SOURCE-LINKED INTELLIGENCE
ENDOPROMPT: Victim-Side Pseudo-References for Utility Degradation
Prompt injection can degrade benign task performance without eliciting harmful content. Yet many attack objectives depend on task labels or predefined target responses. We present ENDOPROMPT, a white-box method that learns utility-degrading prefixes from unlabeled instructions. Its generator takes the request text as input. Clean victim continuations serve as pseudo-references: local search identifies prefixes that reduce continuation likelihood, and preference fitting on comparisons within the same instruction, followed by reward refinement, distills this signal into a generator. At deploymen
Read original source ↗ Open in workspace
- recordType
- paper
- region
- Global
Evidence & attribution
- arXiv · AI, language, vision and robotics · 2026-09-24T15:08:13.000Z
First collected: 2026-09-25T06:12:46.948Z. This is not the publication date.