LOCUS: A Distribution-Free Loss-Quantile Score for Risk-Aware Predictions
This addresses the need for risk-aware predictions in machine learning deployment, offering a practical tool for flagging high-loss events, though it is incremental as it builds on existing predictive distribution methods.
The paper tackles the problem of machine learning models making costly mistakes despite average accuracy by introducing Locus, a distribution-free wrapper that produces a per-input loss-scale reliability score, which in experiments across 13 regression benchmarks effectively ranks risk and reduces large-loss frequency compared to standard heuristics.
Modern machine learning models can be accurate on average yet still make mistakes that dominate deployment cost. We introduce Locus, a distribution-free wrapper that produces a per-input loss-scale reliability score for a fixed prediction function. Rather than quantifying uncertainty about the label, Locus models the realized loss of the prediction function using any engine that outputs a predictive distribution for the loss given an input. A simple split-calibration step turns this function into a distribution-free interpretable score that is comparable across inputs and can be read as an upper loss level. The score is useful on its own for ranking, and it can optionally be thresholded to obtain a transparent flagging rule with distribution-free control of large-loss events. Experiments across 13 regression benchmarks show that Locus yields effective risk ranking and reduces large-loss frequency compared to standard heuristics.