BEGIN:VCALENDAR VERSION:2.0 PRODID:IEEE vTools.Events//EN CALSCALE:GREGORIAN BEGIN:VTIMEZONE TZID:America/New_York BEGIN:DAYLIGHT DTSTART:20260308T030000 TZOFFSETFROM:-0500 TZOFFSETTO:-0400 RRULE:FREQ=YEARLY;BYDAY=2SU;BYMONTH=3 TZNAME:EDT END:DAYLIGHT BEGIN:STANDARD DTSTART:20261101T010000 TZOFFSETFROM:-0400 TZOFFSETTO:-0500 RRULE:FREQ=YEARLY;BYDAY=1SU;BYMONTH=11 TZNAME:EST END:STANDARD END:VTIMEZONE BEGIN:VEVENT DTSTAMP:20260710T154222Z UID:87C49339-8005-40D7-9420-59CE3B0C26BD DTSTART;TZID=America/New_York:20260709T110000 DTEND;TZID=America/New_York:20260709T120000 DESCRIPTION:Networks That Ask for Help: Uncertainty\, guidance\, and trust on the road to AI-native 6G\n\nZero-touch\, self-configuring networks are a defining ambition of AI-native 6G. This talk explores a pragmatic path t oward that goal\, built on two simple design principles\, namely (1) AI-al gorithms complementarity: learned and classical components achieve more wh en composed together\, each supplying what the other lacks (as when a clas sical solver draws global guidance from a language model while lending it algorithmic guarantees in return)\; and (2) calibrated escalation: systems to quantify their own uncertainty and ask (an oracle for labels\, or a hu man operator for clarification) when that uncertainty warrants it\, and le ss and less as confidence is earned. We develop these principles through t hree recent works from our NetAI group. First\, an active machine learning framework for 6G\, in which the network queries an oracle for the labels and samples it is most uncertain about\, cutting data acquisition and anno tation costs while improving mmWave throughput prediction. Second\, an LLM -aided A* algorithm for non-geometric network graphs\, where a classical s olver asks a language model for guidance\, thereby reducing search explora tion while keeping imperfect guidance harmless. Third\, an uncertainty-dri ven intent translation system\, in which a fine-tuned LLM turns operator i ntents into configurations and uses two tiers of uncertainty\, namely pred ictive uncertainty to decide whether to deploy or escalate\, and token-lev el entropy to pinpoint what clarifying questions to ask. These threads rel y on our open-source NWDAF-enabled 5G testbed\, which extends Free5GC with event-driven analytics and an LLM/RAG agentic intent interface\, groundin g the vision in a system that runs today. Together\, they suggest a design philosophy for trustworthy network autonomy: (1) compose generative model s with classical guarantees\, (2) quantify what the system does not know.\ n\nCo-sponsored by: Western Optimized Computing and Communications\n\nSpea ker(s): Dr. Omar Alhussein \n\nRoom: 3102\, Bldg: SEB\, Western Engineeri ng\, London\, Ontario\, Canada LOCATION:Room: 3102\, Bldg: SEB\, Western Engineering\, London\, Ontario\, Canada ORGANIZER:Abdallah.Shami@uwo.ca SEQUENCE:11 SUMMARY:Networks That Ask for Help: Uncertainty\, guidance\, and trust on t he road to AI-native 6G URL;VALUE=URI:https://events.vtools.ieee.org/m/566849 X-ALT-DESC:Description: <br /><p style="margin: 0in\;"><strong><span style= "font-size: 10.0pt\; font-family: 'Helvetica Neue'\,serif\; color: black\; ">Networks That Ask for Help: Uncertainty\, guidance\, and trust on the ro ad to AI-native 6G</span></strong></p>\n<p style="margin: 0in\;"><span sty le="font-size: 10.0pt\; font-family: 'Helvetica Neue'\,serif\; color: blac k\;">Zero-touch\, self-configuring networks are a defining ambition of AI- native 6G. This talk explores a pragmatic path toward that goal\, built on two simple design principles\, namely (1) <em>AI-algorithms complementari ty</em>: learned and classical components achieve more when composed toget her\, each supplying what the other lacks (as when a classical solver draw s global guidance from a language model while lending it algorithmic guara ntees in return)\; and (2) <em>calibrated escalation</em>: systems to quan tify their own uncertainty and ask (an oracle for labels\, or a human oper ator for clarification) when that uncertainty warrants it\, and less and l ess as confidence is earned. We develop these principles through three rec ent works from our NetAI group. First\, an active machine learning framewo rk for 6G\, in which the network queries an oracle for the labels and samp les it is most uncertain about\, cutting data acquisition and annotation c osts while improving mmWave throughput prediction. Second\, an LLM-aided A * algorithm for non-geometric network graphs\, where a classical solver as ks a language model for guidance\, thereby reducing search exploration whi le keeping imperfect guidance harmless. Third\, an uncertainty-driven inte nt translation system\, in which a fine-tuned LLM turns operator intents i nto configurations and uses two tiers of uncertainty\, namely predictive u ncertainty to decide whether to deploy or escalate\, and token-level entro py to pinpoint what clarifying questions to ask. These threads rely on our open-source NWDAF-enabled 5G testbed\, which extends Free5GC with event-d riven analytics and an LLM/RAG agentic intent interface\, grounding the vi sion in a system that runs today. Together\, they suggest a design philoso phy for trustworthy network autonomy: (1) compose generative models with c lassical guarantees\, (2) quantify what the system does not know.</span></ p> END:VEVENT END:VCALENDAR