BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//pretalx//pretalx.com//pydata-amsterdam2026//speaker//9CWPBW
BEGIN:VTIMEZONE
TZID:CET
BEGIN:STANDARD
DTSTART:20001029T040000
RRULE:FREQ=YEARLY;BYDAY=-1SU;BYMONTH=10
TZNAME:CET
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
END:STANDARD
BEGIN:DAYLIGHT
DTSTART:20000326T030000
RRULE:FREQ=YEARLY;BYDAY=-1SU;BYMONTH=3
TZNAME:CEST
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
END:DAYLIGHT
END:VTIMEZONE
BEGIN:VEVENT
UID:pretalx-pydata-amsterdam2026-MDCKAF@pretalx.com
DTSTART;TZID=CET:20260910T111500
DTEND;TZID=CET:20260910T114500
DESCRIPTION:Most ML practitioners working on ranking have had this experien
 ce: the model looks good on paper - AUC\, precision are all healthy - but 
 something feels off when the scores drive real decisions.\n\nThe underlyin
 g cause is often feature neglect. In supervised pointwise ranking\, where 
 a model predicts a relevance score for each item-context pair independentl
 y\, learners frequently over-rely on contextual signals at the expense of 
 item-specific features. This may yield decent classification metrics\, but
  it leads to poorly calibrated scores - and in practice\, we rarely stop a
 t ranking. Model outputs feed downstream tasks: thresholds\, priority queu
 es\, risk tiers\, decision triggers. That's where miscalibration stops bei
 ng a metric problem and becomes a decision-making problem. 	\n\nIn this ta
 lk\, we present architectural interventions to address feature neglect dir
 ectly: item-aware attention mechanisms\, cross-feature interactions\, and 
 feature representation enhancements that explicitly strengthen item-contex
 t relationships. Through real-world experiments\, we show these changes no
 t only improve ranking performance but also produce better-calibrated scor
 es - outputs you can trust beyond the leaderboard.\n\n\nThe talk is target
 ed at Data/ML Scientists and ML Engineers working on recommendation or ran
 king systems who want to move towards more calibrated and robust ranking a
 rchitectures.
DTSTAMP:20260710T150504Z
LOCATION:Main stage
SUMMARY:The Context Trap: Addressing Item Neglect and Calibration in Deep P
 oint-Wise Rankers - Belle Bruinsma\, Akhila Vangara\, Laura Israel
URL:https://pretalx.com/pydata-amsterdam2026/talk/MDCKAF/
END:VEVENT
END:VCALENDAR