Speaker: Subho Majumdar (Head of AI at VIJIL, a US-based startup)
Host: Radhendushka Srivastava
Date: 8 Aug 2025
Time: 3:00 to 4:00 pm
Venue: Ramanujan Hall
Title: Towards Statistical Foundations for Reliable and Defendable Large
Language Models
Abstract: The emergence of Large Language Models (LLMs) has brought in
concomitant concerns about the security and reliability of generative AI
systems. While LLMs promise powerful capabilities in diverse real-world
applications, ensuring that their outputs are resilient to malicious
attacks and consistent across similar inputs has significant methodological
and computational challenges. This situation calls for the revisiting of
modern deep learning architectures through a statistical lens.
I will present on two interconnected themes in this area. First, I will
introduce Representation Noising (RepNoise), a defense mechanism that
protects the weights of open-source LLMs against malicious uses. RepNoise
achieves this through controlled noise injection in the knowledge
representations inside a model that makes it harder to recover harmful
information later. Second, I will discuss my work on the consistency
problem—the equivalent of robustness in LLMs concerned with measuring and
minimizing the sensitivity of LLM outputs to input variations through a
combination of controlled synthetic data generation and fine-tuning.
I will conclude by discussing ongoing work at the intersection of AI
security and statistics, including the development of statistical bounds
for the strength of defense mechanisms like RepNoise, and robustness
frameworks for ensuring AI system reliability in high-stakes applications.