Laplace-based local differential privacy outperforms Gaussian noise under Rényi-DP accounting for bounded [0,1]K prediction outputs¶

Claim · All claims
Evidence · 1 study · design r2 · q2
Currency · newest study 2026 · not yet checked against its sources · not reviewed by a person · recheck by 2029: tool and model accuracy changes quickly

Subclaims¶

q2 i? In the ablation, Gaussian-DP yields MAE worse than both the single-LLM baseline and the full system on all three datasets. → Yagna Manasa Boyapati 2026

Evidence¶

Yagna Manasa Boyapati 2026¶

Yagna Manasa Boyapati, Chong Yu, Tianyu Jiang, Justin Zhan. (2026). Privacy-Preserving Heterogeneous Multi-LLM Federated Inference for Cognitive Diagnosis. arXiv preprint. https://arxiv.org/abs/2609.02947 doi:10.48550/arxiv.2609.02947

q2 · i? · design · r2

Ablation comparison (Table 4) of noise mechanisms under ε=2.0; Gaussian-DP MAE was 0.3147 (ASSIST09), 0.2943 (GSM8K), 0.3165 (UCI), worse than both the full system and the single-LLM baseline.

"Gaussian noise under Rényi-DP accounting (Mironov, 2017) yields MAE worse than both the baseline and full system, confirming that Laplace-based LDP is superior for our bounded [0,1]K prediction setting."

Discussion¶

Related Claims¶

• Residual correction is essential: removing it while retaining LDP yields MAE worse than the single-LLM baseline across all datasets — related
• ε=2.0 provides the best privacy–utility balance, while ε=0.5 introduces excessive noise that degrades performance below baseline on ASSIST09 — related
• Heterogeneous multi-LLM federated inference improves cognitive diagnosis MAE over single-LLM baselines across three educational benchmarks with minimal privacy cost — related