Abstract
An example system includes at least one processor configured to progressively amplify discriminative signals between benign and jailbreak prompts across layer, module, and token levels and employ two independent classifiers that jointly predict jailbreak status based on these amplified representations.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Chen, Yuzhong; Pan, Menghai; and Fan, Xiran, "SYSTEM, METHOD, AND COMPUTER PROGRAM PRODUCT FOR AMPLIFICATION DETECTION AGAINST JAILBREAK ATTACK", Technical Disclosure Commons, (September 29, 2026)
https://www.tdcommons.org/dpubs_series/11907
COinS