PersonaLedger LLM simulator achieves AUC 0.70 for fraud detection at epsilon=1 from DP inputs but shows significant distribution drift due to learned priors overriding input statistics on temporal and demographic features.
Jun Zhang, Graham Cormode, Cecilia M Procopiuc, Divesh Srivastava, and Xiaokui Xiao
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Evaluating LLM Simulators as Differentially Private Data Generators
PersonaLedger LLM simulator achieves AUC 0.70 for fraud detection at epsilon=1 from DP inputs but shows significant distribution drift due to learned priors overriding input statistics on temporal and demographic features.