TY - GEN

T1 - Workshop on Large Language Models' Interpretability and Trustworthiness

T2 - 32nd ACM International Conference on Information and Knowledge Management, CIKM 2023

AU - Saha, Tulika

AU - Saha, Sriparna

AU - Ganguly, Debasis

AU - Mitra, Prasenjit

N1 - Funding information: (4) Prasenjit Mitra is a Professor at The Pennsylvania State Uni-versity and a visiting Professor at the L3S Center at the Leibniz University at Hannover, Germany. He obtained his Ph.D. from Stanford University in 2003 in Electrical Engineering and has been at Penn State since. His research interests are in artifi-cial intelligence, applied machine learning, natural language processing, etc. His research has been supported by the NSF CAREER award, the DoE, DoD, Microsoft Research, Raytheon, Lockheed Martin, Dow Chemicals, McDonnell Foundation, etc. His has published over 200 peer-reviewed papers at top con-ferences and journals, supervised or co-supervised 15-20 Ph.D. dissertations; his work has been widely cited (h-index 60) and over 12,500 citations. Along with his co-authors, he has won the test of time award at the IEEE VIS and a best paper award at ISCRAM, etc. He has been the co-chair of several workshops, including a workshop previously collocated with CIKM. They are listed below: • Program Chair, Big-O(Q)’15: Workshop on Big-Graphs Online Querying in VLDB’15: the 41st International Conference on Very Large Databases (2015). • Program Chair, WIDM’12: The 12th International Workshop on Web Information and Data Management in CIKM’12: the 21st ACM International Conference on Information and Knowledge Management. (2012). • Program Co-Chair, WIDM’09: The 11th International Work-shop on Web Information and Data Management in CIKM’09: the 18th ACM International Conference on Information and Knowledge Management. (2009). • Program Co-Chair, SNAKDD’09: The 2nd International Work-shop on Social Network Mining and Analysis in KDD’08: the 14th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. (2008).

PY - 2023/10/21

Y1 - 2023/10/21

N2 - Large language models (LLMs), when scaled from millions to billions of parameters, have been demonstrated to exhibit the so-called 'emergence' effect, in that they are not only able to produce semantically correct and coherent text, but are also able to adapt themselves surprisingly well with small changes in contexts supplied as inputs (commonly called prompts). Despite producing semantically coherent and potentially relevant text for a given context, LLMs are vulnerable to yield incorrect information. This misinformation generation, or the so-called hallucination problem of an LLM, gets worse when an adversary manipulates the prompts to their own advantage, e.g., generating false propaganda to disrupt communal harmony, generating false information to trap consumers with target consumables etc. Not only does the consumption of an LLM-generated hallucinated content by humans pose societal threats, such misinformation, when used as prompts, may lead to detrimental effects for in-context learning (also known as few-shot prompt learning). With reference to the above-mentioned problems of LLM usage, we argue that it is necessary to foster research on topics related to not only identifying misinformation from LLM-generated content, but also to mitigate the propagation effects of this generated misinformation on downstream predictive tasks thus leading to more robust and effective leveraging in-context learning.

AB - Large language models (LLMs), when scaled from millions to billions of parameters, have been demonstrated to exhibit the so-called 'emergence' effect, in that they are not only able to produce semantically correct and coherent text, but are also able to adapt themselves surprisingly well with small changes in contexts supplied as inputs (commonly called prompts). Despite producing semantically coherent and potentially relevant text for a given context, LLMs are vulnerable to yield incorrect information. This misinformation generation, or the so-called hallucination problem of an LLM, gets worse when an adversary manipulates the prompts to their own advantage, e.g., generating false propaganda to disrupt communal harmony, generating false information to trap consumers with target consumables etc. Not only does the consumption of an LLM-generated hallucinated content by humans pose societal threats, such misinformation, when used as prompts, may lead to detrimental effects for in-context learning (also known as few-shot prompt learning). With reference to the above-mentioned problems of LLM usage, we argue that it is necessary to foster research on topics related to not only identifying misinformation from LLM-generated content, but also to mitigate the propagation effects of this generated misinformation on downstream predictive tasks thus leading to more robust and effective leveraging in-context learning.

KW - Explainability

KW - In-context Learning

KW - Interpretability

KW - Large Language Model

KW - Trustworthiness

UR - http://www.scopus.com/inward/record.url?scp=85178112362&partnerID=8YFLogxK

U2 - 10.1145/3583780.3615311

DO - 10.1145/3583780.3615311

M3 - Conference contribution

AN - SCOPUS:85178112362

T3 - International Conference on Information and Knowledge Management, Proceedings

SP - 5290

EP - 5293

BT - CIKM 2023

PB - Association for Computing Machinery (ACM)

Y2 - 21 October 2023 through 25 October 2023

ER -