C-EDL：基于冲突感知证据深度学习实现鲁棒对抗量化

View PDF HTML (experimental)

Abstract:Reliability of deep learning models is critical for deployment in high-stakes applications, where out-of-distribution or adversarial inputs may lead to detrimental outcomes. Evidential Deep Learning, an efficient paradigm for uncertainty quantification, models predictions as Dirichlet distributions of a single forward pass. However, EDL is particularly vulnerable to adversarially perturbed inputs, making overconfident errors. Conflict-aware Evidential Deep Learning~\mbox{(C-EDL)} is a lightweight post-hoc uncertainty quantification approach that mitigates these issues, enhancing adversarial and OOD robustness without retraining. C-EDL generates diverse, task-preserving transformations per input and quantifies representational disagreement to calibrate uncertainty estimates when needed. C-EDL's conflict-aware prediction adjustment improves detection of OOD and adversarial inputs, maintaining high in-distribution accuracy and low computational overhead. Our experimental evaluation shows that C-EDL significantly outperforms state-of-the-art EDL variants and competitive baselines, achieving substantial reductions in coverage for OOD data (up to $\approx55\%$) and adversarial data (up to $\approx90\%$), across a range of datasets, attack types, and uncertainty metrics.

Comments:	Updated to the published ICLR 2026 version, including revised title. Published version: this https URL
Subjects:	Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
Cite as:	arXiv:2506.05937 [cs.LG]
	(or arXiv:2506.05937v3 [cs.LG] for this version)
	https://doi.org/10.48550/arXiv.2506.05937 arXiv-issued DOI via DataCite

Submission history

From: Charmaine Barker [view email]
[v1] Fri, 6 Jun 2025 10:06:23 UTC (8,267 KB)
[v2] Wed, 4 Mar 2026 11:10:18 UTC (7,416 KB)
[v3] Fri, 2 Oct 2026 09:37:34 UTC (7,416 KB)