Gradual Uncertainty Refinement via Noise-Driven Curriculum: A Post-Hoc Meta-Model for Robust Uncertainty Quantification

Charmaine Barker, Daniel Bethell, Simos Gerasimou
Proceedings of the 42nd Conference on Uncertainty in Artificial Intelligence, PMLR 337:413-447, 2026.

Abstract

Reliable uncertainty quantification remains a major obstacle to the deployment of deep learning models under distributional shift. Existing post-hoc approaches that retrofit pretrained models either inherit misplaced confidence or merely reshape predictions, without advising the model when to be uncertain. We introduce {GUIDE}, a lightweight evidential learning meta-model that attaches to a frozen deep learning model and is explicitly guided on when and how to be uncertain. {GUIDE} identifies salient internal features via a calibration stage and then uses these features to construct a noise-driven curriculum that teaches the model when and how to express uncertainty. {GUIDE} requires no retraining, no architectural modifications, and no manual intermediate-layer selection to the base deep learning model, thus ensuring broad applicability and minimal user intervention. The resulting model avoids distilling overconfidence from the base model, improves out-of-distribution detection ($\approx$ 77%) and adversarial attack detection ($\approx$ 80%), while preserving in-distribution performance. Across diverse benchmarks, {GUIDE} consistently outperforms state-of-the-art approaches, evidencing the need for actively guiding uncertainty to close the gap between predictive confidence and reliability.

Cite this Paper

BibTeX
@InProceedings{pmlr-v337-barker26a, title = {Gradual Uncertainty Refinement via Noise-Driven Curriculum: A Post-Hoc Meta-Model for Robust Uncertainty Quantification}, author = {Barker, Charmaine and Bethell, Daniel and Gerasimou, Simos}, booktitle = {Proceedings of the 42nd Conference on Uncertainty in Artificial Intelligence}, pages = {413--447}, year = {2026}, editor = {Perković, Emilija and Malinsky, Daniel}, volume = {337}, series = {Proceedings of Machine Learning Research}, month = {17--21 Aug}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v337/main/assets/barker26a/barker26a.pdf}, url = {https://proceedings.mlr.press/v337/barker26a.html}, abstract = {Reliable uncertainty quantification remains a major obstacle to the deployment of deep learning models under distributional shift. Existing post-hoc approaches that retrofit pretrained models either inherit misplaced confidence or merely reshape predictions, without advising the model when to be uncertain. We introduce {GUIDE}, a lightweight evidential learning meta-model that attaches to a frozen deep learning model and is explicitly guided on when and how to be uncertain. {GUIDE} identifies salient internal features via a calibration stage and then uses these features to construct a noise-driven curriculum that teaches the model when and how to express uncertainty. {GUIDE} requires no retraining, no architectural modifications, and no manual intermediate-layer selection to the base deep learning model, thus ensuring broad applicability and minimal user intervention. The resulting model avoids distilling overconfidence from the base model, improves out-of-distribution detection ($\approx$ 77%) and adversarial attack detection ($\approx$ 80%), while preserving in-distribution performance. Across diverse benchmarks, {GUIDE} consistently outperforms state-of-the-art approaches, evidencing the need for actively guiding uncertainty to close the gap between predictive confidence and reliability.} }
Endnote
%0 Conference Paper %T Gradual Uncertainty Refinement via Noise-Driven Curriculum: A Post-Hoc Meta-Model for Robust Uncertainty Quantification %A Charmaine Barker %A Daniel Bethell %A Simos Gerasimou %B Proceedings of the 42nd Conference on Uncertainty in Artificial Intelligence %C Proceedings of Machine Learning Research %D 2026 %E Emilija Perković %E Daniel Malinsky %F pmlr-v337-barker26a %I PMLR %P 413--447 %U https://proceedings.mlr.press/v337/barker26a.html %V 337 %X Reliable uncertainty quantification remains a major obstacle to the deployment of deep learning models under distributional shift. Existing post-hoc approaches that retrofit pretrained models either inherit misplaced confidence or merely reshape predictions, without advising the model when to be uncertain. We introduce {GUIDE}, a lightweight evidential learning meta-model that attaches to a frozen deep learning model and is explicitly guided on when and how to be uncertain. {GUIDE} identifies salient internal features via a calibration stage and then uses these features to construct a noise-driven curriculum that teaches the model when and how to express uncertainty. {GUIDE} requires no retraining, no architectural modifications, and no manual intermediate-layer selection to the base deep learning model, thus ensuring broad applicability and minimal user intervention. The resulting model avoids distilling overconfidence from the base model, improves out-of-distribution detection ($\approx$ 77%) and adversarial attack detection ($\approx$ 80%), while preserving in-distribution performance. Across diverse benchmarks, {GUIDE} consistently outperforms state-of-the-art approaches, evidencing the need for actively guiding uncertainty to close the gap between predictive confidence and reliability.
APA
Barker, C., Bethell, D. & Gerasimou, S.. (2026). Gradual Uncertainty Refinement via Noise-Driven Curriculum: A Post-Hoc Meta-Model for Robust Uncertainty Quantification. Proceedings of the 42nd Conference on Uncertainty in Artificial Intelligence, in Proceedings of Machine Learning Research 337:413-447 Available from https://proceedings.mlr.press/v337/barker26a.html.

Related Material