AITOPAI热报
PapersSelected73°

Plasticity Collapse in Machine Unlearning

On the Plasticity Collapse in Continual Machine Unlearning

arXiv cs.LG@Yingdan Shi et al. (5)
Paper
Read the original
Why we picked it

The research addresses a critical challenge for AI compliance with privacy regulations in ASEAN, where models must repeatedly forget data without losing effectiveness.

This paper identifies plasticity collapse as a fundamental limitation in continual machine unlearning. The research shows that sequential unlearning operations accumulate geometric constraints in parameter space, leading to saturated subspaces. Experiments across multiple architectures and datasets confirm two failure modes: forward failure with diminishing forgetting quality and backward failure with spontaneous re-memorization. The findings reveal a critical barrier to long-term reliability of unlearning systems.

Original · arXiv cs.LG

On the Plasticity Collapse in Continual Machine Unlearning

Machine unlearning enables deep neural networks to selectively remove the influence of specific data in response to privacy and regulatory requirements. While prior work largely studies single-shot unlearning, real-world systems must accommodate continual unlearning, where multiple unlearning requests occur sequentially over time. In this work, we identify a fundamental limitation of this setting: plasticity collapse, a progressive breakdown in a model's ability to effectively forget. Through theoretical analysis of continual unlearning dynamics, we show that continual unlearning operations accumulate geometric constraints in parameter space, leading to saturated subspaces that restrict future updates. This structural effect induces two distinct failure modes: (1) Forward failure -- diminishing forgetting quality for subsequent tasks, and (2) Backward failure -- spontaneous re-memorization of previously forgotten information. Extensive experiments across multiple architectures, datasets, and methods in image classification confirm that plasticity collapse is not an artifact of specific implementations, but a pervasive phenomenon inherent to continual unlearning. Our findings reveal a critical barrier to the long-term reliability of machine unlearning systems and motivate the development of plasticity-preserving unlearning algorithms. Our code is available at https://github.com/TIML-Group/Continual-Machine-Unlearning-Plasticity-Collapse

#machine unlearning#plasticity collapse#deep neural networks#privacy#continual learning

Keep reading

Recent stories with the same tag or section
PapersBuilders09:52
Chinese contentSequential deletion Gaussian mechanism for machine unlearning with GDP privacy certification

arXiv has a new theory paper on machine unlearning, proving that Gaussian noise plus GDP can certify that models after sequential deletion are nearly indistinguishable from retraining, with quantitative bounds on how much less noise is needed for the random walk solution.

#machine unlearning#Gaussian differential privacy#privacy#paper interpretation
arXiv cs.LG@Qi Kuang, Yin XiaSource
PapersBuildersSelected21:02
EC²-Bench benchmarks edge-cloud collaboration for long-horizon agents

Cost and privacy trade-offs in agent execution do not move together — once task completion converges, strategy choice is mostly a privacy decision, and the best-scoring setup is also the most exposed. Teams deploying on-device agents can pick a collaboration mode based on edge model size: route execution to the cloud for small models, use cloud advisory for 9B-class ones. The comparison against Tsinghua-affiliated GLM-5.3 makes the findings directly applicable to Chinese open-source model stacks.

#EC2-Bench#OpenBMB#GLM-5.3#Qwen
OpenBMB@OpenBMBSource
PapersBuildersSelected17:32
University of Chicago Paper Finds LLMs Misreport Their Own Cruxes

Sampling multiple answers and diffing them beats asking a model to explain itself — a cheap workflow change that applies to any high-stakes LLM use like research triage or forecasting. The 86x compute saving from the reranker makes it practical to run on everyday tasks. Practitioners building review pipelines or agent self-checks in Singapore and the wider ASEAN developer community can adopt this with any API-based model.

#LLM evaluation#reasoning models#Metaculus#University of Chicago
rohanpaul_ai@rohanpaul_aiSource