Prof. Tanmoy Chakraborty of IIT Delhi

Despite superior performance demonstrated by Transformer-based LLMs across numerous applications involving natural language, their high computational cost, energy consumption, and limited accessibility underscore the need for efficient, interpretable, and adaptable small language models (SLMs). This talk highlights methods to develop economical and interpretable SLMs that rival their larger counterparts in performance without significant computational requirements. Our research emphasizes three key dimensions: economical resource usage, adaptability to diverse and low-resource tasks, and enhanced interpretability. Techniques like competitive knowledge distillation, leveraging student-teacher dynamics, and activation sparsity in manifold-preserving transformers demonstrate significant efficiency gains without compromising performance. We formulate novel decomposer components for LLMs for modularizing problem decomposition and solution generation, allowing smaller models to excel in complex reasoning tasks. We also propose innovative prompt construction and alignment strategies that boost in-context knowledge adaptation in low-resource settings for SLMs. Our findings demonstrate that SLMs can achieve scalability, interpretability, and adaptability, paving the way for broader and sustainable AI accessibility.

Biography:

Tanmoy Chakraborty is a Professor in the Dept. of Electrical Engineering and the Yardi School of AI at IIT Delhi, where he holds the Rajiv Khemani Young Faculty Chair Professorship. He leads the Laboratory for Computational Social Systems (LCS2), a research group that primarily focuses on building economical, adaptable and interpretable language models. He served as the DAAD visiting professor at MPI Saarbrücken, PECFAR visiting professor at TU Munich and Humboldt visiting professor at TU Darmstadt. Tanmoy has received numerous recognitions, including the youngest recipient of the ACM India Outstanding Contribution to Computing Education Award, INSA Young Associate, Ramanujan Fellowship, as well as several faculty awards from industries such as Microsoft, IBM, Google, LinkedIn, JP Morgan, and Adobe. He was the PC Chair of EMNLP'25 and serves as the Editor-in-Chief of ACL Rolling Review. He has authored two textbooks -- "Social Network Analysis" and "Introduction to Large Language Models"- on which he has been offering two NPTEL courses. More details may be found at tanmoychak.com.

Prof. Dipankar Dasgupta of The University of Memphis

AI Agents have been in research and practice for more than 35 years which include autonomous agents, intelligent agents, mobile agents, distributed agents, collaborative agents and some hierarchical multi-agent combinations. These agent architectures incorporated various computational intelligence techniques such as artificial intelligence, neural networks, fuzzy logic, genetic algorithms, data mining, etc. in solving application-specific tasks.

Agentic AIs based on Large Language Models (LLMs) have emerged as new frontier in user-friendly applications, capable of autonomous planning, reasoning, tool use, and adaptive decision-making for complex tasks via agent-to-agent interactions and utilize external components as and when needed.

The talk will discuss the pros and cons of both approaches in terms of safety, reliability and privacy in real-world applications, especially in critical and emerging technologies (CETs). The speaker will address the need for disclosure of agentic AI scope and capabilities, and require agentic AI governance (accountability, assurance, regulation and compliance) for successful businesses in CETs.

Biography:

Dr. Dipankar Dasgupta is a Professor of Computer Science at the University of Memphis since January 1997. He is a leading expert in the application of bio-inspired and machine learning approaches to cyber defense. His groundbreaking works, including digital immunity, negative authentication, cloud insurance model, and adaptive- authentication, have earned recognition in Computer World Magazine and other media outlets. With over 350 publications (including 4 patents), having 23000+ citations, and an h-index of 68 per Google Scholar, Dr. Dasgupta's multidisciplinary research is highly acclaimed. He has received numerous awards, including 2023 Presidential Award for Lifetime Achievement in Research, 2012 Willard R. Sparks Eminent Faculty Award and the 2014 ACM SIGEVO Impact Award. Among other awards: 2007 University’s College of Arts and Sciences awarded the Dunavant Professorship, a 3-year award for exceptional teaching, scholarship, service, and outreach achievement; twice received the College of Arts & Sciences Distinguished Research Award (2002, 2006); the Sigma Xi Research Paper Award (2003);  Early Career Research Award (1999), and Dr. Pat E. Burleson’s Professorship (2007). Dr. Dasgupta’s research has been supported through several funding sources during last 30 years.  

He also received five best paper awards at different international conferences and has been organizing Symposium on Computational Intelligence in Cyber Security at IEEE SSCI since 2007. Dr. Dasgupta is an IEEE Fellow, and NAI Fellow, an ACM Distinguished Speaker (2015-2020 and 2026-2028), an IEEE Distinguished Lecturer (2022-2024) and NSF-Fulbright Distinguished Scholar, 2024). He regularly serves as a panelist and keynote speaker and offers tutorials at leading computer science conferences and has given more than 300 invited talks in different universities and industries. 

As the founding Director of the Center for Information Assurance (CfIA) at the University of Memphis, he has been playing a key role in education, training, and outreach activities on AI based cyber security and Information Assurance since 2004. The center has continued to receive national designations such as Center of Academic Excellence in Education (CAE-CD) and Research (CAE-R). 

Prof. Danda B. Rawat of Howard University

TBA

Prof. Hussein Abbass of University of New South Wales

Recent security incidents involving agentic artificial intelligence have elicited polarized reactions, ranging from dismissals of the events as marketing stunts to apocalyptic framing as existential threats. Beneath these competing public narratives, however, a fundamental research question persists: can we trust agentic artificial intelligence? While framing a complex sociotechnical challenge as a binary question is inherently flawed, doing so provides a useful heuristic to break down the broader problem space into manageable research questions and their associated challenges. In this presentation, I will offer a systems-level perspective on this critical topic and will extend the discussion into ethical and general alignment requirements.

Biography:

Professor Hussein Abbass is a Full Professor at the School of Systems and Computing at UNSW Canberra. He is a Fellow of the IEEE, the Australian Computer Society, the UK Operational Research Society, and the Australian Institute of Managers and Leaders. Having joined UNSW Canberra in 2000 following a decade across industry and academia-and holding a full professorship since 2007-he has previously served as Deputy Head (People) of the School of Engineering & IT (2021-2022), and National President of the Australian Society for Operations Research (2016-2019). His extensive leadership within the IEEE includes serving as the Founding Editor-in-Chief of the IEEE Transactions on Artificial Intelligence (2020-2024), Chair of the IEEE Frank Rosenblatt Award Committee (2024-2025), Vice-Chair of IEEE Standard 7018, and Vice-President for Technical Activities for the IEEE Computational Intelligence Society (2016-2019).