Tag
This paper proposes a method to maximize the area under the ROC curve (AUC) from biased positive-unlabeled data by using confidence values to address the bias in labeled positive data.
The paper introduces Local Reference Geometry (LRG), a lightweight post-hoc feature augmentation module that enhances local feature reliability for minority classes in imbalanced time series classification by measuring and repairing geometry failures.
This paper benchmarks 11 text augmentation methods, including classical, embedding-space, and LLM-based approaches, across 7 imbalanced classification datasets. It finds that retrieval-based oversampling (EmbSMOTE) outperforms LLM-based augmentation, and that preserving class-conditional structure matters more than surface-level diversity.
This paper proposes BS-cRT, a two-stage baseline for long-tailed recognition that trains a backbone with Balanced Softmax and then retrains only the classifier on balanced batches, achieving consistent few-shot accuracy improvements on multiple benchmarks.
RUBRIC is a generator-agnostic filtering framework for imbalanced classification that selects synthetic samples by balancing realism (via a discriminator) and utility (margin-based scoring), improving F1-macro and recall on benchmarks like credit-card fraud detection.