🧪 Test?View on arXiv
RL-ATTACK: A Generalized Black-Box Adversarial Attack on Language Models via Reinforcement Learning
Taikun Udel, Author 2, Author 3, Author 4, Author 5
adversarial attacksreinforcement learningblack-box modelslanguage models
2611.08711
Builder Relevance
16h ago80%
Abstract
This paper presents a reinforcement learning-based framework for generating adversarial attacks on language models in black-box settings.
Reality Card
Core Claim
The proposed RL-ATTACK framework achieves lower post-attack accuracy and greater query efficiency compared to untrained attackers and a strong optimization-based baseline.
Method / Result
Achieves post-attack accuracy of 2% within 20 queries on natural language inference tasks.
Limitations
No single reward design or training setting consistently outperforms others, which may complicate reproducibility.
Paper to code
Verified implementation resources so builders can test the paper’s claims instead of stopping at the abstract.
No verified implementation link has been attached yet. AIBuzzHub will keep this panel separate from unverified search results.