Agenda

20 May 2026 13:00

Addressing Bias, Fairness and Hallucination in LLMs for Heritage Computing

Sala Riunioni B - Edificio ZETA B | Campus Scientifico

Speaker:
Santanu Roy, Heritage Institute of Technology, Kolkata, West Bengal, India

Abstract:
Large Language Models often produce hallucinated responses, where generated content appears convincing but lacks factual or contextual correctness. Such issues become even more critical in domains requiring high interpretability, contextual grounding, and historical authenticity, including heritage preservation, education, healthcare, and cultural analytics. As a case study, we will examines a multimodal AI framework designed for interpreting cultural and mythological wall carvings of Terracotta art of South Asian temples through visual grounding and contextual reasoning . The discussion highlights how domain specific benchmarks, cognitive evaluation metrics, and hallucination mitigation pipelines can support the development of trustworthy AI systems for digital heritage preservation and other sensitive knowledge domains.

Bio sketch:
Santanu Roy obtained a Ph.D in Information Technology from the A.K.Choudhury School of Information Technology, University of Calcutta, India, in 2025 (Supervisor: dr. Soumya Sen, Joint-Supervisor: prof. Agostino Cortesi). From 2022 to 2024 he was head of the Deoartment of Computer Science ain the Future Institute of Engineering and Management, Kolkata. Since 20024 he's Assistant Professor in the Department of Computer Science and Business Systems of the Heritage Institute of Technology.

Language

The event will be held in English

Organized by

Tino Cortesi

Search in the agenda