STING red-teaming framework achieves higher illicit-task completion rates in multi-turn LLM agent interactions than single-turn baselines, with multilingual tests across six languages showing no consistent rise in lower-resource settings.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Helpful to a Fault: Measuring Illicit Assistance in Multi-Turn, Multilingual LLM Agents
STING red-teaming framework achieves higher illicit-task completion rates in multi-turn LLM agent interactions than single-turn baselines, with multilingual tests across six languages showing no consistent rise in lower-resource settings.