Anomaly detection can be performed directly using a primary model's internal neuron output ranges, eliminating the need for expensive external AD models.
March 19, 2026
Original Paper
RangeAD: Fast On-Model Anomaly Detection
arXiv · 2603.17795
The Takeaway
RangeAD leverages information already encoded in the primary model to detect distributional shifts and filter inputs. This 'on-model' approach achieves superior performance on high-dimensional tasks with substantially lower inference overhead, streamlining production ML pipelines.
From the abstract
In practice, machine learning methods commonly require anomaly detection (AD) to filter inputs or detect distributional shifts. Typically, this is implemented by running a separate AD model alongside the primary model. However, this separation ignores the fact that the primary model already encodes substantial information about the target distribution. In this paper, we introduce On-Model AD, a setting for anomaly detection that explicitly leverages access to a related machine learning model. Wi