2026 08 21

Created in August 21, 2026

2026

Two papers are accepted to EMNLP 2026:

  • CODE: A Contradiction-Based Overthinking Attack Framework for Prolonging Retrieval-Augmented Generation. Congratulations to Xiaolei!
  • Attributing and Exploiting Safety Vectors through Global Optimization in Large Language Models. Congratulations to Fengheng!