Confidence elicitation, asking a model to verbalize its uncertainty, can be used as a soft-label signal to craft stronger black-box word-substitution attacks on LLMs.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Confidence Elicitation: A New Attack Vector for Large Language Models
Confidence elicitation, asking a model to verbalize its uncertainty, can be used as a soft-label signal to craft stronger black-box word-substitution attacks on LLMs.