ReservoirTTA: Prolonged Test-time Adaptation for Evolving and Recurring Domains

2025-05-20

AI Summary

This paper addresses long-term test-time adaptation (TTA) under persistent non-stationary test distribution shifts—such as recurrent or gradually evolving domain shifts. To overcome inherent limitations of single-model TTA, including catastrophic forgetting, inter-domain interference, and error accumulation, we propose a multi-model dynamic reservoir framework. Our method introduces an online style-feature clustering–driven, sample-level domain-aware mechanism that enables dynamic reservoir construction, domain-adaptive routing, and plug-and-play parameter updates, augmented by variance-constrained theoretical optimization. Evaluated on ImageNet-C, CIFAR-C, and Cityscapes→ACDC benchmarks, our approach significantly improves both accuracy and stability in long-term TTA. It is the first to achieve robust, continuous adaptation to recurrent and evolving domain shifts, outperforming existing state-of-the-art methods.

Technology Category

• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Computer Vision: Adversarial Attacks & Robustness
• Cognitive Modeling & Cognitive Systems: Adaptive Behavior

Application Category

• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems

Abstract

This paper introduces ReservoirTTA, a novel plug-in framework designed for prolonged test-time adaptation (TTA) in scenarios where the test domain continuously shifts over time, including cases where domains recur or evolve gradually. At its core, ReservoirTTA maintains a reservoir of domain-specialized models -- an adaptive test-time model ensemble -- that both detects new domains via online clustering over style features of incoming samples and routes each sample to the appropriate specialized model, and thereby enables domain-specific adaptation. This multi-model strategy overcomes key limitations of single model adaptation, such as catastrophic forgetting, inter-domain interference, and error accumulation, ensuring robust and stable performance on sustained non-stationary test distributions. Our theoretical analysis reveals key components that bound parameter variance and prevent model collapse, while our plug-in TTA module mitigates catastrophic forgetting of previously encountered domains. Extensive experiments on the classification corruption benchmarks, including ImageNet-C and CIFAR-10/100-C, as well as the Cityscapes$ ightarrow$ACDC semantic segmentation task, covering recurring and continuously evolving domain shifts, demonstrate that ReservoirTTA significantly improves adaptation accuracy and maintains stable performance across prolonged, recurring shifts, outperforming state-of-the-art methods.

Problem

• Adapts models to continuously shifting test domains
• Prevents catastrophic forgetting in evolving domains
• Ensures robust performance in non-stationary distributions

Innovation

• ReservoirTTA uses adaptive model ensemble for domain shifts
• Online clustering detects and routes samples to specialized models
• Multi-model strategy prevents catastrophic forgetting and interference