Indian American Professor's AI Research Accepted at Premier Conference

The University of Arkansas at Little Rock professor's work focuses on improving the resilience of AI models against malicious inputs. The conference, known as NeurIPS, is scheduled to take place in Vancouver, Canada, in December 2025.

Agarwal's research proposes a novel framework that enhances the ability of neural networks to maintain accuracy when faced with adversarial examples. These are inputs deliberately designed to fool the model, a critical concern for applications like autonomous vehicles and medical diagnostics.

The paper, co-authored with two doctoral students from the university, introduces a technique that combines data augmentation with a new training loss function. The approach reportedly reduced error rates by up to 15% in benchmark tests compared to existing methods.

Agarwal has been a faculty member at UA Little Rock since 2012 and leads the Collaborative Social and Information Dynamics Lab. His previous work has been published in journals such as IEEE Transactions on Knowledge and Data Engineering and ACM Transactions on Intelligent Systems and Technology.