Open access
Aug 2026
Associative vs. Distributional: Two Regimes of Backdoor Learning in LoRA-Adapted Code-Generation Models
The results argue against a one-size-fits-all approach to adapter auditing: detection and removal strategies calibrated to one trigger modality can fail outright against the other, and the conditions under which each applies are outlined.
Sai Kiran Chillimuntha, Amrutha Gowri Jayasimha Hanumesh, Jeong Yang
· Journal of Cybersecurity and... · 0 citations