Pointwise Metrics Mislead: An Evaluation Protocol for Multimodal Inverse Problems

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Baattrup, Mads H., Bach, Jörn, Jeppe, Laurids, Labe, Finn, Grohsjean, Alexander, Schwanenberger, Christian, Stelldinger, Peer
Format: Preprint
Veröffentlicht: 2026
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866916037785550848
author Baattrup, Mads H.
Bach, Jörn
Jeppe, Laurids
Labe, Finn
Grohsjean, Alexander
Schwanenberger, Christian
Stelldinger, Peer
author_facet Baattrup, Mads H.
Bach, Jörn
Jeppe, Laurids
Labe, Finn
Grohsjean, Alexander
Schwanenberger, Christian
Stelldinger, Peer
contents Evaluation in scientific reconstruction is dominated by pointwise metrics - RMSE, MAE, per-event resolution - under the implicit assumption that lower error means better reconstruction. We show that this assumption fails structurally for inverse problems with multimodal posteriors. By the law of total variance, point estimators trained to minimize MSE or MAE produce a marginal spectrum strictly narrower than the truth whenever the posterior has nonzero width. The resulting bias is independent of architecture, training, and dataset size, and it compresses precisely the spectral features - tails, modes, shapes - that downstream scientific measurements rely on. We propose a three-part evaluation protocol where each step targets a failure mode the others miss: per-event distributional accuracy via CRPS, population-level marginal accuracy via a spectrum-fidelity diagnostic, and uncertainty trustworthiness via coverage-based calibration. On a synthetic benchmark with an analytic posterior and on a realistic many-to-one inverse problem from particle physics, model rankings reverse between pointwise and distributional metrics, and calibration further separates architectures indistinguishable under CRPS. The evaluation protocol, not the model, determines the scientific conclusion.
format Preprint
id arxiv_https___arxiv_org_abs_2605_22891
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Pointwise Metrics Mislead: An Evaluation Protocol for Multimodal Inverse Problems
Baattrup, Mads H.
Bach, Jörn
Jeppe, Laurids
Labe, Finn
Grohsjean, Alexander
Schwanenberger, Christian
Stelldinger, Peer
Machine Learning
High Energy Physics - Experiment
Evaluation in scientific reconstruction is dominated by pointwise metrics - RMSE, MAE, per-event resolution - under the implicit assumption that lower error means better reconstruction. We show that this assumption fails structurally for inverse problems with multimodal posteriors. By the law of total variance, point estimators trained to minimize MSE or MAE produce a marginal spectrum strictly narrower than the truth whenever the posterior has nonzero width. The resulting bias is independent of architecture, training, and dataset size, and it compresses precisely the spectral features - tails, modes, shapes - that downstream scientific measurements rely on. We propose a three-part evaluation protocol where each step targets a failure mode the others miss: per-event distributional accuracy via CRPS, population-level marginal accuracy via a spectrum-fidelity diagnostic, and uncertainty trustworthiness via coverage-based calibration. On a synthetic benchmark with an analytic posterior and on a realistic many-to-one inverse problem from particle physics, model rankings reverse between pointwise and distributional metrics, and calibration further separates architectures indistinguishable under CRPS. The evaluation protocol, not the model, determines the scientific conclusion.
title Pointwise Metrics Mislead: An Evaluation Protocol for Multimodal Inverse Problems
topic Machine Learning
High Energy Physics - Experiment
url https://arxiv.org/abs/2605.22891