About The Workshop
The striding advances of large language models (LLMs) are revolutionizing many long-standing natural language processing tasks ranging from machine translation to question-answering and dialog systems. However, as LLMs are often built upon massive amounts of text data and subsequently applied in a variety of downstream tasks, building, deploying and operating LLMs entails profound security and trustworthiness challenges, which have attracted intensive research efforts in recent years.
Call For Papers
The primary aim of the proposed workshop is to identify such emerging challenges, discuss novel solutions to address them, and explore new perspectives and constructive views across the full theory/algorithm/application stack.
Topics
The potential topics include but are not limited to- Reliability assurance and assessment of LLMs
- Privacy leakage issues of LLMs
- Copyright protection
- Interpretability of LLMs
- Plagiarism detection and prevention
- Security of LLM deployment
- Backdoor attacks and defenses in LLMs
- Adversarial attacks and defenses in LLMs
- Toxic speech detection and mitigation
- Challenges in new learning paradigms of LLMs (e.g., prompt engineering)
- Fact verification (e.g. hallucinated generation)
Submission
All papers can be submitted through OpenReview through OpenReview: https://openreview.net/group?id=ICLR.cc/2024/Workshop/SeT_LLM
Important Dates
In response to the requests for an extension of the submission deadline and and in consideration of the Lunar Spring Festival, we have decided to extend the deadline for workshop submissions to February 19th.
- Submission Open: Jan 15
- Submission Deadline:
Feb 12Feb 19 - Final Decision Notification: Mar 3
- Camera Ready Deadline:
Apr 3Apr 12
All time are in UTC+0.
Format
Please format your submissions with ICLR 2024 LaTeX style file. The review process for this workshop is double-blinded. Please anonymize your submissions and remove any links that may reveal your identity. Submissions are limited to 4 pages for main contents with unlimited reference and appendix pages. The accepted submissions are allowed with 1 additional page (5 pages in total for main contents) for the camera ready version.
Policies
Submissions that are concurrently under review at other venues are acceptable. All accepted papers are non-archival, and will be made publicly available at Openreview without an official proceeding and reviews. For any questions, please contact us at set-llm-admin@googlegroups.com.
Reviewer recruiment
If you are interested in reviewing submissions, please fill out this form.
Invited Speakers
Tatsunori Hashimoto
Stanford
Graham Neubig
CMU
Bo Li
University of Chicago
Robin Jia
USC
Tom Goldstein
University of Maryland
Chaowei Xiao
University of Wisconsin, Madison
Eric Wallace
OpenAI
Event Schedule
Opening Remarks
Invited Talk 1
Tatsu Hashimoto
Oral Paper Presentation 1
On Prompt-Driven Safeguarding for Large Language ModelsChujie Zheng, Fan Yin, Hao Zhou, Fandong Meng, Jie Zhou, Kai-Wei Chang, Minlie Huang, Nanyun Peng
Oral Paper Presentation 2
Explorations of Self-Repair in Language ModelCody Rushing, Neel Nanda
Invited Talk 2
Graham Neubig
Oral Paper Presentation 3
TOFU: A Task of Fictitious Unlearning for LLMsPratyush Maini, Zhili Feng, Avi Schwarzschild, Zachary Chase Lipton, J Zico Kolter
Oral Paper Presentation 4
Are Large Language Models Bayesian? A Martingale Perspective on In-Context LearningFabian Falck, Ziyu Wang, Christopher C. Holmes
Poster Session A
For all accepted papersLunch Break
Invited Talk 3
Bo Li
Invited Talk 4
Robin Jia
Invited Talk 5
Tom Goldstein
Invited Talk 6
Chaowei Xiao
Invited Talk 7
Eric Wallace
Oral Paper Presentation 5
How Susceptible are Large Language Models to Ideological Manipulation?Kai Chen, Zihao He, Jun Yan, Taiwei Shi, Kristina Lerman
Oral Paper Presentation 6
Assessing the Brittleness of Safety Alignment via Pruning and Low-Rank ModificationsBoyi Wei, Kaixuan Huang, Yangsibo Huang, Tinghao Xie, Xiangyu Qi, Mengzhou Xia, Prateek Mittal, Mengdi Wang, Peter Henderson
Poster Session B
For all accepted papersClosing Remarks
Organizers
This workshop is organized by
Yisen Wang
Peking University
Ting Wang
Stony Brook University
Jinghui Chen
Penn State University
Chaowei Xiao
University of Wisconsin, Madison
Jieyu Zhao
USC
Nanyun Peng
UCLA
Anima Anandkumar
Caltech