aiqbench analysis
This paper introduces Quantized Sufficient Statistics (QSS), a method that accelerates exact machine unlearning by replacing costly retraining with arithmetic subtraction of local prediction corrections. By decoupling global model structure from mutable content statistics, QSS achieves deletion latencies up to 483 times faster than existing approaches while maintaining accuracy comparable to state-of-the-art baselines. The approach offers a practical solution for deploying large models in privacy-sensitive environments where rapid data removal is critical.
Exact Unlearning via Quantized Sufficient Statistics
arXiv:2610.07197v1 Announce Type: new Abstract: Exact unlearning requires a deployed predictor to match one rebuilt without the information named by a deletion request. Existing general-purpose exact methods localize retraining through disjoint shards, but every request still invalidates a model, and smaller shards reduce the data available to each constituent predictor. We introduce Quantized Sufficient Statistics (QSS), which separates a small frozen schema from mutable, sum-decomposable content. The schema learns global structure; the content stores local prediction corrections as additive statistics indexed by quantized regions. Deleting content is therefore exact subtraction rather than optimization. We distinguish two guarantees: QSS-L exactly removes a label while retaining the unlabelled input, whereas QSS-E exactly removes both input and label by learning the schema without deletable examples. A deletion takes the arithmetic fast path with probability $1-\rho$ and triggers a full rebuild with probability $\rho$; all reported expected latencies include both events. Across 15 vision, text, and tabular datasets at $\rho=0.5\%$, QSS-L is within 2 percentage points of SISA on 11 tasks and provides 4--483$\times$ lower expected deletion latency on the low-class-count tasks where a compact schema is effective. QSS-E quantifies the additional accuracy cost of removing every trace of an input.
Abstract:Exact unlearning requires a deployed predictor to match one rebuilt without the information named by a deletion request. Existing general-purpose exact methods localize retraining through disjoint shards, but every request still invalidates a model, and smaller shards reduce the data available to each constituent predictor. We introduce Quantized Sufficient Statistics (QSS), which separates a small frozen schema from mutable, sum-decomposable content. The schema learns global structure; the content stores local prediction corrections as additive statistics indexed by quantized regions. Deleting content is therefore exact subtraction rather than optimization. We distinguish two guarantees: QSS-L exactly removes a label while retaining the unlabelled input, whereas QSS-E exactly removes both input and label by learning the schema without deletable examples. A deletion takes the arithmetic fast path with probability $1-\rho$ and triggers a full rebuild with probability $\rho$; all reported expected latencies include both events. Across 15 vision, text, and tabular datasets at $\rho=0.5\%$, QSS-L is within 2 percentage points of SISA on 11 tasks and provides 4--483$\times$ lower expected deletion latency on the low-class-count tasks where a compact schema is effective. QSS-E quantifies the additional accuracy cost of removing every trace of an input.
33 pages, 20 figures, 15 tables. Accepted at NeurIPS 2026
Machine Learning (cs.LG)
arXiv:2610.07197 [cs.LG]
(or arXiv:2610.07197v1 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2610.07197 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 296 words
Related coverage
A Red Line and Oversight Framework for Government AI Contracts
AI Alignment Forum · Jul 18, 2026
Exact Federated Continual Unlearning for Ridge Heads on Frozen Foundation Models
arXiv · Jun 16, 2026
Less Noise, Same Certificate: Retain Sensitivity for Unlearning
arXiv · Oct 14, 2026
Jev: System One models for Prod, not God — with Diogo Almeida, CEO, TypeSafe AI
Latent Space · Sep 21, 2026