AI Paper Atlas

Research map / Federated learning and privacy / Differential privacy

PEARL: Differentially Private and Entropy-Aware Regulated Language Generation

Seongho Joo, Hyukhun Koh, Kyomin Jung

ICML 2026 · Federated learning and privacy › Differential privacy · approach: Differentially private synthetic data

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
apply PEARL confidence-gap-guided redaction and refill on top of DP decoding (ablation)exponential-mechanism DP decoding (NoRefill) · LeakRAG-Medical eps=3, LLaMA 3.1 8BHalluScore: 2.52 -> 1.87Improved
select refill spans by confidence gap instead of at random (ablation)PEARL · LeakRAG-Medical eps=3, LLaMA 3.1 8BHalluScore: 1.97 -> 1.87Improved
select refill spans by confidence gap instead of at random (ablation)PEARL · LeakRAG-Financial eps=3, LLaMA 3.1 8BBERTScore: 52.30 -> 51.15No clear effect
use confidence-gap-guided rewriting at eps=3PEARL · LeakRAG-Medical, LLaMA 3.1 8BHalluScore: 2.04 (non-private few-shot) -> 1.87Improved
add refill stage after redaction (Refill with GPT vs Full Redaction) (ablation)PEARL · LeakRAG-Medical eps=3, LLaMA 3.1 8BGoldAlign: 2.88 -> 3.06Improved
add redaction (without refill) to DP decoding (ablation)NoRefill · LeakRAG-Medical eps=3, LLaMA 3.1 8BHalluScore: 2.52 -> 2.06Improved
add PEARL on top of refill/redaction at eps=6 (ablation)PEARL · LeakRAG-Medical eps=6, LLaMA 3.1 8BHalluScore: 1.92 -> 1.90No clear effect
increase number of filtered sentences k (ablation)PEARL · LeakRAGHalluScore and leaked PIIImproved
use PEARL instead of PICL private decodingprivate decoding · LeakRAG-Medical eps=8, LLaMA 3.1 8BLeaked PII: 3.60 -> 0.57Improved
use PEARL instead of KSA private decodingprivate decoding · LeakRAG-Medical eps=6, LLaMA 3.1 8BLeaked PII: 0.50 -> 0.53No clear effect
apply DP noise (exponential mechanism) to RAG decodingRAG LLM · LeakRAG, LLaMA 3.1 8BHalluScore: 2.04 (non-private) -> 2.52 (NoRefill, eps=3)Hurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy