analiza aiqbench
This paper introduces novel PAC-Bayesian certificates to rigorously decide whether to adapt machine learning models during test time, distinguishing between radial and direction-sensitive risk assessments. The authors demonstrate that their directional approach can safely accept beneficial updates while rejecting harmful ones even when source performance degrades, providing formal guarantees for deployment under distribution shifts.
Certifying Whether to Adapt: Baseline-Relative PAC-Bayes-MMD Guarantees for Test-Time Adaptation
arXiv:2605.21783v2 Announce Type: replace Abstract: Test-time adaptation (TTA) needs a decision about an update, not only an absolute-risk estimate. We give PAC-Bayesian certificates for the target excess risk of a candidate relative to a pilot-fixed baseline. A radial certificate uses an MMD ambiguity radius; a direction-sensitive certificate retains the sign of the empirical RKHS shift and can therefore certify target improvement even when source risk increases. For finite binary state spaces we derive statewise marginal and signed conditional-shift refinements, with penalties localized to candidate--baseline disagreements. A source-worsening update with target excess risk $-0.576$ is accepted in all 1,000 repetitions by the directional certificate and never by the radial certificate; learned finite-state updates with excess risks $-0.800$ and $-0.401$ are likewise accepted in every repetition. On Covertype, the certificate safely rejects all updates and blocks geographic-shift adaptations whose unconditional deployment increases risk by up to $2.75$ percentage points in prespecified base libraries; an exploratory stress test reaches $5.66$ points. A feature-quantized CIFAR-10-C and ImageNet-C study is structurally vacuous and is reported only as a diagnostic limitation, not as a practical neural certificate.
Abstract:Test-time adaptation (TTA) needs a decision about an update, not only an absolute-risk estimate. We give PAC-Bayesian certificates for the target excess risk of a candidate relative to a pilot-fixed baseline. A radial certificate uses an MMD ambiguity radius; a direction-sensitive certificate retains the sign of the empirical RKHS shift and can therefore certify target improvement even when source risk increases. For finite binary state spaces we derive statewise marginal and signed conditional-shift refinements, with penalties localized to candidate--baseline disagreements. A source-worsening update with target excess risk $-0.576$ is accepted in all 1,000 repetitions by the directional certificate and never by the radial certificate; learned finite-state updates with excess risks $-0.800$ and $-0.401$ are likewise accepted in every repetition. On Covertype, the certificate safely rejects all updates and blocks geographic-shift adaptations whose unconditional deployment increases risk by up to $2.75$ percentage points in prespecified base libraries; an exploratory stress test reaches $5.66$ points. A feature-quantized CIFAR-10-C and ImageNet-C study is structurally vacuous and is reported only as a diagnostic limitation, not as a practical neural certificate.
Machine Learning (cs.LG); Machine Learning (stat.ML)
arXiv:2605.21783 [cs.LG]
(or arXiv:2605.21783v2 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2605.21783 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 · 271 cuvinte