BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//pretalx//pretalx.com//compute-paris-2026//talk//ZMVL9B
BEGIN:VTIMEZONE
TZID:Europe/Paris
BEGIN:STANDARD
DTSTART:20251125T000000
TZNAME:CET
TZOFFSETFROM:+0100
TZOFFSETTO:+0100
END:STANDARD
BEGIN:DAYLIGHT
DTSTART:20260329T030000
RDATE:20270328T030000
TZNAME:CEST
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
END:DAYLIGHT
BEGIN:STANDARD
DTSTART:20261025T030000
RDATE:20271031T030000
TZNAME:CET
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
SUMMARY:Stop Guessing\, Start Measuring. Rigorous Evaluation of Agentic Sy
 stems with Prediction-Powered Inference - Grégoire Martinon
DTSTART;TZID=Europe/Paris:20261125T120000
DTEND;TZID=Europe/Paris:20261125T123000
DTSTAMP:20260930T122033Z
UID:pretalx-compute-paris-2026-ZMVL9B@pretalx.com
DESCRIPTION:Agentic systems are booming\, yet few reach production. A prim
 ary blocker is performance evaluation: LLM-as-Judge is cheap but biased\, 
 while human annotation doesn't scale. The result is a familiar gap\, autom
 ated metrics that look great on paper\, and business experts who remain un
 convinced.\nThis talk introduces GLIDE (Generated Label Inference and Debi
 asing Engine)\, an open-source Python package that bridges this gap using 
 Prediction-Powered Inference. By combining a small set of human labels (~1
 00 annotations) with large volumes of proxy judgments\, GLIDE produces unb
 iased performance estimates with valid confidence intervals\,  powered by 
 state-of-the-art techniques including Prediction-Powered Inference (PPI++)
 \, Active Statistical Inference\, and a suite of samplers that optimize wh
 ere your annotation budget is spent.\nAfter this talk\, you will know what
  statistically rigorous GenAI system evaluation looks like and how to put 
 it in place\nTarget audience: ML and Software Engineers evaluating GenAI s
 ystems\, and researchers in statistical inference. No prior background req
 uired.\nAll code\, tutorials\, and examples are available in the open-sour
 ce repository.
LOCATION:Auditorium
URL:https://pretalx.com/compute-paris-2026/talk/ZMVL9B/
END:VEVENT
END:VCALENDAR