analiza aiqbench
This paper introduces Prediction-Powered Risk Monitoring (PPRM), a semi-supervised method that leverages synthetic labels alongside limited true labels to detect harmful distribution shifts in deployed models. By constructing anytime-valid lower bounds on running risk, the approach provides assumption-free finite-sample guarantees for Type-I error control, addressing the critical challenge of monitoring performance in dynamic environments with scarce labeled data. The technique demonstrates efficacy across diverse tasks, including image classification, large language models, and telecommunications monitoring.
Prediction-Powered Risk Monitoring of Deployed Models for Detecting Harmful Distribution Shifts
arXiv:2602.02229v2 Announce Type: replace Abstract: We study the problem of monitoring model performance in dynamic environments where labeled data are limited. To this end, we propose prediction-powered risk monitoring (PPRM), a semi-supervised risk-monitoring approach based on prediction-powered inference (PPI). PPRM constructs anytime-valid lower bounds on the running risk by combining synthetic labels with a small set of true labels. Harmful shifts are detected via a threshold-based comparison with an upper bound on the nominal risk, satisfying assumption-free finite-sample guarantees on the type-I error. We demonstrate the effectiveness of PPRM through extensive experiments on image classification, large language model (LLM), and telecommunications monitoring tasks.
Abstract:We study the problem of monitoring model performance in dynamic environments where labeled data are limited. To this end, we propose prediction-powered risk monitoring (PPRM), a semi-supervised risk-monitoring approach based on prediction-powered inference (PPI). PPRM constructs anytime-valid lower bounds on the running risk by combining synthetic labels with a small set of true labels. Harmful shifts are detected via a threshold-based comparison with an upper bound on the nominal risk, satisfying assumption-free finite-sample guarantees on the type-I error. We demonstrate the effectiveness of PPRM through extensive experiments on image classification, large language model (LLM), and telecommunications monitoring tasks.
Machine Learning (cs.LG); Signal Processing (eess.SP)
arXiv:2602.02229 [cs.LG]
(or arXiv:2602.02229v2 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2602.02229 Focus to learn more arXiv-issued DOI via DataCite
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 193 cuvinte