BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//CERN//INDICO//EN
BEGIN:VEVENT
SUMMARY:Poravnava in prilagajanje velikih jezikovnih modelov po meri upora
 bnika | Alignment and adaptation of large language models to user needs
DTSTART:20261008T080000Z
DTEND:20261008T120000Z
DTSTAMP:20260930T132000Z
UID:indico-event-3996@indico.ijs.si
CONTACT:events@slaif.si
DESCRIPTION:Speakers: Boshko Koloski (Jožef Stefan Institute)\, Matej Mar
 tinc (Jožef Stefan Institute)\n\nIzvajalec / Course provider: Jožef Stef
 an Institute (JSI)\nPredavatelji / Instructors: Matej Martinc (JSI)\, Bos
 hko Koloski (JSI)\nLearning objectives: Gain practical knowledge of Large 
 Language Model (LLM) alignment using reinforcement learning methods to ens
 ure safety\, reduce bias\, and follow complex user instructions.\nCourse c
 ontent: \nFor the successful implementation of LLMs in business and scient
 ific environments\, models require more than just knowledge\; they need al
 ignment with ethical standards and specific user preferences. This practic
 al workshop will cover:\n\nFundamentals of Alignment: Why simple next-toke
 n prediction is not enough and how to guide models toward helpfulness\, ho
 nesty\, and harmlessness.\nReinforcement Learning from Human Feedback (RLH
 F): An overview of algorithms (e.g.\, PPO\, DPO\, GRPO) for fine-tuning mo
 dels (such as Gemma and GaMS).\nDebiasing: Techniques for identifying and 
 reducing stereotypes and biases in generated responses.\nAdaptation to use
 r needs: How to use the alignment process to ensure a model adheres to the
  specific tone\, style\, or professional constraints of a particular organ
 ization.\n\n \nLearning outcomes:\n\nUnderstanding the process of model a
 lignment using reinforcement learning.\nThe ability to implement technique
 s to reduce bias in model responses.\nPractical knowledge of adapting mode
 l behavior to the specific requirements of a client or research project.\n
 \n\nhttps://indico.ijs.si/event/3996/
IMAGE;VALUE=URI:https://indico.ijs.si/event/3996/logo-3560331715.png
LOCATION:predavalnica Elektronike na Teslovi 30  (Teslova ulica 30 )
URL:https://indico.ijs.si/event/3996/
END:VEVENT
END:VCALENDAR