# Back to the Source: Diffusion-Driven Test-Time Adaptation Source: [Back to the Source: Diffusion-Driven Test-Time Adaptation]() Collection index: [llms.txt]() Data updated: 2026-01-04T05:23:04.931Z Authors: Jin Gao, Jialing Zhang, Xihui Liu, Trevor Darrell, Evan Shelhamer, Dequan Wang Published: 2022-07-07T17:14:10.000Z ## Abstract Test-time adaptation harnesses test inputs to improve the accuracy of a model trained on source data when tested on shifted target data. Existing methods update the source model by (re-)training on each target domain. While effective, re-training is sensitive to the amount and order of the data and the hyperparameters for optimization. We instead update the target data, by projecting all test inputs toward the source domain with a generative diffusion model. Our diffusion-driven adaptation method, DDA, shares its models for classification and generation across all domains. Both models are trained on the source domain, then fixed during testing. We augment diffusion with image guidance and self-ensembling to automatically decide how much to adapt. Input adaptation by DDA is more robust than prior model adaptation approaches across a variety of corruptions, architectures, and data regimes on the ImageNet-C benchmark. With its input-wise updates, DDA succeeds where model adaptation degrades on too little data in small batches, dependent data in non-uniform order, or mixed data with multiple corruptions. ## Recorded benchmark results A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage. - [Image Classification on ImageNet-C-bar level 5 1.0](): Image Classification; ImageNet-C-bar level 5 1.0; Accuracy: 46.3. - [Lesion Segmentation on In-house Topcon → Spectralis (test)](): Lesion Segmentation; In-house Topcon → Spectralis (test); GA DSC: 53.2. - [Retinal Fluid Segmentation on RETOUCH Cirrus → Spectralis (test)](): Lesion Segmentation; RETOUCH Cirrus → Spectralis (test); IRF: 50. - [Lesion Segmentation on RETOUCH Topcon → Spectralis (test)](): Lesion Segmentation; RETOUCH Topcon → Spectralis (test); Intraretinal Fluid Score: 38. - [Image Classification on ImageNet-C (level 5)](): Image Classification; ImageNet-C level 5; Avg Top-1 Acc (ImageNet-C L5): 49.4. ## Original sources and implementation [Original paper]() [PDF]()