Unlearning Under Imbalance: Benchmarking Fairness in Multimodal LLM Unlearning

• URL: http://arxiv.org/abs/2607.21300v1
• Date: Thu, 23 Jul 2026 13:24:34 GMT
• Title: Unlearning Under Imbalance: Benchmarking Fairness in Multimodal LLM Unlearning
• Abstract summary: We present FAIRGET, the first benchmark that evaluates unlearning under unbalanced, realistic, forget requests.<n>We also propose FAUN, the first unlearning algorithm for MLLMs that forgets unlearning data while preserving model fairness.
• Score: 33.59393479055872
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Machine unlearning has emerged as a tool for removing personal data from trained models to comply with recent AI regulations. To evaluate unlearning effectiveness in multimodal large language models (MLLMs), prior works fine-tune models on fictitious identities, simulating unlearning requests on subsets of these IDs, which are typically uniformly distributed. However, in realistic scenarios, people from different demographic groups may request to be unlearned at different frequencies, potentially altering the model's internal beliefs for these groups and leading to biased behaviors. To fill this gap, we propose FAIRGET, the first Visual Question Answering benchmark that evaluates unlearning under unbalanced, realistic, forget requests. These requests are designed to simulate multiple realistic scenarios, ranging from simple to challenging settings, that lead to biased unlearned models if fairness is not accounted for. Additionally, we propose FAUN, the first unlearning algorithm for MLLMs that forgets unlearning data while preserving model fairness. FAUN exploits a bias-aware activation steering mechanism to unlearn identities while accounting for the unbalanced nature of the forget data. Experiments on FAIRGET and the established FIUBench demonstrate our method's superiority both in unlearning quality and fairness.

Related papers

• Multi-Objective Learning for Diffusion Models: A Statistical Theory under Semi-Supervised Learning [83.65494573724904]
  We develop a principled MOL framework for diffusion models with limited data.<n>We propose a two-stage training procedure that first fits lightweight specialist models from limited paired data.<n>We establish bounds generalization showing that the required number of paired samples only depends on the complexity of the specialist model classes.
  arXiv  Detail & Related papers  (2026-05-24T18:19:40Z)
• FaLW: A Forgetting-aware Loss Reweighting for Long-tailed Unlearning [24.734154431191538]
  FaLW is a plug-and-play, instance-wise dynamic loss reweighting method.<n>It assesses the unlearning state of each sample by comparing its predictive probability to the distribution of unseen data from the same class.<n>Experiments demonstrate that FaLW achieves superior performance.
  arXiv  Detail & Related papers  (2026-01-26T16:21:01Z)
• Non-Invasive Fairness in Learning through the Lens of Data Drift [88.37640805363317]
  We show how to improve the fairness of Machine Learning models without altering the data or the learning algorithm. We use a simple but key insight: the divergence of trends between different populations, and, consecutively, between a learned model and minority populations, is analogous to data drift. We explore two strategies (model-splitting and reweighing) to resolve this drift, aiming to improve the overall conformance of models to the underlying data.
  arXiv  Detail & Related papers  (2023-03-30T17:30:42Z)
• Debiasing Vision-Language Models via Biased Prompts [79.04467131711775]
  We propose a general approach for debiasing vision-language foundation models by projecting out biased directions in the text embedding. We show that debiasing only the text embedding with a calibrated projection matrix suffices to yield robust classifiers and fair generative models.
  arXiv  Detail & Related papers  (2023-01-31T20:09:33Z)
• Synthetic Model Combination: An Instance-wise Approach to Unsupervised Ensemble Learning [92.89846887298852]
  Consider making a prediction over new test data without any opportunity to learn from a training set of labelled data. Give access to a set of expert models and their predictions alongside some limited information about the dataset used to train them.
  arXiv  Detail & Related papers  (2022-10-11T10:20:31Z)
• Fairness Reprogramming [42.65700878967251]
  We propose a new generic fairness learning paradigm, called FairReprogram, which incorporates the model reprogramming technique. Specifically, FairReprogram considers the case where models can not be changed and appends to the input a set of perturbations, called the fairness trigger. We show both theoretically and empirically that the fairness trigger can effectively obscure demographic biases in the output prediction of fixed ML models.
  arXiv  Detail & Related papers  (2022-09-21T09:37:00Z)
• Fair Classification via Transformer Neural Networks: Case Study of an Educational Domain [0.0913755431537592]
  This paper presents a preliminary investigation of fairness constraint in transformer networks on Law School Student neural datasets. We have employed fairness metrics for evaluation and check the trade-off between fairness and accuracy.
  arXiv  Detail & Related papers  (2022-06-03T06:34:16Z)
• FairIF: Boosting Fairness in Deep Learning via Influence Functions with Validation Set Sensitive Attributes [51.02407217197623]
  We propose a two-stage training algorithm named FAIRIF. It minimizes the loss over the reweighted data set where the sample weights are computed. We show that FAIRIF yields models with better fairness-utility trade-offs against various types of bias.
  arXiv  Detail & Related papers  (2022-01-15T05:14:48Z)
• Certifiable Machine Unlearning for Linear Models [1.484852576248587]
  Machine unlearning is the task of updating machine learning (ML) models after a subset of the training data they were trained on is deleted. We present an experimental study of the three state-of-the-art approximate unlearning methods for linear models.
  arXiv  Detail & Related papers  (2021-06-29T05:05:58Z)
• Self-Damaging Contrastive Learning [92.34124578823977]
  Unlabeled data in reality is commonly imbalanced and shows a long-tail distribution. This paper proposes a principled framework called Self-Damaging Contrastive Learning to automatically balance the representation learning without knowing the classes. Our experiments show that SDCLR significantly improves not only overall accuracies but also balancedness.
  arXiv  Detail & Related papers  (2021-06-06T00:04:49Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.