Consensus sampling aggregates k distributions to achieve risk competitive with the average of the safest s, abstaining on low agreement, and formalizes this via R-robustness that bounds leakage and adversarial influence for generative AI safety.
hints” which can then be provided to psmall. 14 •In addition to generating hints, use the large models as “gatekeepers
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.AI 1years
2025 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Consensus Sampling for Safer Generative AI
Consensus sampling aggregates k distributions to achieve risk competitive with the average of the safest s, abstaining on low agreement, and formalizes this via R-robustness that bounds leakage and adversarial influence for generative AI safety.