aiqbench analysis
RAPTOR addresses critical inefficiencies in differentially private fine-tuning for sparse Mixture-of-Experts models by distinguishing between shared layers and expert-specific computations. The framework introduces role-aware optimization strategies that prevent privacy noise from degrading performance on low-load experts while maintaining rigorous $(\varepsilon,\delta)$-DP guarantees. Experimental results on architectures like Switch Transformer demonstrate consistent improvements over standard baselines, particularly under tight privacy budgets.
RAPTOR: Role-Aware Private Training for Mixture-of-Experts
arXiv:2609.05770v1 Announce Type: new Abstract: Differentially private (DP) fine-tuning methods treat sparse Mixture-of-Experts (MoE) models as a single dense block, ignoring that shared layers see all data while experts only see routed records. We identify and formally characterize three resulting failure modes: global clipping suppresses expert gradients, batch-level normalization dilutes sparse expert updates, and fixed privacy noise degrades signal-to-noise ratio on low-load experts. We introduce RAPTOR - a Role-Aware Private Training framework, which alternates shared and expert optimization and targets each failure directly, using expert-specific clipping and noise together with a public expected-owner denominator and a count-independent update schedule that avoids conditioning on private, realized expert counts. We prove the resulting mechanism satisfies $(\varepsilon,\delta)$-DP: because each record is assigned to exactly one owner expert, per-expert mechanisms within a layer compose in parallel, so updating all $E$ experts costs no more, in privacy terms, than updating one, with shared and expert streams composing sequentially across training. We further derive a bias-variance decomposition of the public-denominator estimator showing its bias grows predictably with routing imbalance, yielding a privacy-free rule for selecting which layer to protect from routing entropy measured on a small public corpus. Experiments on Switch Transformer and OLMoE fine-tuning across GLUE tasks, and on DeepSeek-VL2-Tiny, show consistent gains over standard DP baselines across several privacy levels ($\varepsilon$), with the largest margins typically at the tightest budgets. Code and models are publicly available: https://github.com/leduckhai/RAPTOR
Authors:Duc Dm, Khai Le-Duc, Nguyen Do, Minh Son Hoang, Florent Draye, Thai Hoang, Hoang Phuong Dam, Jiarui Liu, Chris Ngo, Terry Jingchen Zhang, Anh Le Duc Tran, Nhat Do Minh, Minh Ngoc Le, My T. Thai, Ran Xu, Silvio Savarese, Mona Diab, Bernhard Schölkopf, Zhijing Jin, Huy L. Nguyen, Daeyoung Kim
Abstract:Differentially private (DP) fine-tuning methods treat sparse Mixture-of-Experts (MoE) models as a single dense block, ignoring that shared layers see all data while experts only see routed records. We identify and formally characterize three resulting failure modes: global clipping suppresses expert gradients, batch-level normalization dilutes sparse expert updates, and fixed privacy noise degrades signal-to-noise ratio on low-load experts. We introduce RAPTOR - a Role-Aware Private Training framework, which alternates shared and expert optimization and targets each failure directly, using expert-specific clipping and noise together with a public expected-owner denominator and a count-independent update schedule that avoids conditioning on private, realized expert counts. We prove the resulting mechanism satisfies $(\varepsilon,\delta)$-DP: because each record is assigned to exactly one owner expert, per-expert mechanisms within a layer compose in parallel, so updating all $E$ experts costs no more, in privacy terms, than updating one, with shared and expert streams composing sequentially across training. We further derive a bias-variance decomposition of the public-denominator estimator showing its bias grows predictably with routing imbalance, yielding a privacy-free rule for selecting which layer to protect from routing entropy measured on a small public corpus. Experiments on Switch Transformer and OLMoE fine-tuning across GLUE tasks, and on DeepSeek-VL2-Tiny, show consistent gains over standard DP baselines across several privacy levels ($\varepsilon$), with the largest margins typically at the tightest budgets. Code and models are publicly available: this https URL
Machine Learning (cs.LG); Computation and Language (cs.CL)
arXiv:2609.05770 [cs.LG]
(or arXiv:2609.05770v1 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2609.05770 arXiv-issued DOI via DataCite (pending registration)
Full text under the publisher's open licence · 303 words
Related coverage
[AINews] not much happened today
Latent Space · Sep 10, 2026
The Continual Constitutional Convention: How language models revise the documents that govern them
AI Alignment Forum · Oct 13, 2026
J++ Lens: Jacobian Filtering Enables More Faithful Workspace Lenses
LessWrong · Oct 8, 2026
[AINews] Reality Checks on AI News (Yegge shuts down Gas Town, Databricks’ +60% Astra cost)
Latent Space · Sep 17, 2026