{"ID":2895721,"CreatedAt":"2026-06-01T04:54:23.091178241Z","UpdatedAt":"2026-06-01T04:54:23.091178241Z","DeletedAt":null,"paper_url":"https://arxiv.org/abs/2507.08607","arxiv_id":"2507.08607","title":"BayesTTA: Continual-Temporal Test-Time Adaptation for Vision-Language Models via Gaussian Discriminant Analysis","abstract":"Vision-language models (VLMs) such as CLIP achieve strong zero-shot recognition but degrade significantly under \\textit{temporally evolving distribution shifts} common in real-world scenarios (e.g., gradual illumination or seasonal changes). Existing continual test-time adaptation (CTTA) methods are typically built around sudden and severe distribution shifts and neglect temporal continuity, leading to three core defects: limited memory cache restricts long-range distribution modeling, causing catastrophic forgetting; entropy-based confidence becomes unreliable under temporal drift, worsening error accumulation; and static visual representations misalign with evolving inputs. We formalize this practical problem as \\textit{Continual-Temporal Test-Time Adaptation (CT-TTA)}, where test distributions evolve gradually over time. To address it, we propose \\textit{BayesTTA}, a Bayesian adaptation framework that enforces temporally consistent predictions and dynamically aligns visual representations. Specifically, BayesTTA incrementally estimates class-conditional Gaussian mixture distributions without storing raw data, adaptively selects covariance structures through statistical hypothesis testing, and performs calibrated inference using Gaussian discriminant analysis (GDA). These calibrated predictions supervise self-paced adaptation of normalization layers, ensuring efficient and stable representation alignment. We establish a comprehensive CT-TTA benchmark across four temporally evolving datasets and further evaluate generalization on ten standard TTA datasets. Extensive experiments show that BayesTTA consistently outperforms state-of-the-art methods, achieving significant gains while maintaining efficiency. Code is available at \\href{https://github.com/cuishuang99/BayesTTA}{https://github.com/cuishuang99/BayesTTA}.","short_abstract":"Vision-language models (VLMs) such as CLIP achieve strong zero-shot recognition but degrade significantly under \\textit{temporally evolving distribution shifts} common in real-world scenarios (e.g., gradual illumination or seasonal changes). Existing continual test-time adaptation (CTTA) methods are typically built aro...","url_abs":"https://arxiv.org/abs/2507.08607","url_pdf":"https://arxiv.org/pdf/2507.08607v1","authors":"[\"Shuang Cui\",\"Jinglin Xu\",\"Yi Li\",\"Xiongxin Tang\",\"Jiangmeng Li\",\"Jiahuan Zhou\",\"Fanjiang Xu\",\"Fuchun Sun\",\"Hui Xiong\"]","published":"2025-07-11T14:02:54Z","proceeding":"cs.CV","tasks":"[\"cs.CV\"]","methods":"[\"Language Model\"]","has_code":false,"code_links":[{"ID":612214,"CreatedAt":"2026-06-01T04:54:23.091178241Z","UpdatedAt":"2026-06-01T04:54:23.091178241Z","DeletedAt":null,"paper_id":2895721,"paper_url":"https://arxiv.org/abs/2507.08607","paper_title":"BayesTTA: Continual-Temporal Test-Time Adaptation for Vision-Language Models via Gaussian Discriminant Analysis","repo_url":"https://github.com/cuishuang99/BayesTTA","is_official":false,"mentioned_in_paper":false,"mentioned_in_github":true,"github_stars":0}]}