Back to Source: Open-Set Continual Test-Time Adaptation via Domain Compensation

2026-04-23

AI Summary

This work addresses the dual challenges of domain shift and unknown-class samples in open-set continual test-time adaptation (OCTTA) by proposing the DOCO framework. DOCO dynamically partitions incoming samples at inference time to distinguish between known and unknown instances, learns source-aligned domain-compensating prompts, and propagates these prompts to unknown samples, thereby establishing a closed-loop synergy between domain adaptation and out-of-distribution (OOD) detection. A structure-preserving regularization is introduced to mitigate semantic distortion and effectively isolate the semantic novelty of unknown samples. Extensive experiments demonstrate that DOCO significantly outperforms existing continual TTA (CTTA) and open-set TTA (OSTTA) methods across multiple benchmarks, achieving joint optimization of both tasks for the first time under the OCTTA setting and establishing a new state of the art.

Technology Category

• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Search and Optimization: Mixed Discrete/Continuous Search
• Constraint Satisfaction and Optimization: Distributed CSP/Optimization

Application Category

• Economics, Online Markets and Human Computation: Humans versus LLMs for data annotation and labeling
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation

Abstract

Test-Time Adaptation (TTA) aims to mitigate distributional shifts between training and test domains during inference time. However, existing TTA methods fall short in the realistic scenario where models face both continually changing domains and the simultaneous emergence of unknown semantic classes, a challenging setting we term Open-set Continual Test-Time Adaptation (OCTTA). The coupling of domain and semantic shifts often collapses the feature space, severely degrading both classification and out-of-distribution detection. To tackle this, we propose DOmain COmpensation (DOCO), a lightweight and effective framework that robustly performs domain adaptation and OOD detection in a synergistic, closed loop. DOCO first performs dynamic, adaptation-conditioned sample splitting to separate likely ID from OOD samples. Then, using only the ID samples, it learns a domain compensation prompt by aligning feature statistics with the source domain, guided by a structural preservation regularizer that prevents semantic distortion. This learned prompt is then propagated to the OOD samples within the same batch, effectively isolating their semantic novelty for more reliable detection. Extensive experiments on multiple challenging benchmarks demonstrate that DOCO outperforms prior CTTA and OSTTA methods, establishing a new state-of-the-art for the demanding OCTTA setting.

Problem

• Test-Time Adaptation
• Open-Set Recognition
• Continual Domain Shift
• Out-of-Distribution Detection
• Distributional Shift

Innovation

• Open-set Continual Test-Time Adaptation
• Domain Compensation
• Test-Time Adaptation
• Out-of-Distribution Detection
• Feature Alignment