Unsupervised Layer-wise Score Aggregation for Textual OOD Detection
This addresses robustness and security needs for AI-based systems by enhancing OOD detection in text, though it is incremental as it builds on existing anomaly score methods.
The paper tackled the problem of textual out-of-distribution (OOD) detection by proposing an unsupervised method to aggregate layer-wise anomaly scores, showing that the last layer is often suboptimal. The method achieved near-oracle performance on an extended benchmark with up to 77 classes, improving robustness and consistency.
Out-of-distribution (OOD) detection is a rapidly growing field due to new robustness and security requirements driven by an increased number of AI-based systems. Existing OOD textual detectors often rely on an anomaly score (e.g., Mahalanobis distance) computed on the embedding output of the last layer of the encoder. In this work, we observe that OOD detection performance varies greatly depending on the task and layer output. More importantly, we show that the usual choice (the last layer) is rarely the best one for OOD detection and that far better results could be achieved if the best layer were picked. To leverage this observation, we propose a data-driven, unsupervised method to combine layer-wise anomaly scores. In addition, we extend classical textual OOD benchmarks by including classification tasks with a greater number of classes (up to 77), which reflects more realistic settings. On this augmented benchmark, we show that the proposed post-aggregation methods achieve robust and consistent results while removing manual feature selection altogether. Their performance achieves near oracle's best layer performance.