MS student, Tufts University
1 paper at NeurIPS 2025
We propose BAM-ICL, a novel budgeted adversarial manipulation hijacking attack framework for in-context learning.
