Saved in:
Bibliographic Details
Main Authors: Mohammadi-Seif, Abolfazl, Baeza-Yates, Ricardo
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2605.02544
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911644147253248
author Mohammadi-Seif, Abolfazl
Baeza-Yates, Ricardo
author_facet Mohammadi-Seif, Abolfazl
Baeza-Yates, Ricardo
contents The widespread adoption of machine learning in critical applications demands techniques to mitigate high-consequence errors. Our method utilizes a dual-classifier GBDT pipeline to distinguish routine human-like errors from high-risk non-human misclassifications. Evaluated across three domains, animal breed classification, skin lesion diagnosis (ISIC 2018), and prostate histopathology (SICAPv2), our framework demonstrates robust safety improvements. To address real-world deployment concerns, our results confirm the pipeline introduces negligible inference latency (1.60% overhead for the animal dataset, 1.84% for ISIC, and 1.70% for SICAPv2) while outperforming traditional Maximum Class Probability (MCP) baselines in correction precision. Our conservative correction strategy successfully reduced dangerous non-human errors by 34.1% in ISIC and 12.57% in SICAPv2, improving super-class diagnostic safety to 90.41% and 92.13% respectively. This proves that safety-critical reliability can be substantially enhanced post-hoc without expensive model retraining. keywords: Error Analysis, Post-hoc Correction, Trustworthy AI.
format Preprint
id arxiv_https___arxiv_org_abs_2605_02544
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Improving Model Safety by Targeted Error Correction
Mohammadi-Seif, Abolfazl
Baeza-Yates, Ricardo
Artificial Intelligence
Computer Vision and Pattern Recognition
The widespread adoption of machine learning in critical applications demands techniques to mitigate high-consequence errors. Our method utilizes a dual-classifier GBDT pipeline to distinguish routine human-like errors from high-risk non-human misclassifications. Evaluated across three domains, animal breed classification, skin lesion diagnosis (ISIC 2018), and prostate histopathology (SICAPv2), our framework demonstrates robust safety improvements. To address real-world deployment concerns, our results confirm the pipeline introduces negligible inference latency (1.60% overhead for the animal dataset, 1.84% for ISIC, and 1.70% for SICAPv2) while outperforming traditional Maximum Class Probability (MCP) baselines in correction precision. Our conservative correction strategy successfully reduced dangerous non-human errors by 34.1% in ISIC and 12.57% in SICAPv2, improving super-class diagnostic safety to 90.41% and 92.13% respectively. This proves that safety-critical reliability can be substantially enhanced post-hoc without expensive model retraining. keywords: Error Analysis, Post-hoc Correction, Trustworthy AI.
title Improving Model Safety by Targeted Error Correction
topic Artificial Intelligence
Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2605.02544