PromptInject shows that simple adversarial prompts can cause goal hijacking and prompt leaking in GPT-3, exploiting its stochastic behavior.
Textattack: A framework for adversarial attacks, data augmentation, and adversarial training in nlp
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2022 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Ignore Previous Prompt: Attack Techniques For Language Models
PromptInject shows that simple adversarial prompts can cause goal hijacking and prompt leaking in GPT-3, exploiting its stochastic behavior.