Ignore This Title and HackAPrompt: Exposing Systemic Vulnerabilities of LLMs Through a Global Prompt Hacking Competition
Open accessAbstract
Sander Schulhoff, Jeremy Pinto, Anaum Khan, Louis-François Bouchard, Chenglei Si, Svetlina Anati, Valen Tagliabue, Anson Kost, Christopher Carnahan, Jordan Boyd-Graber. Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing. 2023.
Integrity signals
Registry
Cites
Statistics
Claims the paper makes
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
Existing safety mechanisms leave systems vulnerable to sophisticated exploits that a static rule set cannot anticipate.
“This leaves systems vulnerable to sophisticated exploits that a static rule set cannot anticipate (Schulhoff et al., 2023).”
From VeriGuard: Enhancing LLM Agent Safety via Verified Code Generation (2025) · cites this paper as Schulhoff et al., 2023
The citation lists haven’t been fetched yet
References 36 · Cited by 54
- Authors
- Schulhoff, S., Pinto, J., Khan, A., Bouchard, L.-F., Si, C., Anati, S., Tagliabue, V., Kost, A., Carnahan, C., & Boyd-Graber, J.
- Affiliations
- University of Maryland, College Park (US)
- Stanford University (US)
- Technical University of Sofia (BG)
- University of Milan (IT)
- New York University (US)
- University of Arizona (US)
- Corresponding author
- Schulhoff, Sander
- ORCID
- Boyd-Graber, Jordan
- Venue
- Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing
- Published
- 2023
- Pages
- 4945-4977
- OpenAlex
- W4389523893
- Cited by
- 54
- References
- 36
- Citations by year
- 2026: 27 · 2025: 21 · 2024: 5 · 2023: 1
- FWCI
- 9.57
- Type
- conference paper
- Language
- en
- Open access
- hybrid
- doi.org
- Licence
- cc-by
- Locations
- Conference on Empirical Methods in Natural Language Processing (EMNLP) (publishedVersion, cc-by)
- UA Campus Repository (The University of Arizona) (submittedVersion, cc-by-nc-sa)
- Topics
- Physical Sciences › Computer Science › Artificial Intelligence › Adversarial Robustness in Machine Learning
- Physical Sciences › Computer Science › Artificial Intelligence › Topic Modeling
- Physical Sciences › Computer Science › Artificial Intelligence › Natural Language Processing Techniques
- Keywords
- large language models, prompt injection, systemic vulnerability, adversarial prompts
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
