Profile-DP: Differential Privacy for User-Controlled Profile Release in Personalized LLM Inference
We introduce Profile-DP, a policy-aware local differential privacy framework designed specifically for user-profile release in personalized LLM inference.
ProofPaper ↗
Key points
- Personalized large language models (LLMs) increasingly rely on profile-conditioned inference, pairing public requests with rich, user-specific facts.
- Profile-DP provably bounds the release of user-withheld facts while treating public instructions, profile schemas, and authorized facts as public side information.
- By generating a privatized profile that composes with the public instruction prior to answer generation, our framework decouples formal privacy guarantees from downstream model behavior.
- We develop two finite-domain mechanisms satisfying Profile-DP: DP-SLOT independently perturbs protected attributes, while DP-BLOCK uses public semantic decisions to coordinate their release and jointly sample related attributes.
Sources (1)
- [1]Profile-DP: Differential Privacy for User-Controlled Profile Release in Personalized LLM InferencearXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 19, 07:44 PM
We introduce Profile-DP, a policy-aware local differential privacy framework designed specifically for user-profile release in personalized LLM inference.
Personalized large language models (LLMs) increasingly rely on profile-conditioned inference, pairing public requests with rich, user-specific facts.
Extractive summary: sentences quoted from the sources.
Before this
- Oct 19, 2026Introducing Clairic, a new System One class of extraction models
- Oct 19, 2026Push-based KV transfer to improve TTFT and TPOT for disaggregated inference in vLLM
- Oct 19, 2026Rethinking LLM Fine-Tuning via Weight Space Reparameterization: Preserving Safety during Downstream Adaptation
- Oct 19, 2026From Black Box to Controller: Inference-Time Steering of LLM via Feature Amplification
- Oct 18, 2026CorrectRTL: Boost RTL Generation Quality Through Automated Self-Correction with Instruction Decomposition
- Oct 17, 2026When LLM-Generated Refactorings Ripple: An Empirical Study of Repository Adoption Readiness