Fast Exact Unlearning for In-Context Learning Data for LLMs

Andrei Ioan Muresanu, Anvith Thudi, Michael R. Zhang, Nicolas Papernot
Proceedings of the 42nd International Conference on Machine Learning, PMLR 267:45272-45288, 2025.

Abstract

Modern machine learning models are expensive to train, and there is a growing concern about the challenge of retroactively removing specific training data. Achieving exact unlearning in deep learning pipelines—producing models as if certain data had never been included in training—remains an open problem. In this paper, we revisit exact unlearning in deep learning and show that for large language models (LLMs) we can efficiently exactly unlearn “fine-tuning data" (the data used to adapt a pre-trained model). This follows from two observations. First, we can use in-context learning to adapt the LLM to the fine-tuning dataset instead of SGD based algorithms. Second, we show that accurate in-context learning can be done with quantized k-means, which allows for effectively constant time unlearning operations. Our evaluation shows that this unlearning recipe has similar performance to fine-tuning alternatives, but vastly reduces the unlearning costs. Our study also highlights the need for new measures of unlearning cost when adapting the learning algorithm to have faster unlearn operations.

Cite this Paper

BibTeX
@InProceedings{pmlr-v267-muresanu25a, title = {Fast Exact Unlearning for In-Context Learning Data for {LLM}s}, author = {Muresanu, Andrei Ioan and Thudi, Anvith and Zhang, Michael R. and Papernot, Nicolas}, booktitle = {Proceedings of the 42nd International Conference on Machine Learning}, pages = {45272--45288}, year = {2025}, editor = {Singh, Aarti and Fazel, Maryam and Hsu, Daniel and Lacoste-Julien, Simon and Berkenkamp, Felix and Maharaj, Tegan and Wagstaff, Kiri and Zhu, Jerry}, volume = {267}, series = {Proceedings of Machine Learning Research}, month = {13--19 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v267/main/assets/muresanu25a/muresanu25a.pdf}, url = {https://proceedings.mlr.press/v267/muresanu25a.html}, abstract = {Modern machine learning models are expensive to train, and there is a growing concern about the challenge of retroactively removing specific training data. Achieving exact unlearning in deep learning pipelines—producing models as if certain data had never been included in training—remains an open problem. In this paper, we revisit exact unlearning in deep learning and show that for large language models (LLMs) we can efficiently exactly unlearn “fine-tuning data" (the data used to adapt a pre-trained model). This follows from two observations. First, we can use in-context learning to adapt the LLM to the fine-tuning dataset instead of SGD based algorithms. Second, we show that accurate in-context learning can be done with quantized k-means, which allows for effectively constant time unlearning operations. Our evaluation shows that this unlearning recipe has similar performance to fine-tuning alternatives, but vastly reduces the unlearning costs. Our study also highlights the need for new measures of unlearning cost when adapting the learning algorithm to have faster unlearn operations.} }
Endnote
%0 Conference Paper %T Fast Exact Unlearning for In-Context Learning Data for LLMs %A Andrei Ioan Muresanu %A Anvith Thudi %A Michael R. Zhang %A Nicolas Papernot %B Proceedings of the 42nd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2025 %E Aarti Singh %E Maryam Fazel %E Daniel Hsu %E Simon Lacoste-Julien %E Felix Berkenkamp %E Tegan Maharaj %E Kiri Wagstaff %E Jerry Zhu %F pmlr-v267-muresanu25a %I PMLR %P 45272--45288 %U https://proceedings.mlr.press/v267/muresanu25a.html %V 267 %X Modern machine learning models are expensive to train, and there is a growing concern about the challenge of retroactively removing specific training data. Achieving exact unlearning in deep learning pipelines—producing models as if certain data had never been included in training—remains an open problem. In this paper, we revisit exact unlearning in deep learning and show that for large language models (LLMs) we can efficiently exactly unlearn “fine-tuning data" (the data used to adapt a pre-trained model). This follows from two observations. First, we can use in-context learning to adapt the LLM to the fine-tuning dataset instead of SGD based algorithms. Second, we show that accurate in-context learning can be done with quantized k-means, which allows for effectively constant time unlearning operations. Our evaluation shows that this unlearning recipe has similar performance to fine-tuning alternatives, but vastly reduces the unlearning costs. Our study also highlights the need for new measures of unlearning cost when adapting the learning algorithm to have faster unlearn operations.
APA
Muresanu, A.I., Thudi, A., Zhang, M.R. & Papernot, N.. (2025). Fast Exact Unlearning for In-Context Learning Data for LLMs. Proceedings of the 42nd International Conference on Machine Learning, in Proceedings of Machine Learning Research 267:45272-45288 Available from https://proceedings.mlr.press/v267/muresanu25a.html.

Related Material