BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//pretalx//pretalx.coscup.org//coscup-2026//talk//NUPSVF
BEGIN:VTIMEZONE
TZID:CST
BEGIN:STANDARD
DTSTART:20000101T000000
RRULE:FREQ=YEARLY;BYMONTH=1
TZNAME:CST
TZOFFSETFROM:+0800
TZOFFSETTO:+0800
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
UID:pretalx-coscup-2026-NUPSVF@pretalx.coscup.org
DTSTART;TZID=CST:20260809T100500
DTEND;TZID=CST:20260809T103500
DESCRIPTION:要考期中考？AI 沒過也會有這天。開發 Agent Workf
 low 時要怎麼分辨。表現變好是 Model 的功勞，還是好的 Wo
 rkflow 帶來的功勞？這時就需要一套好的驗證方法：期中
 考！但是道高一尺，魔高一丈。你想得到的，AI 也想得
 到。上網作弊、偷看答案，只有你想不到，沒有 AI 做不
 到。本節的故事背景，是我們在開發開源 Agent Workflow 時
 ，設計內部 benchmark 所碰到的各種大小事。\n\nHow do you eva
 luate the performance of an agent workflow? When things get better\, is it
  the model doing the work\, or the workflow itself? A good benchmark is li
 ke a midterm exam. It tells you who actually studied.\nBut AI is smarter t
 han you expect. It will cheat by searching the web\, peek at other answers
 \, and find every loophole you forgot to close. AI always finds a way to s
 urprise you.\nIn this session\, we'll share the story behind building a be
 nchmark for an open-source agent workflow. Every twist\, every workaround\
 , and every time the AI quietly outsmarted our test design.
DTSTAMP:20260713T142433Z
LOCATION:AU
SUMMARY:AI 也要期中考？ The Story Behind Building a Benchmark for Age
 nt Workflow Development - Kent Huang
URL:https://pretalx.coscup.org/coscup-2026/talk/NUPSVF/
END:VEVENT
END:VCALENDAR