aiqbench analysis
The paper argues that the usual split of model uncertainty into aleatoric and epistemic parts depends on the model rather than the data, and that under model mismatch some training data can actively hurt performance on a specific task. It proposes an influence-function method for Gaussian process models that removes such detrimental samples, and reports gains in Bayesian optimization, model-based reinforcement learning, and transductive learning. For ML practitioners, the practical question is whether the gains hold outside the benchmarked settings and how costly the influence computation is at scale.
Forgetting to Improve: Principled Data Removal in Active Learning
arXiv:2611.04054v1 Announce Type: cross Abstract: The uncertainty of a statistical model is most commonly factorized into an aleatoric and an epistemic part. This factorization changes how predictions are interpreted in downstream decision tasks. Importantly, except for the idealistic scenario with no model mismatch, the quantification is a characteristic of the model and not the data generating process. In this paper, we propose Forgetting to Improve a method that reduces this discrepancy by incorporating the task into the modeling framework. Our key insight is to acknowledge that in scenarios of model mismatch, data can have a detrimental effect on the modeling for a specific task. Based on this insight, we propose an influence function for Gaussian process models that allows for principled removal of detrimental data samples. We showcase the flexibility of this approach by demonstrating significant improvements across a range of tasks, including Bayesian optimization, model-based reinforcement learning, and transductive learning.
Abstract:The uncertainty of a statistical model is most commonly factorized into an aleatoric and an epistemic part. This factorization changes how predictions are interpreted in downstream decision tasks. Importantly, except for the idealistic scenario with no model mismatch, the quantification is a characteristic of the model and not the data generating process. In this paper, we propose Forgetting to Improve a method that reduces this discrepancy by incorporating the task into the modeling framework. Our key insight is to acknowledge that in scenarios of model mismatch, data can have a detrimental effect on the modeling for a specific task. Based on this insight, we propose an influence function for Gaussian process models that allows for principled removal of detrimental data samples. We showcase the flexibility of this approach by demonstrating significant improvements across a range of tasks, including Bayesian optimization, model-based reinforcement learning, and transductive learning.
Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
arXiv:2611.04054 [cs.AI]
(or arXiv:2611.04054v1 [cs.AI] for this version)
https://doi.org/10.48550/arXiv.2611.04054 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 · 243 words
Related coverage
Forgetting is Everywhere
arXiv · Jul 28, 2026
Distance Is Not Enough: Forget-Retain Alignment Gap Predicts LLM Relearning Robustness
arXiv · Aug 27, 2026
SAUL: Sharpness-Aware Augmented-Lagrangian Unlearning
arXiv · Aug 19, 2026
Towards Mitigating Excessive Forgetting in LLM Unlearning via Entanglement-Guidance with Proxy Constraint
arXiv · Sep 24, 2026