Abstract
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 deployment, the generator produces one prefix per request without further victim-side search. Across four instruction-tuned models and the complete splits of seven benign benchmarks, ENDOPROMPT yields a mean utility change of -26.8 percentage points; 27 of 28 cells are negative. Failure analysis reveals output expansion and prefix reuse; the controls do not establish a degradation advantage from request matching. Victim-derived supervision can reveal utility weaknesses without benchmark feedback or prescribed failure responses. The code will be released upon acceptance.
Keywords
Subject
Publication details
- Journal
- Not available
- Open access
- Green open access
Cite this article
APA 7
Wu, Q., Feng, Z., Yu, Y., Luo, Y., & Cheng, H. (2026). ENDOPROMPT: Victim-Side Pseudo-References for Utility Degradation. https://omanscience.com/en/articles/endoprompt-victim-side-pseudo-references-for-utility-degradation
MLA 9
Wu, Qingyu, et al. "ENDOPROMPT: Victim-Side Pseudo-References for Utility Degradation." https://omanscience.com/en/articles/endoprompt-victim-side-pseudo-references-for-utility-degradation.
Chicago (author–date)
Wu, Qingyu, Zeyu Feng, Yongda Yu, Yuzhe Luo, and Hua Cheng. 2026. "ENDOPROMPT: Victim-Side Pseudo-References for Utility Degradation." https://omanscience.com/en/articles/endoprompt-victim-side-pseudo-references-for-utility-degradation.
Harvard
Wu, Q., Feng, Z., Yu, Y., Luo, Y. and Cheng, H. (2026) 'ENDOPROMPT: Victim-Side Pseudo-References for Utility Degradation', Available at: https://omanscience.com/en/articles/endoprompt-victim-side-pseudo-references-for-utility-degradation.
Vancouver
Wu Q, Feng Z, Yu Y, Luo Y, Cheng H. ENDOPROMPT: Victim-Side Pseudo-References for Utility Degradation. https://omanscience.com/en/articles/endoprompt-victim-side-pseudo-references-for-utility-degradation
IEEE
Q. Wu, Z. Feng, Y. Yu, Y. Luo, and H. Cheng, "ENDOPROMPT: Victim-Side Pseudo-References for Utility Degradation," https://omanscience.com/en/articles/endoprompt-victim-side-pseudo-references-for-utility-degradation.