u03c7\u03c1\u03cc\u03bd\u03bf, \u03ac\u03c1\u03b8\u03c1\u03c9\u03c3\u03b7 \u03ba\u03b1\u03b9 3D \u03c3\u03c5\u03bd\u03c4\u03b5\u03c4\u03b1\u03b3\u03bc\u03ad\u03bd\u03b7, \u03c4\u03bf SPARC \u03c0\u03b1\u03c1\u03ac\u03b3\u03b5\u03b9 interval \u03b3\u03cd\u03c1\u03c9 \u03b1\u03c0\u03cc \u03c4\u03b7\u03bd \u03c0\u03c1\u03bf\u03b2\u03bb\u03b5\u03c0\u03cc\u03bc\u03b5\u03bd\u03b7 \u03bc\u03ad\u03c3\u03b7 \u03b8\u03ad\u03c3\u03b7. \u0397 \u03c3\u03c5\u03bb\u03bb\u03bf\u03b3\u03ae \u03b1\u03c5\u03c4\u03ce\u03bd \u03c4\u03c9\u03bd axis-aligned intervals \u03c3\u03c7\u03b7\u03bc\u03b1\u03c4\u03af\u03b6\u03b5\u03b9 prediction tube. \u03a4\u03bf \u03c3\u03cd\u03c3\u03c4\u03b7\u03bc\u03b1 \u03bc\u03c0\u03bf\u03c1\u03b5\u03af \u03ad\u03c4\u03c3\u03b9 \u03bd\u03b1 \u03c3\u03c5\u03bd\u03b4\u03ad\u03c3\u03b5\u03b9 \u03ad\u03bd\u03b1 \u03c3\u03c5\u03b3\u03ba\u03b5\u03ba\u03c1\u03b9\u03bc\u03ad\u03bd\u03bf envelope \u03bc\u03b5 collision checks, personal-space constraints \u03ae conservative fallback.<\/p>\n<p>\u039f \u03ba\u03cd\u03c1\u03b9\u03bf\u03c2 \u03c3\u03c4\u03cc\u03c7\u03bf\u03c2 \u03b5\u03af\u03bd\u03b1\u03b9 95% marginal coverage \u03b3\u03b9\u03b1 \u03c4\u03bf induced score distribution, \u03cc\u03c7\u03b9 \u03b5\u03b3\u03b3\u03cd\u03b7\u03c3\u03b7 \u03cc\u03c4\u03b9 \u03bf\u03bb\u03cc\u03ba\u03bb\u03b7\u03c1\u03b7 \u03b7 \u03c0\u03bb\u03ae\u03c1\u03b7\u03c2 \u03c4\u03c1\u03bf\u03c7\u03b9\u03ac \u03b8\u03b1 \u03b2\u03c1\u03af\u03c3\u03ba\u03b5\u03c4\u03b1\u03b9 \u03c4\u03b1\u03c5\u03c4\u03cc\u03c7\u03c1\u03bf\u03bd\u03b1 \u03bc\u03ad\u03c3\u03b1 \u03c3\u03c4\u03bf \u03c3\u03cd\u03bd\u03bf\u03bb\u03bf. \u03a4\u03bf split conformal calibration \u03c7\u03c1\u03b7\u03c3\u03b9\u03bc\u03bf\u03c0\u03bf\u03b9\u03b5\u03af standardized residuals \u03ba\u03b1\u03b9 quantiles \u03b1\u03bd\u03ac horizon \u03ba\u03b1\u03b9 \u03ac\u03c1\u03b8\u03c1\u03c9\u03c3\u03b7, \u03bc\u03b5 pooling \u03c3\u03c4\u03b9\u03c2 \u03c4\u03c1\u03b5\u03b9\u03c2 coordinates. \u03a4\u03bf paper \u03b5\u03be\u03b5\u03c4\u03ac\u03b6\u03b5\u03b9 \u03b5\u03c0\u03b9\u03c0\u03bb\u03ad\u03bf\u03bd Mondrian tubes \u03b1\u03bd\u03ac \u03ba-bin \u03ba\u03b1\u03b9 trajectory-level ellipsoids, \u03b1\u03bb\u03bb\u03ac \u03b1\u03c5\u03c4\u03ac \u03b5\u03af\u03bd\u03b1\u03b9 \u03b4\u03b9\u03b1\u03c6\u03bf\u03c1\u03b5\u03c4\u03b9\u03ba\u03ad\u03c2 \u03ad\u03bd\u03bd\u03bf\u03b9\u03b5\u03c2 \u03ba\u03ac\u03bb\u03c5\u03c8\u03b7\u03c2 \u03ba\u03b1\u03b9 \u03b4\u03b5\u03bd \u03c0\u03c1\u03ad\u03c0\u03b5\u03b9 \u03bd\u03b1 \u03c0\u03b1\u03c1\u03bf\u03c5\u03c3\u03b9\u03ac\u03b6\u03bf\u03bd\u03c4\u03b1\u03b9 \u03c9\u03c2 \u03c4\u03bf \u03b2\u03b1\u03c3\u03b9\u03ba\u03cc \u03b1\u03c0\u03bf\u03c4\u03ad\u03bb\u03b5\u03c3\u03bc\u03b1.<\/p>\n<h2 id=\"peiramatiko-protokollo\">\u03a4\u03bf \u03c0\u03b5\u03b9\u03c1\u03b1\u03bc\u03b1\u03c4\u03b9\u03ba\u03cc \u03c0\u03c1\u03c9\u03c4\u03cc\u03ba\u03bf\u03bb\u03bb\u03bf \u03ba\u03b1\u03b9 \u03c4\u03b1 \u03b4\u03b5\u03b4\u03bf\u03bc\u03ad\u03bd\u03b1<\/h2>\n<p>\u03a4\u03bf Human3.6M \u03b5\u03af\u03bd\u03b1\u03b9 \u03c4\u03bf \u03ba\u03cd\u03c1\u03b9\u03bf benchmark. \u0397 \u03b1\u03be\u03b9\u03bf\u03bb\u03cc\u03b3\u03b7\u03c3\u03b7 \u03b5\u03c0\u03b5\u03ba\u03c4\u03b5\u03af\u03bd\u03b5\u03c4\u03b1\u03b9 \u03c3\u03b5 AMASS, LaFAN1, CMU-MoCap, 3DPW \u03ba\u03b1\u03b9 \u03c3\u03c4\u03b1 HRI-style datasets CHICO, HA4M \u03ba\u03b1\u03b9 AnDy, \u03b4\u03b7\u03bc\u03b9\u03bf\u03c5\u03c1\u03b3\u03ce\u03bd\u03c4\u03b1\u03c2 \u03b5\u03bd\u03bd\u03ad\u03b1 dataset\/protocol blocks. \u03a3\u03c4\u03b1 \u03bc\u03b7 Human3.6M \u03b4\u03b5\u03b4\u03bf\u03bc\u03ad\u03bd\u03b1, \u03b5\u03ba\u03c4\u03cc\u03c2 \u03b1\u03bd \u03b4\u03b7\u03bb\u03ce\u03bd\u03b5\u03c4\u03b1\u03b9 \u03b4\u03b9\u03b1\u03c6\u03bf\u03c1\u03b5\u03c4\u03b9\u03ba\u03ac, \u03c4\u03bf \u03bc\u03bf\u03bd\u03c4\u03ad\u03bb\u03bf \u03c0\u03b1\u03c1\u03b1\u03c4\u03b7\u03c1\u03b5\u03af 50 frames \u03ba\u03b1\u03b9 \u03c0\u03c1\u03bf\u03b2\u03bb\u03ad\u03c0\u03b5\u03b9 25.<\/p>\n<p>\u03a4\u03bf evaluation protocol \u03c7\u03c1\u03b7\u03c3\u03b9\u03bc\u03bf\u03c0\u03bf\u03b9\u03b5\u03af deterministic permutation \u03bc\u03b5 seed 304. \u03a4\u03b1 \u03c0\u03c1\u03ce\u03c4\u03b1 512 windows \u03c7\u03c1\u03b7\u03c3\u03b9\u03bc\u03bf\u03c0\u03bf\u03b9\u03bf\u03cd\u03bd\u03c4\u03b1\u03b9 \u03bc\u03cc\u03bd\u03bf \u03b3\u03b9\u03b1 split conformal calibration \u03ba\u03b1\u03b9 \u03c4\u03b1 \u03b5\u03c0\u03cc\u03bc\u03b5\u03bd\u03b1 1.024 \u03b1\u03c0\u03bf\u03ba\u03bb\u03b5\u03b9\u03c3\u03c4\u03b9\u03ba\u03ac \u03b3\u03b9\u03b1 metrics, \u03bc\u03b5 \u03c3\u03c4\u03cc\u03c7\u03bf \u03b1=0,05. \u03a4\u03bf training split \u03c0\u03c1\u03bf\u03c3\u03b1\u03c1\u03bc\u03cc\u03b6\u03b5\u03b9 backbone, covariance heads \u03ba\u03b1\u03b9 conjugate statistics. Held-out training windows \u03b5\u03c0\u03b9\u03bb\u03ad\u03b3\u03bf\u03c5\u03bd \u03c0\u03ad\u03bd\u03c4\u03b5 uncertainty\/calibration hyperparameters, \u03b5\u03bd\u03ce \u03c4\u03bf disjoint evaluation set \u03c7\u03c1\u03b7\u03c3\u03b9\u03bc\u03bf\u03c0\u03bf\u03b9\u03b5\u03af\u03c4\u03b1\u03b9 \u03bc\u03af\u03b1 \u03c6\u03bf\u03c1\u03ac.<\/p>\n<aside class=\"td-article-note\">\n<p><strong>\u0397 \u03b4\u03b9\u03b1\u03af\u03c1\u03b5\u03c3\u03b7 \u03c4\u03c9\u03bd \u03b4\u03b5\u03b4\u03bf\u03bc\u03ad\u03bd\u03c9\u03bd \u03b5\u03af\u03bd\u03b1\u03b9 \u03bc\u03ad\u03c1\u03bf\u03c2 \u03c4\u03bf\u03c5 \u03b1\u03c0\u03bf\u03c4\u03b5\u03bb\u03ad\u03c3\u03bc\u03b1\u03c4\u03bf\u03c2.<\/strong> \u0391\u03bd \u03c4\u03bf calibration set \u03b5\u03c0\u03b7\u03c1\u03b5\u03ac\u03c3\u03b5\u03b9 weights \u03ae hyperparameters, \u03ae \u03b1\u03bd \u03c4\u03b1 evaluation windows \u03c7\u03c1\u03b7\u03c3\u03b9\u03bc\u03bf\u03c0\u03bf\u03b9\u03b7\u03b8\u03bf\u03cd\u03bd \u03b5\u03c0\u03b1\u03bd\u03b5\u03b9\u03bb\u03b7\u03bc\u03bc\u03ad\u03bd\u03b1 \u03b3\u03b9\u03b1 tuning, \u03b7 \u03b1\u03bd\u03b1\u03c6\u03b5\u03c1\u03cc\u03bc\u03b5\u03bd\u03b7 \u03ba\u03ac\u03bb\u03c5\u03c8\u03b7 \u03b4\u03b5\u03bd \u03b1\u03bd\u03c4\u03b9\u03c3\u03c4\u03bf\u03b9\u03c7\u03b5\u03af \u03c0\u03b9\u03b1 \u03c3\u03c4\u03bf \u03c0\u03c1\u03c9\u03c4\u03cc\u03ba\u03bf\u03bb\u03bb\u03bf \u03c4\u03b7\u03c2 \u03b5\u03c1\u03b3\u03b1\u03c3\u03af\u03b1\u03c2. \u03a3\u03c4\u03b7\u03bd \u03c0\u03b1\u03c1\u03b1\u03b3\u03c9\u03b3\u03ae \u03c7\u03c1\u03b5\u03b9\u03ac\u03b6\u03b5\u03c4\u03b1\u03b9 \u03b1\u03bd\u03c4\u03af\u03c3\u03c4\u03bf\u03b9\u03c7\u03b7 \u03c0\u03b5\u03b9\u03b8\u03b1\u03c1\u03c7\u03af\u03b1 \u03b5\u03ba\u03b4\u03cc\u03c3\u03b5\u03c9\u03bd \u03ba\u03b1\u03b9 \u03c7\u03c1\u03bf\u03bd\u03b9\u03ba\u03ce\u03bd \u03bf\u03c1\u03af\u03c9\u03bd.<\/p>\n<\/aside>\n<h2 id=\"vasika-apotelesmata\">\u03a4\u03b1 \u03b2\u03b1\u03c3\u03b9\u03ba\u03ac \u03b1\u03c0\u03bf\u03c4\u03b5\u03bb\u03ad\u03c3\u03bc\u03b1\u03c4\u03b1 \u03c7\u03c9\u03c1\u03af\u03c2 \u03c5\u03c0\u03b5\u03c1\u03b2\u03bf\u03bb\u03ad\u03c2<\/h2>\n<p>\u0395\u03c0\u03b5\u03b9\u03b4\u03ae \u03bf\u03b9 \u03bc\u03bf\u03bd\u03ac\u03b4\u03b5\u03c2 \u03b4\u03b9\u03b1\u03c6\u03ad\u03c1\u03bf\u03c5\u03bd \u03bc\u03b5\u03c4\u03b1\u03be\u03cd datasets, \u03b7 \u03b5\u03c1\u03b3\u03b1\u03c3\u03af\u03b1 \u03b1\u03bd\u03b1\u03c6\u03ad\u03c1\u03b5\u03b9 mean ranks. \u03a4\u03bf SPARC \u03b5\u03af\u03c7\u03b5 rank 1,00 \u03c3\u03c4\u03bf NLL \u03ba\u03b1\u03b9 2,69 \u03c3\u03c4\u03bf \u03c3\u03c5\u03bd\u03b4\u03c5\u03b1\u03c3\u03bc\u03ad\u03bd\u03bf MPJPE+NLL, \u03c4\u03b1 \u03ba\u03b1\u03bb\u03cd\u03c4\u03b5\u03c1\u03b1 \u03b1\u03c0\u03bf\u03c4\u03b5\u03bb\u03ad\u03c3\u03bc\u03b1\u03c4\u03b1 \u03bc\u03b5\u03c4\u03b1\u03be\u03cd \u03c4\u03c9\u03bd \u03c0\u03b1\u03c1\u03bf\u03c5\u03c3\u03b9\u03b1\u03b6\u03cc\u03bc\u03b5\u03bd\u03c9\u03bd \u03bc\u03b5\u03b8\u03cc\u03b4\u03c9\u03bd. \u03a3\u03c4\u03bf point accuracy \u03b5\u03af\u03c7\u03b5 MPJPE rank 4,39, \u03ac\u03c1\u03b1 \u03b4\u03b5\u03bd \u03ae\u03c4\u03b1\u03bd \u03ba\u03b1\u03b8\u03bf\u03bb\u03b9\u03ba\u03cc\u03c2 \u03bd\u03b9\u03ba\u03b7\u03c4\u03ae\u03c2. \u03a3\u03c4\u03bf \u03c0\u03bb\u03ac\u03c4\u03bf\u03c2 \u03c4\u03c9\u03bd conformal 95% tubes \u03b5\u03af\u03c7\u03b5 rank 2,50.<\/p>\n<div class=\"td-chart td-chart--metrics\">\n<div class=\"td-chart-head\">\n<p class=\"td-chart-title\">\u03a4\u03bf SPARC \u03c3\u03b5 \u03c4\u03ad\u03c3\u03c3\u03b5\u03c1\u03b9\u03c2 \u03b5\u03c0\u03b1\u03bb\u03b7\u03b8\u03b5\u03c5\u03bc\u03ad\u03bd\u03b5\u03c2 \u03bc\u03b5\u03c4\u03c1\u03b9\u03ba\u03ad\u03c2<\/p>\n<p class=\"td-chart-subtitle\">\u039f\u03b9 \u03b1\u03c1\u03b9\u03b8\u03bc\u03bf\u03af \u03b5\u03af\u03bd\u03b1\u03b9 mean ranks \u03c3\u03c4\u03bf \u03c3\u03c5\u03b3\u03ba\u03b5\u03ba\u03c1\u03b9\u03bc\u03ad\u03bd\u03bf cross-dataset \u03c0\u03c1\u03c9\u03c4\u03cc\u03ba\u03bf\u03bb\u03bb\u03bf\u00b7 \u03bc\u03b9\u03ba\u03c1\u03cc\u03c4\u03b5\u03c1\u03bf \u03b5\u03af\u03bd\u03b1\u03b9 \u03ba\u03b1\u03bb\u03cd\u03c4\u03b5\u03c1\u03bf \u03ba\u03b1\u03b9 \u03b4\u03b5\u03bd \u03b1\u03c0\u03bf\u03c4\u03b5\u03bb\u03bf\u03cd\u03bd \u03c0\u03bf\u03c3\u03bf\u03c3\u03c4\u03ac \u03b5\u03c0\u03b9\u03c4\u03c5\u03c7\u03af\u03b1\u03c2.<\/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\">1,00<\/span><span class=\"td-metric-label\">mean rank \u03c3\u03c4\u03bf Gaussian NLL<\/span><\/div>\n<div class=\"td-metric-card\"><span class=\"td-metric-value\">2,69<\/span><span class=\"td-metric-label\">mean rank \u03c3\u03c4\u03bf combined MPJPE + NLL<\/span><\/div>\n<div class=\"td-metric-card\"><span class=\"td-metric-value\">4,39<\/span><span class=\"td-metric-label\">mean rank \u03c3\u03c4\u03bf point-forecast MPJPE<\/span><\/div>\n<div class=\"td-metric-card\"><span class=\"td-metric-value\">2,50<\/span><span class=\"td-metric-label\">mean rank \u03c3\u03c4\u03bf conformal tube width W95<\/span><\/div>\n<\/div>\n<\/div>\n<\/div>\n<p>\u03a3\u03c4\u03bf Human3.6M, \u03c4\u03bf \u03c5\u03c8\u03b7\u03bb\u03cc\u03c4\u03b5\u03c1\u03bf \u03ba-decile \u03b5\u03af\u03c7\u03b5 1,79 \u03c6\u03bf\u03c1\u03ad\u03c2 \u03c4\u03bf \u03bc\u03ad\u03c3\u03bf MPJPE. \u0397 \u03b1\u03c6\u03b1\u03af\u03c1\u03b5\u03c3\u03b7 \u03b1\u03c5\u03c4\u03bf\u03cd \u03c4\u03bf\u03c5 decile \u03bc\u03b5\u03af\u03c9\u03c3\u03b5 \u03c4\u03bf MPJPE \u03c0\u03b5\u03c1\u03af\u03c0\u03bf\u03c5 9%. \u039f\u03b9 thresholds \u03b5\u03af\u03c7\u03b1\u03bd \u03c1\u03c5\u03b8\u03bc\u03b9\u03c3\u03c4\u03b5\u03af \u03c3\u03b5 held-out data \u03ba\u03b1\u03b9 \u03bf\u03b9 \u03c3\u03c5\u03b3\u03b3\u03c1\u03b1\u03c6\u03b5\u03af\u03c2 \u03b4\u03b7\u03bb\u03ce\u03bd\u03bf\u03c5\u03bd \u03cc\u03c4\u03b9 \u03b4\u03b5\u03bd \u03c5\u03c0\u03ac\u03c1\u03c7\u03b5\u03b9 formal guarantee \u03b3\u03b9\u03b1 \u03b1\u03c5\u03c4\u03cc \u03c4\u03bf monitoring \u03b1\u03c0\u03bf\u03c4\u03ad\u03bb\u03b5\u03c3\u03bc\u03b1. \u0397 \u03c3\u03c9\u03c3\u03c4\u03ae \u03b1\u03bd\u03ac\u03b3\u03bd\u03c9\u03c3\u03b7 \u03b5\u03af\u03bd\u03b1\u03b9 risk ranking \u03c0\u03bf\u03c5 \u03b1\u03be\u03af\u03b6\u03b5\u03b9 \u03c0\u03b5\u03c1\u03b1\u03b9\u03c4\u03ad\u03c1\u03c9 validation, \u03cc\u03c7\u03b9 \u03c0\u03b9\u03c3\u03c4\u03bf\u03c0\u03bf\u03af\u03b7\u03c3\u03b7 \u03cc\u03c4\u03b9 \u03ba\u03ac\u03b8\u03b5 high-\u03ba \u03c0\u03b5\u03c1\u03af\u03c0\u03c4\u03c9\u03c3\u03b7 \u03b8\u03b1 \u03b1\u03c0\u03bf\u03c4\u03cd\u03c7\u03b5\u03b9.<\/p>\n<h2 id=\"trade-offs-ablation\">\u03a4\u03b9 \u03b1\u03c0\u03bf\u03ba\u03ac\u03bb\u03c5\u03c8\u03b5 \u03c4\u03bf ablation \u03b3\u03b9\u03b1 \u03c4\u03b1 trade-offs<\/h2>\n<p>\u03a4\u03bf component ablation \u03b4\u03b5\u03af\u03c7\u03bd\u03b5\u03b9 \u03cc\u03c4\u03b9 \u03c4\u03b1 \u03ba\u03ad\u03c1\u03b4\u03b7 \u03b4\u03b5\u03bd \u03c0\u03c1\u03bf\u03ad\u03c1\u03c7\u03bf\u03bd\u03c4\u03b1\u03b9 \u03b1\u03c0\u03cc \u03ad\u03bd\u03b1\u03bd \u03bc\u03cc\u03bd\u03bf \u03bc\u03b7\u03c7\u03b1\u03bd\u03b9\u03c3\u03bc\u03cc. \u039c\u03ad\u03c3\u03b1 \u03c3\u03c4\u03b7\u03bd FC-Out \u03bf\u03b9\u03ba\u03bf\u03b3\u03ad\u03bd\u03b5\u03b9\u03b1, \u03c4\u03bf \u03b2\u03b1\u03c3\u03b9\u03ba\u03cc \u03ba variant \u03b5\u03af\u03c7\u03b5 MPJPE+NLL mean rank 5,28. \u0397 hybrid matrix-normal \u03b4\u03bf\u03bc\u03ae \u03c4\u03bf \u03b2\u03b5\u03bb\u03c4\u03af\u03c9\u03c3\u03b5 \u03c3\u03b5 3,94 \u03ba\u03b1\u03b9 \u03c4\u03bf GraphJ coupling \u03c3\u03b5 2,47. \u03a4\u03bf SPARC \u03bc\u03b5 time-coupled shrinkage \u03b5\u03af\u03c7\u03b5 3,17 \u03c3\u03b5 \u03b1\u03c5\u03c4\u03cc \u03c4\u03bf within-family metric.<\/p>\n<p>\u0397 time coupling \u03b4\u03b5\u03bd \u03c0\u03b1\u03c1\u03bf\u03c5\u03c3\u03b9\u03ac\u03b6\u03b5\u03c4\u03b1\u03b9 \u03c9\u03c2 \u03ba\u03b1\u03b8\u03b1\u03c1\u03ae \u03bd\u03af\u03ba\u03b7 \u03c3\u03c4\u03bf NLL. \u0395\u03af\u03bd\u03b1\u03b9 calibration-efficiency knob: \u03b8\u03c5\u03c3\u03b9\u03ac\u03b6\u03b5\u03b9 \u03bc\u03ad\u03c1\u03bf\u03c2 \u03c4\u03bf\u03c5 NLL rank \u03c3\u03b5 \u03c3\u03c7\u03ad\u03c3\u03b7 \u03bc\u03b5 \u03c4\u03bf uncoupled hybrid MN-GraphJ, \u03b1\u03bb\u03bb\u03ac \u03c5\u03c0\u03bf\u03c3\u03c4\u03b7\u03c1\u03af\u03b6\u03b5\u03b9 \u03ba\u03b1\u03bb\u03cd\u03c4\u03b5\u03c1\u03bf calibrated tube trade-off \u03c3\u03c4\u03bf \u03c0\u03bb\u03ae\u03c1\u03b5\u03c2 benchmark. \u038c\u03c0\u03c9\u03c2 \u03ba\u03b1\u03b9 \u03c3\u03c4\u03bf <a href=\"https:\/\/twodots.gr\/ai-agents-behavioral-testing\/\">behavioral testing \u03c4\u03c9\u03bd AI agents<\/a>, \u03b7 \u03c4\u03b5\u03bb\u03b9\u03ba\u03ae \u03b5\u03c0\u03b9\u03bb\u03bf\u03b3\u03ae \u03c0\u03c1\u03ad\u03c0\u03b5\u03b9 \u03bd\u03b1 \u03b1\u03ba\u03bf\u03bb\u03bf\u03c5\u03b8\u03b5\u03af \u03c4\u03bf failure mode \u03c0\u03bf\u03c5 \u03ad\u03c7\u03b5\u03b9 \u03c0\u03c1\u03b1\u03b3\u03bc\u03b1\u03c4\u03b9\u03ba\u03cc \u03ba\u03cc\u03c3\u03c4\u03bf\u03c2, \u03cc\u03c7\u03b9 \u03ad\u03bd\u03b1 \u03c3\u03c5\u03bd\u03bf\u03bb\u03b9\u03ba\u03cc score \u03c7\u03c9\u03c1\u03af\u03c2 \u03b5\u03c0\u03b9\u03c7\u03b5\u03b9\u03c1\u03b7\u03c3\u03b9\u03b1\u03ba\u03cc \u03c0\u03bb\u03b1\u03af\u03c3\u03b9\u03bf.<\/p>\n<h2 id=\"single-pass-compute\">\u0388\u03bd\u03b1 forward pass \u03b4\u03b5\u03bd \u03c3\u03b7\u03bc\u03b1\u03af\u03bd\u03b5\u03b9 \u03bc\u03b7\u03b4\u03b5\u03bd\u03b9\u03ba\u03cc \u03ba\u03cc\u03c3\u03c4\u03bf\u03c2<\/h2>\n<p>\u038c\u03bb\u03b1 \u03c4\u03b1 structured heads \u03c4\u03b7\u03c2 \u03b5\u03c1\u03b3\u03b1\u03c3\u03af\u03b1\u03c2, \u03c3\u03c5\u03bc\u03c0\u03b5\u03c1\u03b9\u03bb\u03b1\u03bc\u03b2\u03b1\u03bd\u03bf\u03bc\u03ad\u03bd\u03bf\u03c5 \u03c4\u03bf\u03c5 SPARC, \u03c7\u03c1\u03b5\u03b9\u03ac\u03b6\u03bf\u03bd\u03c4\u03b1\u03b9 \u03ad\u03bd\u03b1 forward pass \u03c3\u03c4\u03bf inference. \u03a4\u03bf \u03ba \u03c0\u03c1\u03bf\u03c3\u03b8\u03ad\u03c4\u03b5\u03b9 horizon-wise quadratic forms \u03ba\u03b1\u03b9 \u03b1\u03c0\u03bf\u03c6\u03b5\u03cd\u03b3\u03b5\u03b9 Monte Carlo sampling \u03ae ensembles. \u0397 \u03c0\u03c1\u03bf\u03c3\u03b1\u03c1\u03bc\u03bf\u03b3\u03ae \u03c4\u03bf\u03c5 inverse sufficient-statistics matrix \u03b3\u03af\u03bd\u03b5\u03c4\u03b1\u03b9 offline, \u03b5\u03bd\u03ce \u03c4\u03bf conformal calibration \u03c7\u03c1\u03b7\u03c3\u03b9\u03bc\u03bf\u03c0\u03bf\u03b9\u03b5\u03af \u03c4\u03b1 512 held-out windows.<\/p>\n<p>\u03a4\u03bf \u00absingle-pass\u00bb \u03c0\u03b5\u03c1\u03b9\u03b3\u03c1\u03ac\u03c6\u03b5\u03b9 \u03c4\u03bf inference pattern, \u03cc\u03c7\u03b9 \u03bf\u03bb\u03cc\u03ba\u03bb\u03b7\u03c1\u03bf \u03c4\u03bf lifecycle. \u03a0\u03b1\u03c1\u03b1\u03bc\u03ad\u03bd\u03bf\u03c5\u03bd \u03c4\u03bf \u03ba\u03cc\u03c3\u03c4\u03bf\u03c2 \u03b5\u03ba\u03c0\u03b1\u03af\u03b4\u03b5\u03c5\u03c3\u03b7\u03c2 \u03c4\u03bf\u03c5 forecaster, \u03c4\u03bf fitting \u03c4\u03c9\u03bd covariance heads, \u03b7 \u03b1\u03c0\u03bf\u03b8\u03ae\u03ba\u03b5\u03c5\u03c3\u03b7 \u03ba\u03b1\u03b9 \u03ad\u03ba\u03b4\u03bf\u03c3\u03b7 \u03c4\u03c9\u03bd statistics, \u03b7 \u03c3\u03c5\u03bd\u03c4\u03ae\u03c1\u03b7\u03c3\u03b7 calibration sets \u03ba\u03b1\u03b9 \u03b7 \u03c0\u03b1\u03c1\u03b1\u03ba\u03bf\u03bb\u03bf\u03cd\u03b8\u03b7\u03c3\u03b7 drift. \u0397 <a href=\"https:\/\/twodots.gr\/enterprise-ai-harnesses-diakyvernisi-architektoniki\/\">\u03b4\u03b9\u03b1\u03ba\u03c5\u03b2\u03ad\u03c1\u03bd\u03b7\u03c3\u03b7 \u03c9\u03c2 \u03bc\u03ad\u03c1\u03bf\u03c2 \u03c4\u03b7\u03c2 \u03b1\u03c1\u03c7\u03b9\u03c4\u03b5\u03ba\u03c4\u03bf\u03bd\u03b9\u03ba\u03ae\u03c2<\/a> \u03b5\u03af\u03bd\u03b1\u03b9 \u03c3\u03b7\u03bc\u03b1\u03bd\u03c4\u03b9\u03ba\u03ae \u03b1\u03ba\u03c1\u03b9\u03b2\u03ce\u03c2 \u03b5\u03c0\u03b5\u03b9\u03b4\u03ae \u03b1\u03c5\u03c4\u03ac \u03c4\u03b1 artifacts \u03c0\u03c1\u03ad\u03c0\u03b5\u03b9 \u03bd\u03b1 \u03b5\u03af\u03bd\u03b1\u03b9 \u03b1\u03bd\u03b1\u03c0\u03b1\u03c1\u03b1\u03b3\u03ce\u03b3\u03b9\u03bc\u03b1 \u03ba\u03b1\u03b9 \u03b5\u03bb\u03ad\u03b3\u03be\u03b9\u03bc\u03b1.<\/p>\n<h2 id=\"exchangeability-periorismoi\">\u0397 \u03b5\u03b3\u03b3\u03cd\u03b7\u03c3\u03b7 \u03c3\u03c4\u03b1\u03bc\u03b1\u03c4\u03ac \u03b5\u03ba\u03b5\u03af \u03cc\u03c0\u03bf\u03c5 \u03c3\u03c0\u03ac\u03b5\u03b9 \u03b7 exchangeability<\/h2>\n<p>\u0397 split conformal prediction \u03b4\u03af\u03bd\u03b5\u03b9 finite-sample marginal validity \u03cc\u03c4\u03b1\u03bd calibration \u03ba\u03b1\u03b9 evaluation \u03c0\u03b1\u03c1\u03b1\u03b4\u03b5\u03af\u03b3\u03bc\u03b1\u03c4\u03b1 \u03b5\u03af\u03bd\u03b1\u03b9 exchangeable. \u03a3\u03b5 \u03c0\u03c1\u03b1\u03b3\u03bc\u03b1\u03c4\u03b9\u03ba\u03ad\u03c2 \u03c7\u03c1\u03bf\u03bd\u03b9\u03ba\u03ad\u03c2 \u03c1\u03bf\u03ad\u03c2, \u03b1\u03bb\u03bb\u03b1\u03b3\u03ad\u03c2 \u03c3\u03c5\u03bc\u03c0\u03b5\u03c1\u03b9\u03c6\u03bf\u03c1\u03ac\u03c2, \u03bd\u03ad\u03b1 \u03ba\u03b1\u03b8\u03ae\u03ba\u03bf\u03bd\u03c4\u03b1, \u03b4\u03b9\u03b1\u03c6\u03bf\u03c1\u03b5\u03c4\u03b9\u03ba\u03ac \u03c3\u03ce\u03bc\u03b1\u03c4\u03b1 \u03ae distribution shift \u03bc\u03c0\u03bf\u03c1\u03bf\u03cd\u03bd \u03bd\u03b1 \u03c0\u03b1\u03c1\u03b1\u03b2\u03b9\u03ac\u03c3\u03bf\u03c5\u03bd \u03b1\u03c5\u03c4\u03ae \u03c4\u03b7\u03bd \u03c5\u03c0\u03cc\u03b8\u03b5\u03c3\u03b7. \u03a4\u03bf \u03af\u03b4\u03b9\u03bf \u03c4\u03bf paper \u03b6\u03b7\u03c4\u03ac \u03c4\u03cc\u03c4\u03b5 \u03b7 calibration \u03bd\u03b1 \u03b4\u03b9\u03b1\u03b2\u03ac\u03b6\u03b5\u03c4\u03b1\u03b9 \u03b5\u03bc\u03c0\u03b5\u03b9\u03c1\u03b9\u03ba\u03ac \u03ba\u03b1\u03b9 \u03cc\u03c7\u03b9 \u03c9\u03c2 unconditional guarantee.<\/p>\n<p>\u03a5\u03c0\u03ac\u03c1\u03c7\u03bf\u03c5\u03bd \u03ba\u03b1\u03b9 \u03ac\u03bb\u03bb\u03bf\u03b9 \u03c0\u03b5\u03c1\u03b9\u03bf\u03c1\u03b9\u03c3\u03bc\u03bf\u03af. \u0397 \u03c0\u03bf\u03b9\u03cc\u03c4\u03b7\u03c4\u03b1 \u03c4\u03bf\u03c5 \u03ba \u03b5\u03be\u03b1\u03c1\u03c4\u03ac\u03c4\u03b1\u03b9 \u03b1\u03c0\u03cc learned features. \u03a0\u03bb\u03bf\u03c5\u03c3\u03b9\u03cc\u03c4\u03b5\u03c1\u03b5\u03c2 \u03c3\u03c5\u03c3\u03c7\u03b5\u03c4\u03af\u03c3\u03b5\u03b9\u03c2 \u03bc\u03c0\u03bf\u03c1\u03b5\u03af \u03bd\u03b1 \u03b1\u03c0\u03b1\u03b9\u03c4\u03bf\u03cd\u03bd \u03c0\u03c5\u03ba\u03bd\u03cc\u03c4\u03b5\u03c1\u03b1 coordinate models \u03ba\u03b1\u03b9 \u03b5\u03c0\u03b9\u03c0\u03bb\u03ad\u03bf\u03bd compute. \u03a4\u03b1 \u03b1\u03c0\u03bf\u03c4\u03b5\u03bb\u03ad\u03c3\u03bc\u03b1\u03c4\u03b1 \u03b9\u03c3\u03c7\u03cd\u03bf\u03c5\u03bd \u03b3\u03b9\u03b1 \u03c4\u03b1 datasets, \u03c4\u03b1 splits \u03ba\u03b1\u03b9 \u03c4\u03b9\u03c2 \u03bc\u03b5\u03c4\u03c1\u03b9\u03ba\u03ad\u03c2 \u03c4\u03b7\u03c2 \u03bc\u03b5\u03bb\u03ad\u03c4\u03b7\u03c2. \u039f\u03b9 split-seed, fit-seed \u03ba\u03b1\u03b9 repeated-HPO diagnostics \u03b5\u03bd\u03b9\u03c3\u03c7\u03cd\u03bf\u03c5\u03bd \u03c4\u03b7\u03bd \u03b1\u03be\u03b9\u03bf\u03bb\u03cc\u03b3\u03b7\u03c3\u03b7, \u03b1\u03bb\u03bb\u03ac \u03b4\u03b5\u03bd \u03b1\u03bd\u03c4\u03b9\u03ba\u03b1\u03b8\u03b9\u03c3\u03c4\u03bf\u03cd\u03bd validation \u03c3\u03c4\u03bf \u03c0\u03c1\u03b1\u03b3\u03bc\u03b1\u03c4\u03b9\u03ba\u03cc \u03c0\u03b5\u03c1\u03b9\u03b2\u03ac\u03bb\u03bb\u03bf\u03bd.<\/p>\n<h2 id=\"risk-interface-epicheiriseis\">\u0391\u03c0\u03cc \u03c4\u03b7 \u03c1\u03bf\u03bc\u03c0\u03bf\u03c4\u03b9\u03ba\u03ae \u03c3\u03b5 \u03ad\u03bd\u03b1 \u03b3\u03b5\u03bd\u03b9\u03ba\u03cc risk interface<\/h2>\n<p>\u0397 \u03ac\u03bc\u03b5\u03c3\u03b7 \u03b5\u03c6\u03b1\u03c1\u03bc\u03bf\u03b3\u03ae \u03c4\u03bf\u03c5 SPARC \u03b5\u03af\u03bd\u03b1\u03b9 \u03b7 \u03c0\u03c1\u03cc\u03b2\u03bb\u03b5\u03c8\u03b7 \u03b1\u03bd\u03b8\u03c1\u03ce\u03c0\u03b9\u03bd\u03b7\u03c2 \u03ba\u03af\u03bd\u03b7\u03c3\u03b7\u03c2 \u03b3\u03b9\u03b1 HRI, human-aware navigation \u03ba\u03b1\u03b9 safety filters. \u0394\u03b5\u03bd \u03b1\u03c0\u03bf\u03b4\u03b5\u03b9\u03ba\u03bd\u03cd\u03b5\u03b9 \u03cc\u03c4\u03b9 \u03b7 \u03af\u03b4\u03b9\u03b1 \u03b1\u03c1\u03c7\u03b9\u03c4\u03b5\u03ba\u03c4\u03bf\u03bd\u03b9\u03ba\u03ae \u03c0\u03c1\u03ad\u03c0\u03b5\u03b9 \u03bd\u03b1 \u03bc\u03b5\u03c4\u03b1\u03c6\u03b5\u03c1\u03b8\u03b5\u03af \u03b1\u03c5\u03c4\u03bf\u03cd\u03c3\u03b9\u03b1 \u03c3\u03b5 demand forecasting, marketing attribution \u03ae recommendations. \u03a0\u03c1\u03bf\u03c3\u03c6\u03ad\u03c1\u03b5\u03b9 \u03cc\u03bc\u03c9\u03c2 \u03ad\u03bd\u03b1 \u03b5\u03bb\u03ad\u03b3\u03be\u03b9\u03bc\u03bf design principle: \u03c0\u03c1\u03b9\u03bd \u03ad\u03bd\u03b1 confidence signal \u03bf\u03b4\u03b7\u03b3\u03ae\u03c3\u03b5\u03b9 \u03c3\u03b5 \u03b4\u03c1\u03ac\u03c3\u03b7, \u03c0\u03c1\u03ad\u03c0\u03b5\u03b9 \u03bd\u03b1 \u03b5\u03af\u03bd\u03b1\u03b9 \u03c3\u03b1\u03c6\u03ad\u03c2 \u03c4\u03b9 \u03bc\u03b5\u03c4\u03c1\u03ac, \u03c3\u03b5 \u03c0\u03bf\u03b9\u03b1 \u03b4\u03b5\u03b4\u03bf\u03bc\u03ad\u03bd\u03b1 \u03b2\u03b1\u03b8\u03bc\u03bf\u03bd\u03bf\u03bc\u03ae\u03b8\u03b7\u03ba\u03b5 \u03ba\u03b1\u03b9 \u03c0\u03bf\u03b9\u03b1 policy \u03b5\u03bd\u03b5\u03c1\u03b3\u03bf\u03c0\u03bf\u03b9\u03b5\u03af.<\/p>\n<p>\u0397 \u03b4\u03b9\u03ac\u03ba\u03c1\u03b9\u03c3\u03b7 \u03b1\u03c5\u03c4\u03ae \u03ad\u03c7\u03b5\u03b9 \u03b1\u03be\u03af\u03b1 \u03ba\u03b1\u03b9 \u03c3\u03c4\u03b1 \u03b5\u03c0\u03b9\u03c7\u03b5\u03b9\u03c1\u03b7\u03c3\u03b9\u03b1\u03ba\u03ac AI workflows. \u03a4\u03bf <a href=\"https:\/\/twodots.gr\/why2speak-exigisi-ai-agent-allazei-apofasi\/\">rationale \u03b5\u03bd\u03cc\u03c2 AI agent \u03bc\u03c0\u03bf\u03c1\u03b5\u03af \u03bd\u03b1 \u03b1\u03bb\u03bb\u03ac\u03b6\u03b5\u03b9 \u03c4\u03b7\u03bd \u03b1\u03c0\u03cc\u03c6\u03b1\u03c3\u03ae \u03c4\u03bf\u03c5<\/a>, \u03b5\u03bd\u03ce \u03bf\u03b9 <a href=\"https:\/\/twodots.gr\/ai-agents-paragogi-leitourgikes-astochies\/\">\u03bb\u03b5\u03b9\u03c4\u03bf\u03c5\u03c1\u03b3\u03b9\u03ba\u03ad\u03c2 \u03b1\u03c3\u03c4\u03bf\u03c7\u03af\u03b5\u03c2 \u03c3\u03c4\u03b7\u03bd \u03c0\u03b1\u03c1\u03b1\u03b3\u03c9\u03b3\u03ae<\/a> \u03c3\u03c5\u03c7\u03bd\u03ac \u03b2\u03c1\u03af\u03c3\u03ba\u03bf\u03bd\u03c4\u03b1\u03b9 \u03ad\u03be\u03c9 \u03b1\u03c0\u03cc \u03c4\u03bf \u03af\u03b4\u03b9\u03bf \u03c4\u03bf \u03bc\u03bf\u03bd\u03c4\u03ad\u03bb\u03bf. \u0388\u03bd\u03b1 risk interface \u03c0\u03c1\u03ad\u03c0\u03b5\u03b9 \u03b5\u03c0\u03bf\u03bc\u03ad\u03bd\u03c9\u03c2 \u03bd\u03b1 \u03c3\u03c5\u03bd\u03b4\u03ad\u03b5\u03c4\u03b1\u03b9 \u03bc\u03b5 observable state, thresholds, human escalation, safe mode \u03ba\u03b1\u03b9 action logs.<\/p>\n<h2 id=\"deployment-checklist\">\u0395\u03c0\u03c4\u03ac \u03ad\u03bb\u03b5\u03b3\u03c7\u03bf\u03b9 \u03c0\u03c1\u03b9\u03bd \u03b1\u03c0\u03cc deployment<\/h2>\n<p>\u03a4\u03bf SPARC \u03ba\u03ac\u03bd\u03b5\u03b9 \u03c4\u03b7\u03bd \u03b1\u03b2\u03b5\u03b2\u03b1\u03b9\u03cc\u03c4\u03b7\u03c4\u03b1 \u03c0\u03b9\u03bf \u03c3\u03c5\u03b3\u03ba\u03b5\u03ba\u03c1\u03b9\u03bc\u03ad\u03bd\u03b7, \u03b1\u03bb\u03bb\u03ac \u03b4\u03b5\u03bd \u03b1\u03c0\u03bf\u03c6\u03b1\u03c3\u03af\u03b6\u03b5\u03b9 \u03bc\u03cc\u03bd\u03bf \u03c4\u03bf\u03c5 \u03c0\u03ce\u03c2 \u03b8\u03b1 \u03c7\u03c1\u03b7\u03c3\u03b9\u03bc\u03bf\u03c0\u03bf\u03b9\u03b7\u03b8\u03b5\u03af. \u0397 \u03c0\u03b1\u03c1\u03b1\u03ba\u03ac\u03c4\u03c9 \u03b1\u03ba\u03bf\u03bb\u03bf\u03c5\u03b8\u03af\u03b1 \u03bc\u03b5\u03c4\u03b1\u03c4\u03c1\u03ad\u03c0\u03b5\u03b9 \u03c4\u03b1 \u03b5\u03c1\u03b5\u03c5\u03bd\u03b7\u03c4\u03b9\u03ba\u03ac \u03b5\u03c5\u03c1\u03ae\u03bc\u03b1\u03c4\u03b1 \u03c3\u03b5 \u03c0\u03c1\u03b1\u03ba\u03c4\u03b9\u03ba\u03cc pilot plan \u03c7\u03c9\u03c1\u03af\u03c2 \u03bd\u03b1 \u03b5\u03c0\u03b5\u03ba\u03c4\u03b5\u03af\u03bd\u03b5\u03b9 \u03c4\u03bf\u03c5\u03c2 \u03b9\u03c3\u03c7\u03c5\u03c1\u03b9\u03c3\u03bc\u03bf\u03cd\u03c2 \u03c4\u03b7\u03c2 \u03bc\u03b5\u03bb\u03ad\u03c4\u03b7\u03c2.<\/p>\n<div class=\"td-step-list\">\n<p class=\"td-step-list-title\">\u0391\u03c0\u03cc \u03c4\u03bf prediction tube \u03c3\u03b5 \u03b1\u03c3\u03c6\u03b1\u03bb\u03ae \u03b5\u03c0\u03b9\u03c7\u03b5\u03b9\u03c1\u03b7\u03c3\u03b9\u03b1\u03ba\u03ae \u03c0\u03bf\u03bb\u03b9\u03c4\u03b9\u03ba\u03ae<\/p>\n<ol>\n<li><span class=\"td-step-kicker\">\u0392\u03ae\u03bc\u03b1 1<\/span><strong>\u039f\u03c1\u03af\u03c3\u03c4\u03b5 \u03c0\u03bf\u03b9\u03bf \u03c3\u03c6\u03ac\u03bb\u03bc\u03b1 \u03ad\u03c7\u03b5\u03b9 \u03c0\u03c1\u03b1\u03b3\u03bc\u03b1\u03c4\u03b9\u03ba\u03cc \u03ba\u03cc\u03c3\u03c4\u03bf\u03c2<\/strong>\n<p>\u039e\u03b5\u03c7\u03c9\u03c1\u03af\u03c3\u03c4\u03b5 point error, \u03c5\u03c0\u03b5\u03c1\u03b2\u03bf\u03bb\u03b9\u03ba\u03ae \u03b2\u03b5\u03b2\u03b1\u03b9\u03cc\u03c4\u03b7\u03c4\u03b1, \u03c0\u03bf\u03bb\u03cd \u03c0\u03bb\u03b1\u03c4\u03cd interval \u03ba\u03b1\u03b9 \u03ba\u03b1\u03b8\u03c5\u03c3\u03c4\u03b5\u03c1\u03b7\u03bc\u03ad\u03bd\u03bf fallback. \u0397 \u03bc\u03b5\u03c4\u03c1\u03b9\u03ba\u03ae \u03c0\u03c1\u03ad\u03c0\u03b5\u03b9 \u03bd\u03b1 \u03b1\u03ba\u03bf\u03bb\u03bf\u03c5\u03b8\u03b5\u03af \u03c4\u03b7 \u03b6\u03b7\u03bc\u03b9\u03ac \u03c0\u03bf\u03c5 \u03bc\u03c0\u03bf\u03c1\u03b5\u03af \u03bd\u03b1 \u03c0\u03c1\u03bf\u03ba\u03b1\u03bb\u03ad\u03c3\u03b5\u03b9 \u03b7 \u03b1\u03c0\u03cc\u03c6\u03b1\u03c3\u03b7.<\/p>\n<\/li>\n<li><span class=\"td-step-kicker\">\u0392\u03ae\u03bc\u03b1 2<\/span><strong>\u039a\u03b1\u03b8\u03bf\u03c1\u03af\u03c3\u03c4\u03b5 \u03c4\u03b7\u03bd \u03ad\u03bd\u03bd\u03bf\u03b9\u03b1 \u03ba\u03ac\u03bb\u03c5\u03c8\u03b7\u03c2<\/strong>\n<p>\u0391\u03c0\u03bf\u03c6\u03b1\u03c3\u03af\u03c3\u03c4\u03b5 \u03b1\u03bd \u03c7\u03c1\u03b5\u03b9\u03ac\u03b6\u03b5\u03c3\u03c4\u03b5 marginal intervals, joint-wise \u03ba\u03ac\u03bb\u03c5\u03c8\u03b7 \u03ae trajectory-level set. \u03a4\u03bf 95% marginal coverage \u03c4\u03bf\u03c5 \u03b2\u03b1\u03c3\u03b9\u03ba\u03bf\u03cd SPARC \u03b4\u03b5\u03bd \u03b5\u03af\u03bd\u03b1\u03b9 \u03b5\u03b3\u03b3\u03cd\u03b7\u03c3\u03b7 \u03b3\u03b9\u03b1 \u03bf\u03bb\u03cc\u03ba\u03bb\u03b7\u03c1\u03b7 \u03c4\u03c1\u03bf\u03c7\u03b9\u03ac.<\/p>\n<\/li>\n<li><span class=\"td-step-kicker\">\u0392\u03ae\u03bc\u03b1 3<\/span><strong>\u039a\u03bb\u03b5\u03b9\u03b4\u03ce\u03c3\u03c4\u03b5 train, held-out, calibration \u03ba\u03b1\u03b9 evaluation \u03c3\u03cd\u03bd\u03bf\u03bb\u03b1<\/strong>\n<p>\u03a4\u03b5\u03ba\u03bc\u03b7\u03c1\u03b9\u03ce\u03c3\u03c4\u03b5 \u03b5\u03ba\u03b4\u03cc\u03c3\u03b5\u03b9\u03c2, \u03c7\u03c1\u03bf\u03bd\u03b9\u03ba\u03ac \u03c0\u03b1\u03c1\u03ac\u03b8\u03c5\u03c1\u03b1 \u03ba\u03b1\u03b9 \u03c7\u03c1\u03ae\u03c3\u03b7 \u03ba\u03ac\u03b8\u03b5 split. \u039c\u03b7\u03bd \u03b5\u03c0\u03b9\u03c4\u03c1\u03ad\u03c8\u03b5\u03c4\u03b5 \u03c3\u03c4\u03bf calibration \u03ae \u03c4\u03bf \u03c4\u03b5\u03bb\u03b9\u03ba\u03cc evaluation \u03bd\u03b1 \u03b3\u03af\u03bd\u03b5\u03b9 \u03ba\u03c1\u03c5\u03c6\u03cc tuning set.<\/p>\n<\/li>\n<li><span class=\"td-step-kicker\">\u0392\u03ae\u03bc\u03b1 4<\/span><strong>\u0395\u03bb\u03ad\u03b3\u03be\u03c4\u03b5 \u03b1\u03bd \u03c4\u03bf \u03ba \u03ba\u03b1\u03c4\u03b1\u03c4\u03ac\u03c3\u03c3\u03b5\u03b9 \u03c4\u03bf\u03c5\u03c2 \u03b4\u03b9\u03ba\u03bf\u03cd\u03c2 \u03c3\u03b1\u03c2 \u03ba\u03b9\u03bd\u03b4\u03cd\u03bd\u03bf\u03c5\u03c2<\/strong>\n<p>\u039c\u03b5\u03c4\u03c1\u03ae\u03c3\u03c4\u03b5 high-\u03ba failure concentration \u03b1\u03bd\u03ac \u03c7\u03ce\u03c1\u03bf, \u03ba\u03af\u03bd\u03b7\u03c3\u03b7 \u03ba\u03b1\u03b9 \u03c5\u03c0\u03bf\u03bf\u03bc\u03ac\u03b4\u03b1. \u03a4\u03bf 1,79\u00d7 \u03c4\u03bf\u03c5 Human3.6M \u03b4\u03b5\u03bd \u03bc\u03b5\u03c4\u03b1\u03c6\u03ad\u03c1\u03b5\u03c4\u03b1\u03b9 \u03b1\u03c5\u03c4\u03cc\u03bc\u03b1\u03c4\u03b1 \u03c3\u03c4\u03b7 \u03b4\u03b9\u03ba\u03ae \u03c3\u03b1\u03c2 \u03c1\u03bf\u03ae.<\/p>\n<\/li>\n<li><span class=\"td-step-kicker\">\u0392\u03ae\u03bc\u03b1 5<\/span><strong>\u03a3\u03c5\u03bd\u03b4\u03ad\u03c3\u03c4\u03b5 thresholds \u03bc\u03b5 \u03c3\u03b1\u03c6\u03ae fallback<\/strong>\n<p>\u0393\u03b9\u03b1 \u03ba\u03ac\u03b8\u03b5 risk band \u03bf\u03c1\u03af\u03c3\u03c4\u03b5 \u03b5\u03c0\u03b9\u03b2\u03c1\u03ac\u03b4\u03c5\u03bd\u03c3\u03b7, \u03bc\u03b5\u03b3\u03b1\u03bb\u03cd\u03c4\u03b5\u03c1\u03bf safety margin, abstention \u03ae \u03b1\u03bd\u03b8\u03c1\u03ce\u03c0\u03b9\u03bd\u03b7 \u03c0\u03b1\u03c1\u03ad\u03bc\u03b2\u03b1\u03c3\u03b7. \u0397 \u03c0\u03bf\u03bb\u03b9\u03c4\u03b9\u03ba\u03ae \u03c7\u03c1\u03b5\u03b9\u03ac\u03b6\u03b5\u03c4\u03b1\u03b9 \u03b4\u03bf\u03ba\u03b9\u03bc\u03ae \u03c5\u03c0\u03cc latency \u03ba\u03b1\u03b9 capacity constraints.<\/p>\n<\/li>\n<li><span class=\"td-step-kicker\">\u0392\u03ae\u03bc\u03b1 6<\/span><strong>\u03a0\u03b1\u03c1\u03b1\u03ba\u03bf\u03bb\u03bf\u03c5\u03b8\u03ae\u03c3\u03c4\u03b5 coverage, width \u03ba\u03b1\u03b9 drift \u03bc\u03b1\u03b6\u03af<\/strong>\n<p>\u0388\u03bd\u03b1 interval \u03bc\u03c0\u03bf\u03c1\u03b5\u03af \u03bd\u03b1 \u03b4\u03b9\u03b1\u03c4\u03b7\u03c1\u03b5\u03af coverage \u03b5\u03c0\u03b5\u03b9\u03b4\u03ae \u03ad\u03b3\u03b9\u03bd\u03b5 \u03c5\u03c0\u03b5\u03c1\u03b2\u03bf\u03bb\u03b9\u03ba\u03ac \u03c0\u03bb\u03b1\u03c4\u03cd. \u03a0\u03b1\u03c1\u03b1\u03ba\u03bf\u03bb\u03bf\u03c5\u03b8\u03ae\u03c3\u03c4\u03b5 \u03c4\u03b1\u03c5\u03c4\u03cc\u03c7\u03c1\u03bf\u03bd\u03b1 empirical coverage, mean width, MPJPE, \u03ba distribution \u03ba\u03b1\u03b9 \u03b1\u03bb\u03bb\u03b1\u03b3\u03ad\u03c2 \u03c0\u03bb\u03b7\u03b8\u03c5\u03c3\u03bc\u03bf\u03cd.<\/p>\n<\/li>\n<li><span class=\"td-step-kicker\">\u0392\u03ae\u03bc\u03b1 7<\/span><strong>\u039a\u03c1\u03b1\u03c4\u03ae\u03c3\u03c4\u03b5 replayable \u03b5\u03ba\u03b4\u03cc\u03c3\u03b5\u03b9\u03c2 \u03ba\u03b1\u03b9 \u03b1\u03bd\u03b8\u03c1\u03ce\u03c0\u03b9\u03bd\u03bf override<\/strong>\n<p>\u0391\u03c0\u03bf\u03b8\u03b7\u03ba\u03b5\u03cd\u03c3\u03c4\u03b5 model, covariance head, sufficient statistics, calibration quantiles, thresholds \u03ba\u03b1\u03b9 action outcome. \u0397 \u03b1\u03be\u03af\u03b1 \u03c4\u03bf\u03c5 <a href=\"https:\/\/twodots.gr\/dyna-2-rompot-mathainei-apo-anthropino-vinteo\/\">robot learning \u03b1\u03c0\u03cc \u03b1\u03bd\u03b8\u03c1\u03ce\u03c0\u03b9\u03bd\u03b7 \u03ba\u03af\u03bd\u03b7\u03c3\u03b7<\/a> \u03bc\u03b5\u03b3\u03b1\u03bb\u03ce\u03bd\u03b5\u03b9 \u03cc\u03c4\u03b1\u03bd \u03b7 \u03bf\u03bc\u03ac\u03b4\u03b1 \u03bc\u03c0\u03bf\u03c1\u03b5\u03af \u03bd\u03b1 \u03b1\u03bd\u03b1\u03ba\u03b1\u03c4\u03b1\u03c3\u03ba\u03b5\u03c5\u03ac\u03c3\u03b5\u03b9 \u03b3\u03b9\u03b1\u03c4\u03af \u03c4\u03bf \u03c3\u03cd\u03c3\u03c4\u03b7\u03bc\u03b1 \u03ac\u03c6\u03b7\u03c3\u03b5 \u03ae \u03c0\u03b5\u03c1\u03b9\u03cc\u03c1\u03b9\u03c3\u03b5 \u03bc\u03b9\u03b1 \u03b5\u03bd\u03ad\u03c1\u03b3\u03b5\u03b9\u03b1.<\/p>\n<\/li>\n<\/ol>\n<\/div>\n<h2 id=\"symperasma-sparc\">\u03a4\u03bf \u03c3\u03c5\u03bc\u03c0\u03ad\u03c1\u03b1\u03c3\u03bc\u03b1: \u03b3\u03c1\u03ae\u03b3\u03bf\u03c1\u03b7 \u03c0\u03c1\u03cc\u03b2\u03bb\u03b5\u03c8\u03b7 \u03bc\u03b5 \u03b5\u03b9\u03bb\u03b9\u03ba\u03c1\u03b9\u03bd\u03ae \u03cc\u03c1\u03b9\u03b1<\/h2>\n<p>\u0397 \u03c3\u03b7\u03bc\u03b1\u03bd\u03c4\u03b9\u03ba\u03cc\u03c4\u03b5\u03c1\u03b7 \u03c3\u03c5\u03bd\u03b5\u03b9\u03c3\u03c6\u03bf\u03c1\u03ac \u03c4\u03bf\u03c5 SPARC \u03b4\u03b5\u03bd \u03b5\u03af\u03bd\u03b1\u03b9 \u03cc\u03c4\u03b9 \u03ba\u03b5\u03c1\u03b4\u03af\u03b6\u03b5\u03b9 \u03c0\u03b1\u03bd\u03c4\u03bf\u03cd \u03c3\u03c4\u03b7\u03bd \u03b1\u03ba\u03c1\u03af\u03b2\u03b5\u03b9\u03b1 \u03c4\u03b7\u03c2 \u03ba\u03b5\u03bd\u03c4\u03c1\u03b9\u03ba\u03ae\u03c2 \u03c0\u03c1\u03cc\u03b2\u03bb\u03b5\u03c8\u03b7\u03c2. \u0395\u03af\u03bd\u03b1\u03b9 \u03cc\u03c4\u03b9 \u03c3\u03c5\u03bd\u03b4\u03ad\u03b5\u03b9 deterministic forecaster, structured aleatoric covariance, \u03b1\u03bd\u03b1\u03bb\u03c5\u03c4\u03b9\u03ba\u03cc epistemic scale \u03ba\u03b1\u03b9 conformal calibration \u03c3\u03b5 \u03ad\u03bd\u03b1 \u03c3\u03c5\u03bd\u03b5\u03ba\u03c4\u03b9\u03ba\u03cc interface. \u03a3\u03c4\u03bf \u03c3\u03c5\u03b3\u03ba\u03b5\u03ba\u03c1\u03b9\u03bc\u03ad\u03bd\u03bf benchmark \u03b5\u03af\u03c7\u03b5 \u03c4\u03b7\u03bd \u03ba\u03b1\u03bb\u03cd\u03c4\u03b5\u03c1\u03b7 \u03ba\u03b1\u03c4\u03ac\u03c4\u03b1\u03be\u03b7 NLL \u03ba\u03b1\u03b9 combined MPJPE+NLL, \u03b1\u03bd\u03c4\u03b1\u03b3\u03c9\u03bd\u03b9\u03c3\u03c4\u03b9\u03ba\u03cc point error \u03ba\u03b1\u03b9 \u03b1\u03c0\u03bf\u03b4\u03bf\u03c4\u03b9\u03ba\u03ac calibrated tubes.<\/p>\n<p>\u03a4\u03b1\u03c5\u03c4\u03cc\u03c7\u03c1\u03bf\u03bd\u03b1, \u03b7 \u03b5\u03c1\u03b3\u03b1\u03c3\u03af\u03b1 \u03ba\u03c1\u03b1\u03c4\u03ac \u03ba\u03b1\u03b8\u03b1\u03c1\u03ac \u03c4\u03b1 \u03cc\u03c1\u03b9\u03ac \u03c4\u03b7\u03c2: \u03cc\u03c7\u03b9 uniform MPJPE win, \u03cc\u03c7\u03b9 strong conditional guarantee, \u03cc\u03c7\u03b9 universal OOD detector \u03ba\u03b1\u03b9 \u03cc\u03c7\u03b9 formal guarantee \u03b3\u03b9\u03b1 \u03c4\u03b7\u03bd \u03ba-based selective policy. \u0391\u03c5\u03c4\u03ae \u03b7 \u03b1\u03ba\u03c1\u03af\u03b2\u03b5\u03b9\u03b1 \u03b5\u03af\u03bd\u03b1\u03b9 \u03ba\u03bf\u03bc\u03bc\u03ac\u03c4\u03b9 \u03c4\u03b7\u03c2 \u03b1\u03be\u03af\u03b1\u03c2 \u03c4\u03b7\u03c2. \u03a3\u03b5 safety-sensitive AI, \u03b1\u03be\u03b9\u03bf\u03c0\u03b9\u03c3\u03c4\u03af\u03b1 \u03b4\u03b5\u03bd \u03b5\u03af\u03bd\u03b1\u03b9 \u03bc\u03cc\u03bd\u03bf \u03bd\u03b1 \u03be\u03ad\u03c1\u03b5\u03b9\u03c2 \u03c4\u03b9 \u03c0\u03c1\u03bf\u03b2\u03bb\u03ad\u03c0\u03b5\u03b9 \u03c4\u03bf \u03bc\u03bf\u03bd\u03c4\u03ad\u03bb\u03bf. \u0395\u03af\u03bd\u03b1\u03b9 \u03bd\u03b1 \u03be\u03ad\u03c1\u03b5\u03b9\u03c2 \u03c0\u03cc\u03c3\u03bf \u03c7\u03ce\u03c1\u03bf \u03c0\u03c1\u03ad\u03c0\u03b5\u03b9 \u03bd\u03b1 \u03b1\u03c6\u03ae\u03c3\u03b5\u03b9\u03c2 \u03b3\u03b9\u03b1 \u03cc\u03c3\u03b1 \u03b4\u03b5\u03bd \u03b3\u03bd\u03c9\u03c1\u03af\u03b6\u03b5\u03b9.<\/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 confidence score \u03c3\u03b5 \u03b5\u03bb\u03b5\u03b3\u03c7\u03cc\u03bc\u03b5\u03bd\u03b7 \u03c1\u03bf\u03ae<\/p>\n<h3>\u03a3\u03c7\u03b5\u03b4\u03b9\u03ac\u03c3\u03c4\u03b5 AI automations \u03bc\u03b5 thresholds, validation \u03ba\u03b1\u03b9 \u03b1\u03c3\u03c6\u03b1\u03bb\u03ae fallbacks<\/h3>\n<p>\u0397 TWO DOTS \u03c7\u03b1\u03c1\u03c4\u03bf\u03b3\u03c1\u03b1\u03c6\u03b5\u03af risk signals, human approvals, monitoring, action logs \u03ba\u03b1\u03b9 fallback paths \u03ce\u03c3\u03c4\u03b5 \u03ad\u03bd\u03b1 AI workflow \u03bd\u03b1 \u03c0\u03b1\u03c1\u03b1\u03bc\u03ad\u03bd\u03b5\u03b9 \u03c7\u03c1\u03ae\u03c3\u03b9\u03bc\u03bf \u03c7\u03c9\u03c1\u03af\u03c2 \u03bd\u03b1 \u03ba\u03c1\u03cd\u03b2\u03b5\u03b9 \u03c4\u03b1 \u03cc\u03c1\u03b9\u03b1 \u03c4\u03b7\u03c2 \u03c0\u03c1\u03cc\u03b2\u03bb\u03b5\u03c8\u03ae\u03c2 \u03c4\u03bf\u03c5.<\/p>\n<div class=\"td-service-cta-actions\"><a class=\"td-service-cta-button\" href=\"https:\/\/twodots.gr\/aftomatismoi-epicheiriseon-ai\/\">\u0394\u03b5\u03af\u03c4\u03b5 \u03c4\u03bf\u03c5\u03c2 \u0391\u03c5\u03c4\u03bf\u03bc\u03b1\u03c4\u03b9\u03c3\u03bc\u03bf\u03cd\u03c2 \u0395\u03c0\u03b9\u03c7\u03b5\u03b9\u03c1\u03ae\u03c3\u03b5\u03c9\u03bd &amp; AI<\/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\">\u0427\u0435\u0441\u0442\u043e \u0437\u0430\u0434\u0430\u0432\u0430\u043d\u0438 \u0432\u044a\u043f\u0440\u043e\u0441\u0438<\/p>\n<details class=\"td-faq-item\">\n<summary class=\"td-faq-title\">\u03a4\u03b9 \u03c3\u03b7\u03bc\u03b1\u03af\u03bd\u03b5\u03b9 SPARC;<\/summary>\n<div class=\"td-faq-content\">\n<p>Single-Pass Adaptive Risk Calibration. \u0395\u03af\u03bd\u03b1\u03b9 uncertainty layer \u03b3\u03b9\u03b1 human motion forecasting \u03c0\u03bf\u03c5 \u03c3\u03c5\u03bd\u03b4\u03c5\u03ac\u03b6\u03b5\u03b9 Bayesian last-layer scaling, structured Gaussian covariance \u03ba\u03b1\u03b9 split conformal calibration.<\/p>\n<\/div>\n<\/details>\n<details class=\"td-faq-item\">\n<summary class=\"td-faq-title\">\u03a7\u03c1\u03b5\u03b9\u03ac\u03b6\u03b5\u03c4\u03b1\u03b9 \u03c4\u03bf SPARC \u03c0\u03bf\u03bb\u03bb\u03ac inference passes;<\/summary>\n<div class=\"td-faq-content\">\n<p>\u038c\u03c7\u03b9. \u03a7\u03c1\u03b7\u03c3\u03b9\u03bc\u03bf\u03c0\u03bf\u03b9\u03b5\u03af \u03ad\u03bd\u03b1 forward pass \u03ba\u03b1\u03b9 \u03ba\u03bb\u03b5\u03b9\u03c3\u03c4\u03bf\u03cd \u03c4\u03cd\u03c0\u03bf\u03c5 quadratic-form \u03c5\u03c0\u03bf\u03bb\u03bf\u03b3\u03b9\u03c3\u03bc\u03bf\u03cd\u03c2 \u03b3\u03b9\u03b1 \u03c4\u03bf \u03ba, \u03c7\u03c9\u03c1\u03af\u03c2 Monte Carlo sampling \u03ae ensemble inference.<\/p>\n<\/div>\n<\/details>\n<details class=\"td-faq-item\">\n<summary class=\"td-faq-title\">\u03a4\u03b9 \u03bc\u03b5\u03c4\u03c1\u03ac \u03c4\u03bf \u03ba \u03c3\u03c4\u03bf SPARC;<\/summary>\n<div class=\"td-faq-content\">\n<p>\u039c\u03b5\u03c4\u03c1\u03ac feature-space leverage \u03c3\u03b5 \u03c3\u03c7\u03ad\u03c3\u03b7 \u03bc\u03b5 \u03c4\u03bf training design \u03ba\u03b1\u03b9 \u03bb\u03b5\u03b9\u03c4\u03bf\u03c5\u03c1\u03b3\u03b5\u03af \u03c9\u03c2 epistemic inflation scale. \u0397 \u03b1\u03be\u03b9\u03bf\u03c0\u03b9\u03c3\u03c4\u03af\u03b1 \u03c4\u03bf\u03c5 \u03b5\u03be\u03b1\u03c1\u03c4\u03ac\u03c4\u03b1\u03b9 \u03b1\u03c0\u03cc \u03c4\u03b1 learned features \u03ba\u03b1\u03b9 \u03b4\u03b5\u03bd \u03b5\u03af\u03bd\u03b1\u03b9 \u03ba\u03b1\u03b8\u03bf\u03bb\u03b9\u03ba\u03cc\u03c2 OOD detector.<\/p>\n<\/div>\n<\/details>\n<details class=\"td-faq-item\">\n<summary class=\"td-faq-title\">\u03a4\u03b9 \u03b5\u03b3\u03b3\u03c5\u03ce\u03bd\u03c4\u03b1\u03b9 \u03c4\u03b1 95% prediction tubes;<\/summary>\n<div class=\"td-faq-content\">\n<p>Finite-sample marginal validity \u03b3\u03b9\u03b1 \u03c4\u03bf conformal score distribution \u03c5\u03c0\u03cc exchangeability. \u0394\u03b5\u03bd \u03b5\u03b3\u03b3\u03c5\u03ce\u03bd\u03c4\u03b1\u03b9 \u03b1\u03c5\u03c4\u03cc\u03bc\u03b1\u03c4\u03b1 conditional coverage \u03c3\u03b5 \u03ba\u03ac\u03b8\u03b5 \u03c5\u03c0\u03bf\u03bf\u03bc\u03ac\u03b4\u03b1 \u03bf\u03cd\u03c4\u03b5 \u03cc\u03c4\u03b9 \u03bf\u03bb\u03cc\u03ba\u03bb\u03b7\u03c1\u03b7 \u03b7 \u03c4\u03c1\u03bf\u03c7\u03b9\u03ac \u03ba\u03b1\u03bb\u03cd\u03c0\u03c4\u03b5\u03c4\u03b1\u03b9 \u03c4\u03b1\u03c5\u03c4\u03cc\u03c7\u03c1\u03bf\u03bd\u03b1.<\/p>\n<\/div>\n<\/details>\n<details class=\"td-faq-item\">\n<summary class=\"td-faq-title\">\u03a3\u03b5 \u03c0\u03bf\u03b9\u03b1 \u03b4\u03b5\u03b4\u03bf\u03bc\u03ad\u03bd\u03b1 \u03b1\u03be\u03b9\u03bf\u03bb\u03bf\u03b3\u03ae\u03b8\u03b7\u03ba\u03b5 \u03c4\u03bf SPARC;<\/summary>\n<div class=\"td-faq-content\">\n<p>\u03a3\u03b5 Human3.6M, AMASS, LaFAN1, CMU-MoCap, 3DPW, CHICO, HA4M \u03ba\u03b1\u03b9 AnDy, \u03bf\u03c1\u03b3\u03b1\u03bd\u03c9\u03bc\u03ad\u03bd\u03b1 \u03c3\u03b5 \u03b5\u03bd\u03bd\u03ad\u03b1 dataset\/protocol blocks.<\/p>\n<\/div>\n<\/details>\n<details class=\"td-faq-item\">\n<summary class=\"td-faq-title\">\u0389\u03c4\u03b1\u03bd \u03c0\u03c1\u03ce\u03c4\u03bf \u03c3\u03b5 \u03ba\u03ac\u03b8\u03b5 metric;<\/summary>\n<div class=\"td-faq-content\">\n<p>\u038c\u03c7\u03b9. \u0395\u03af\u03c7\u03b5 mean rank 1,00 \u03c3\u03c4\u03bf NLL \u03ba\u03b1\u03b9 2,69 \u03c3\u03c4\u03bf combined MPJPE+NLL, \u03b1\u03bb\u03bb\u03ac mean rank 4,39 \u03c3\u03c4\u03bf MPJPE, \u03ac\u03c1\u03b1 \u03b4\u03b5\u03bd \u03ae\u03c4\u03b1\u03bd uniform point-accuracy winner.<\/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 \u03ba \u03bd\u03b1 \u03b5\u03bd\u03b5\u03c1\u03b3\u03bf\u03c0\u03bf\u03b9\u03b5\u03af fallback;<\/summary>\n<div class=\"td-faq-content\">\n<p>\u039c\u03c0\u03bf\u03c1\u03b5\u03af \u03bd\u03b1 \u03c7\u03c1\u03b7\u03c3\u03b9\u03bc\u03bf\u03c0\u03bf\u03b9\u03b7\u03b8\u03b5\u03af \u03c9\u03c2 lightweight risk-ranking signal \u03bc\u03b5 thresholds \u03b1\u03c0\u03cc held-out data. \u0397 \u03bc\u03b5\u03bb\u03ad\u03c4\u03b7 \u03b4\u03b5\u03bd \u03b4\u03af\u03bd\u03b5\u03b9 formal guarantee \u03cc\u03c4\u03b9 \u03ba\u03ac\u03b8\u03b5 high-\u03ba \u03c0\u03b5\u03c1\u03af\u03c0\u03c4\u03c9\u03c3\u03b7 \u03b8\u03b1 \u03b1\u03c0\u03bf\u03c4\u03cd\u03c7\u03b5\u03b9.<\/p>\n<\/div>\n<\/details>\n<details class=\"td-faq-item\">\n<summary class=\"td-faq-title\">\u03a0\u03bf\u03b9\u03bf\u03c2 \u03b5\u03af\u03bd\u03b1\u03b9 \u03bf \u03b2\u03b1\u03c3\u03b9\u03ba\u03cc\u03c2 \u03c0\u03b5\u03c1\u03b9\u03bf\u03c1\u03b9\u03c3\u03bc\u03cc\u03c2 \u03c0\u03c1\u03b9\u03bd \u03b1\u03c0\u03cc production \u03c7\u03c1\u03ae\u03c3\u03b7;<\/summary>\n<div class=\"td-faq-content\">\n<p>\u0397 conformal \u03b5\u03b3\u03b3\u03cd\u03b7\u03c3\u03b7 \u03b2\u03b1\u03c3\u03af\u03b6\u03b5\u03c4\u03b1\u03b9 \u03c3\u03c4\u03b7\u03bd exchangeability \u03ba\u03b1\u03b9 \u03c4\u03bf epistemic signal \u03c3\u03c4\u03b7 learned representation. Distribution shift, temporal dependence \u03ba\u03b1\u03b9 \u03ba\u03b1\u03ba\u03ce\u03c2 \u03b5\u03c5\u03b8\u03c5\u03b3\u03c1\u03b1\u03bc\u03bc\u03b9\u03c3\u03bc\u03ad\u03bd\u03b1 features \u03b1\u03c0\u03b1\u03b9\u03c4\u03bf\u03cd\u03bd \u03bd\u03ad\u03b1 \u03b5\u03bc\u03c0\u03b5\u03b9\u03c1\u03b9\u03ba\u03ae \u03b5\u03c0\u03b9\u03ba\u03cd\u03c1\u03c9\u03c3\u03b7.<\/p>\n<\/div>\n<\/details>\n<\/div>\n<\/section>\n<div class=\"td-source-list\">\n<p id=\"piges\" class=\"td-source-list-title\">\u0418\u0437\u0442\u043e\u0447\u043d\u0438\u0446\u0438<\/p>\n<ul>\n<li><a href=\"https:\/\/arxiv.org\/abs\/2608.20802\" target=\"_blank\" rel=\"noopener\">arXiv \u2014 SPARC: Single-Pass Scaling for Motion Forecasting with Conformal Bayesian Last Layers<\/a><\/li>\n<li><a href=\"https:\/\/arxiv.org\/abs\/2207.01567\" target=\"_blank\" rel=\"noopener\">arXiv \u2014 Back to MLP: A Simple Baseline for Human Motion Prediction<\/a><\/li>\n<li><a href=\"https:\/\/arxiv.org\/abs\/2208.07308\" target=\"_blank\" rel=\"noopener\">arXiv \u2014 Pose Forecasting in Industrial Human-Robot Collaboration<\/a><\/li>\n<li><a href=\"https:\/\/arxiv.org\/abs\/1905.03222\" target=\"_blank\" rel=\"noopener\">arXiv \u2014 Conformalized Quantile Regression<\/a><\/li>\n<\/ul>\n<\/div>","protected":false},"excerpt":{"rendered":"<p>\u03a4\u03bf SPARC \u03c3\u03c5\u03bd\u03b4\u03c5\u03ac\u03b6\u03b5\u03b9 Bayesian last layer, structured covariance \u03ba\u03b1\u03b9 conformal calibration \u03b3\u03b9\u03b1 \u03b3\u03c1\u03ae\u03b3\u03bf\u03c1\u03b7 \u03c0\u03c1\u03cc\u03b2\u03bb\u03b5\u03c8\u03b7 \u03b1\u03bd\u03b8\u03c1\u03ce\u03c0\u03b9\u03bd\u03b7\u03c2 \u03ba\u03af\u03bd\u03b7\u03c3\u03b7\u03c2 \u03bc\u03b5 \u03b5\u03bb\u03ad\u03b3\u03be\u03b9\u03bc\u03b1 \u03cc\u03c1\u03b9\u03b1.<\/p>","protected":false},"author":1,"featured_media":90933,"comment_status":"open","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"_gspb_post_css":"","content-type":"","footnotes":""},"categories":[199],"tags":[19887,19886,19885,17395,19884],"class_list":["post-90929","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-techniti-noimosyni","tag-bayesian-ai","tag-conformal-prediction","tag-human-motion-forecasting","tag-robotics","tag-sparc"],"blocksy_meta":{"styles_descriptor":{"styles":{"desktop":"","tablet":"","mobile":""},"google_fonts":[],"version":7}},"_links":{"self":[{"href":"https:\/\/twodots.gr\/bg\/wp-json\/wp\/v2\/posts\/90929","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/twodots.gr\/bg\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/twodots.gr\/bg\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/twodots.gr\/bg\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/twodots.gr\/bg\/wp-json\/wp\/v2\/comments?post=90929"}],"version-history":[{"count":1,"href":"https:\/\/twodots.gr\/bg\/wp-json\/wp\/v2\/posts\/90929\/revisions"}],"predecessor-version":[{"id":90934,"href":"https:\/\/twodots.gr\/bg\/wp-json\/wp\/v2\/posts\/90929\/revisions\/90934"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/twodots.gr\/bg\/wp-json\/wp\/v2\/media\/90933"}],"wp:attachment":[{"href":"https:\/\/twodots.gr\/bg\/wp-json\/wp\/v2\/media?parent=90929"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/twodots.gr\/bg\/wp-json\/wp\/v2\/categories?post=90929"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/twodots.gr\/bg\/wp-json\/wp\/v2\/tags?post=90929"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}