BEGIN:VCALENDAR
VERSION:2.0
PRODID:Data::ICal 0.24
BEGIN:VEVENT
DESCRIPTION:   'Title: FuzzLLM: A Fuzzing Framework for Discovering Jailbre
 ak\n   Vulnerabilities in Large Language Models\n   When: Saturday\, Aug 1
 0\, 15:00 - 15:59 PDT\n   Where: LVCC West/Floor 3/W314-W316 - [1]Map\n\n 
   Description:\n\n   Jailbreak vulnerabilities in Large Language Models (L
 LMs)\, which\n   exploit meticulously crafted prompts to elicit content th
 at violates\n   service guidelines\, have captured the attention of resear
 ch\n   communities. While model owners can defend against individual\n   j
 ailbreak prompts through safety training strategies\, this relatively\n   
 passive approach struggles to handle the broader category of similar\n   j
 ailbreaks. To tackle this issue\, we introduce FuzzLLM\, an automated\n   
 fuzzing framework designed to proactively test and discover jailbreak\n   
 vulnerabilities in LLMs. We utilize templates to capture the\n   structura
 l integrity of a prompt and isolate key features of a\n   jailbreak class 
 as constraints. By integrating different base classes\n   into powerful co
 mbo attacks and varying the elements of constraints\n   and prohibited que
 stions\, FuzzLLM enables efficient testing with\n   reduced manual effort.
  Extensive experiments demonstrate FuzzLLM's\n   effectiveness and compreh
 ensiveness in vulnerability discovery across\n   various LLMs.\n\n   Speak
 erBio:  Ian G. Harris\, Professor of Computer Science at\n   University of
  California Irvine\n\n   Ian G. Harris is Professor of Computer Science at
  the University of\n   California Irvine. He received his BS degree in Com
 puter Science from\n   Massachusetts Institute of Technology in 1990. He r
 eceived his MS and\n   PhD degrees in Computer Science from the University
  of California San\n   Diego in 1992 and 1997 respectively. He was a membe
 r of the faculty in\n   the Electrical and Computer Engineering Department
  at the University\n   of Massachusetts Amherst from 1997 until June 2003.
 \n\n   '\n\n   1. #LVCCW_Level3_North\n\n\n
DTEND:20240810T225900Z
DTSTART:20240810T220000Z
LOCATION:BICV - LVCC West/Floor 3/W314-W316
SUMMARY:FuzzLLM: A Fuzzing Framework for Discovering Jailbreak Vulnerabilit
 ies in Large Language Models
END:VEVENT
END:VCALENDAR