WASP fuses multiple pretrained language models with differentially private sample voting and contrastive prompts to synthesize task-specific text data, improving downstream classifier accuracy over single-model baselines when only about 100 private examples are available.
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
-
Contrastive Private Data Synthesis via Weighted Multi-PLM Fusion
WASP fuses multiple pretrained language models with differentially private sample voting and contrastive prompts to synthesize task-specific text data, improving downstream classifier accuracy over single-model baselines when only about 100 private examples are available.