Conference · 2026
A Metacognitive Framework for Test-Time Adaptation: Taxonomy and Research Challenges
Abstract
In Test-Time Adaptation (TTA), when out-of-distribution sam-
ples interleave unpredictably with corrupted in-distribution
samples, adaptation without corrective measures is prone to
significant performance decrease. Existing protocols rely on
unreliable factors (e.g., predictive entropy) which are unable
to distinguish corrupted in-distribution data from semantically
novel open-set samples. To address this issue, we formal-
ize TTA through a two-level metacognitive framework that
decouples task execution from control, facilitating a unified
taxonomy that categorizes existing TTA protocols across three
core axes, i.e., which knowledge is monitored, which part of
the models are targeted for adaptation, and how adaptation
occurs. We provide evidence that robust corrective measures
in TTA require monitoring grounded in structured internal
representations to reliably estimate model competence. Lastly,
we frame monitoring as a resource-allocation problem, exam-
ine neuromorphic substrates for bounded self-regulation, and
outline key open challenges.
BibTeX
@inproceedings{loi2026metacognitive,
title = {A Metacognitive Framework for Test-Time Adaptation: Taxonomy and Research Challenges},
author = {Loi, Dario and Onorato, Gabriele and Coviello, Antonella and Rifat, Shahriar and Bastian, Nathaniel D. and Restuccia, Francesco},
booktitle = {Proceedings of the AAAI Fall Symposium on Artificial Metacognition (METACOG)},
series = {AAAI Fall Symposium Series},
year = {2026},
month = {November},
address = {Arlington, VA, USA},
note = {To appear},
}