BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//pretalx//pretalx.com//euroscipy-2026//speaker//9B7HES
BEGIN:VTIMEZONE
TZID:Poland
BEGIN:DAYLIGHT
DTSTART:20250721T000000
TZNAME:CEST
TZOFFSETFROM:+0200
TZOFFSETTO:+0200
END:DAYLIGHT
BEGIN:STANDARD
DTSTART:20251026T030000
RDATE:20261025T030000
TZNAME:CET
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
END:STANDARD
BEGIN:DAYLIGHT
DTSTART:20260329T030000
RDATE:20270328T030000
TZNAME:CEST
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
END:DAYLIGHT
END:VTIMEZONE
BEGIN:VEVENT
SUMMARY:The Illusion of Compliance: Auditing LLM-as-a-Judge Systems - Vasu
  Sharma
DTSTART;TZID=Poland:20260721T144000
DTEND;TZID=Poland:20260721T151000
DTSTAMP:20260911T035710Z
UID:pretalx-euroscipy-2026-8KCT3D@pretalx.com
DESCRIPTION:LLM-as-a-Judge systems are increasingly deployed in high-stake
 s settings - screening job applicants\, triaging medical cases\, assessing
  credit risk\, and flagging legal exposure. As the EU AI Act takes effect 
 in August 2026 with penalties up to €35M for biased high-risk systems\, 
 organizations are investing heavily in fairness audits. But passing a bias
  check does not guarantee fairness. Standard Python fairness pipelines rar
 ely detect this shift. In a controlled hiring experiment on real resumes\,
  we demonstrate how alignment and potentially bias-mitigation techniques c
 an reduce aggregate disparities while redistributing harm across intersect
 ional subgroups.
LOCATION:Room 1.19 (Ground Floor\, Shannon)
URL:https://pretalx.com/euroscipy-2026/talk/8KCT3D/
END:VEVENT
BEGIN:VEVENT
SUMMARY:A Hands-On Introduction to Mechanistic Interpretability - Vasu Sha
 rma
DTSTART;TZID=Poland:20260723T090000
DTEND;TZID=Poland:20260723T103000
DTSTAMP:20260911T035710Z
UID:pretalx-euroscipy-2026-H9UDLT@pretalx.com
DESCRIPTION:Large language models (LLMs) have become central to modern sci
 entific computing\, yet for most practitioners they remain opaque systems 
 - input goes in\, text comes out\, and the internal mechanism is a mystery
 . Mechanistic interpretability (MI) is the emerging discipline of reverse-
 engineering what specific components of a neural network actually *do*.\nU
 sing Andrej Karpathy's `microgpt` - a fully self-contained\, 200-line\, de
 pendency-free GPT implementation in pure Python - as our subject\, we syst
 ematically dissect what a trained language model has learned. No PyTorch\,
  no specialised ML frameworks: just the familiar tools applied to a genuin
 ely novel problem.\n\nThe model is tiny by design: 4\,192 parameters\, a 2
 7-token vocabulary (a–z + a special token)\, trained on 32\,000 names in
  roughly one minute on a laptop. This makes it the ideal subject for inter
 pretability work - every attention weight is inspectable\, every embedding
  printable\, every head ablatable. The scientific question driving the tut
 orial is: *"What has this model actually learned about the structure of na
 mes?"*
LOCATION:Room 1.38 (Ground Floor\, Turing)
URL:https://pretalx.com/euroscipy-2026/talk/H9UDLT/
END:VEVENT
END:VCALENDAR