Order-Aware Test-Time Adaptation: Leveraging Temporal Dynamics for Robust Streaming Inference

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Kim, Young Kyung, Schlesinger, Oded, Wu, Qiangqiang, Di Martino, J. Matías, Sapiro, Guillermo
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910004288684032
author Kim, Young Kyung
Schlesinger, Oded
Wu, Qiangqiang
Di Martino, J. Matías
Sapiro, Guillermo
author_facet Kim, Young Kyung
Schlesinger, Oded
Wu, Qiangqiang
Di Martino, J. Matías
Sapiro, Guillermo
contents Test-Time Adaptation (TTA) enables pre-trained models to adjust to distribution shift by learning from unlabeled test-time streams. However, existing methods typically treat these streams as independent samples, overlooking the supervisory signal inherent in temporal dynamics. To address this, we introduce Order-Aware Test-Time Adaptation (OATTA). We formulate test-time adaptation as a gradient-free recursive Bayesian estimation task, using a learned dynamic transition matrix as a temporal prior to refine the base model's predictions. To ensure safety in weakly structured streams, we introduce a likelihood-ratio gate (LLR) that reverts to the base predictor when temporal evidence is absent. OATTA is a lightweight, model-agnostic module that incurs negligible computational overhead. Extensive experiments across image classification, wearable and physiological signal analysis, and language sentiment analysis demonstrate its universality; OATTA consistently boosts established baselines, improving accuracy by up to 6.35%. Our findings establish that modeling temporal dynamics provides a critical, orthogonal signal beyond standard order-agnostic TTA approaches.
format Preprint
id arxiv_https___arxiv_org_abs_2601_21012
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Order-Aware Test-Time Adaptation: Leveraging Temporal Dynamics for Robust Streaming Inference
Kim, Young Kyung
Schlesinger, Oded
Wu, Qiangqiang
Di Martino, J. Matías
Sapiro, Guillermo
Machine Learning
Test-Time Adaptation (TTA) enables pre-trained models to adjust to distribution shift by learning from unlabeled test-time streams. However, existing methods typically treat these streams as independent samples, overlooking the supervisory signal inherent in temporal dynamics. To address this, we introduce Order-Aware Test-Time Adaptation (OATTA). We formulate test-time adaptation as a gradient-free recursive Bayesian estimation task, using a learned dynamic transition matrix as a temporal prior to refine the base model's predictions. To ensure safety in weakly structured streams, we introduce a likelihood-ratio gate (LLR) that reverts to the base predictor when temporal evidence is absent. OATTA is a lightweight, model-agnostic module that incurs negligible computational overhead. Extensive experiments across image classification, wearable and physiological signal analysis, and language sentiment analysis demonstrate its universality; OATTA consistently boosts established baselines, improving accuracy by up to 6.35%. Our findings establish that modeling temporal dynamics provides a critical, orthogonal signal beyond standard order-agnostic TTA approaches.
title Order-Aware Test-Time Adaptation: Leveraging Temporal Dynamics for Robust Streaming Inference
topic Machine Learning
url https://arxiv.org/abs/2601.21012