[HUGGINGFACE]score: 0.47
LogMILP Localizes Anomalous Log Entries Using Only Bag-Level Labels via Counterfactual Perturbation
May 8, 2026
LogMILP uses multi-instance learning with prototype-guided structural modeling and counterfactual perturbation consistency regularization to localize individual anomalous log entries in networked systems using only coarse bag-level anomaly labels.
paper
HOW THIS AFFECTS YOU
●
builderYou can improve log anomaly localization in production systems without the cost of instance-level annotation, using only existing bag-level labels.
●
researcherCounterfactual perturbation consistency as a regularizer for MIL-based localization is a transferable technique for other weakly supervised detection tasks.