Today at the IEEE COMPSAC Symposium on Software Engineering Technologies & Applications (COMPSAC-SETA 2025) Mosarrat Rumman presented “A Contrastive Learning Approach to Bug Severity Classification with Large Language Model Embeddings” (co-authors Anushka Zaman, Emon Roy, Jeremy Bradbury). This research shows that contrastive fine-tuning for LLMs can improve semantic separation and boost generalization on imbalanced, diverse bug severity datasets (NASA PITS, Mozilla).
