# Exact Unlearning via Quantized Sufficient Statistics

arXiv: https://arxiv.org/abs/2610.07197 · HTML: https://mlsanity.com/paper/2610.07197/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Placed in Machine unlearning, one of 12 directions in LLM safety and security.
- Builds on 10 of its 27 references that are in the map.
- First posted on arXiv 5 Oct 2026.

## Placed in

- LLM safety and security / [Machine unlearning](https://mlsanity.com/safety/machine-unlearning/)

## Builds on (most cited first, 10 of 10)

- [Concentrated Differential Privacy: Simplifications, Extensions, and Lower Bounds](https://mlsanity.com/paper/1605.02065/) (2016)
- [LLM Unlearning via Neural Activation Redirection](https://mlsanity.com/paper/2502.07218/) (2025)
- [Fast Exact Unlearning for In-Context Learning Data for LLMs](https://mlsanity.com/paper/2402.00751/) (2024)
- [Targeted Unlearning with Single Layer Unlearning Gradient](https://mlsanity.com/paper/2407.11867/) (2024)
- [MU-Bench: A Multitask Multimodal Benchmark for Machine Unlearning](https://mlsanity.com/paper/2406.14796/) (2024)
- [Exact Federated Continual Unlearning for Ridge Heads on Frozen Foundation Models](https://mlsanity.com/paper/2603.12977/) (2026)
- [Towards Efficient and Exact Forgetting Services in Pre-Trained-Model-based Continual Learning](https://mlsanity.com/paper/2505.12239/) (2025)
- [Forget by Uncertainty: Orthogonal Entropy Unlearning for Quantized Neural Networks](https://mlsanity.com/paper/2602.00567/) (2026)
- [Provable Unlearning with Gradient Ascent on Two-Layer ReLU Neural Networks](https://mlsanity.com/paper/2510.14844/) (2025)
- [QUAIL: Quantization Aware Unlearning for Mitigating Misinformation in LLMs](https://mlsanity.com/paper/2601.15538/) (2026)

## Abstract (arXiv)

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-ρ and triggers a full rebuild with probability ρ; all reported expected latencies include both events. Across 15 vision, text, and tabular datasets at ρ=0.5%, QSS-L is within 2 percentage points of SISA on 11 tasks and provides 4--483× 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.