When Rule Violations Are Rare: Chimera Training for Logical Anomaly Detection

2026-05-24

AI Summary

This work addresses the scarcity of genuine rule-violating samples in training data by proposing a logic-rule-based structured semantic anomaly detection method. The approach compiles logical rules into directed acyclic graphs (DAGs) and introduces a neural rule evaluator equipped with a subtree MLP gating mechanism. By performing “chimeric training” at the feature level, it generates operand-level counterfactual samples that provide intermediate supervision signals without requiring real anomalies. This strategy effectively covers sparse ground-truth configurations and mitigates shortcut learning. Evaluated on CLEVRER, OpenImages, and VidOR datasets, the method significantly improves rule-level anomaly detection performance in terms of AUROC, outperforming existing baselines—particularly on compositional and relational rules—while simultaneously producing both anomaly scores and rule-level attributions.

Abstract

Many practical anomalies are not merely rare inputs, but violations of semantic constraints: objects co-occur in structured ways, actions imply preconditions, and events satisfy temporal or relational regularities. We study anomaly detection in this setting, where constraints are given as logical rules over learned visual concepts, but real rule violations are rare or absent during training. We propose a neural rule evaluator that compiles each constraint into a directed acyclic graph and learns feature-aware subtree MLP gates for its internal logical operators. Each gate maps child features and edge-level negations to a parent representation and a rule-satisfaction probability, with intermediate supervision obtained from exact Boolean propagation over ground-truth concept labels. The key difficulty is that same-image training data often provide insufficient coverage of informative truth configurations and also allow shortcut solutions. To address this, we introduce chimera training: an operand-level counterfactual construction at the feature level. Instead of mixing input images, we concatenate subtree features from different samples; each operand keeps the hard truth label of the sample it came from, and the chimera target is obtained by applying the node's logical operator to those inherited labels. This supplies supervised logical counterexamples without requiring real anomalous images. Across CLEVRER, OpenImages, and VidOR, the resulting evaluator improves rule-level anomaly AUROC over independent-events and same-image semantic-training baselines, especially for compositional and relational rules. The method yields both scalar anomaly scores and rule-level attributions.

Problem

• logical anomaly detection
• rule violations
• semantic constraints
• rare anomalies
• supervised counterexamples

Innovation

• chimera training
• logical anomaly detection
• neural rule evaluator
• counterfactual feature construction
• semantic constraints