← Back to all news
arXiv Machine Learning

Privacy Without Regret: Differentially Private Inference-Time Alignment

The paper introduces Private Best-of-N (PrivBoN), a method that adds calibrated Gumbel noise to reward scores during inference-time alignment, achieving both ε-differential privacy and KL-regularized alignment. When the privacy budget exceeds a critical threshold ε*, the noise becomes regret-optimal, matching the theoretical alignment skyline. The authors also propose Private Inference-Time Pessimism (PrivITP), which uses χ^2-regularized rejection sampling and a two-phase Gaussian mechanism to provide ex-post (ε,δ)-DP with a privacy cost independent of the number of responses, and demonstrate that both methods outperform standard Best-of-N across multiple models and datasets.

Related stories

arXiv Machine Learning
6d ago

Differentially Private Sparse Reward Estimation with Preference Feedback

The paper introduces a differentially private framework for sparse reward estimation in reinforcement learning with human feedback (RLHF). It assumes that only a small subset of features influences human preferences and provides theoretical error bounds for both Sample-DP and Label-DP settings under the Bradley-Terry-Luce model. Empirical results show that the proposed methods outperform dense baselines on sentiment generation and safety alignment tasks, achieving a better privacy‑utility trade‑off.

By Meng Ding, Mingxi Lei, Jie Zhang, Jinyan Liu, Di Wang
llmsreinforcement-learningbenchmarkssafety
More like this →
arXiv AI
Sep 10

PAC-Private Autoregressive Generation: Calibrating Noise to Ensemble Disagreement

The paper introduces PAC‑Private Autoregressive Generation, a method that calibrates noise based on ensemble disagreement across overlapping ‘worlds’ of a private corpus, thereby extending PAC privacy from classification to text generation. By training adapters on a frozen public model and using posterior‑weighted disagreement to add noise only when predictions vary, the approach achieves strong privacy guarantees while preserving most of the fine‑tuning benefit. Experiments on WikiText‑103 with GPT‑2‑small show 74 % of the fine‑tuning gain retained with a per‑token budget of 2⁻³², and membership‑inference success bounded to 51.08 % after one million tokens, outperforming PMixED under matched conditions.

By Mina Mirzadehsarcheshmeh, Amir Keyvan Khandani
llmsfine-tuningsafety
More like this →