{
  "abstract": "Deep neural networks often suffer significant accuracy degradation when exposed to real-world image corruptions and distribution shifts. To overcome the limitations of fixed, input-agnostic test-time augmentation (TTA), an adaptive framework is proposed that learns per-sample transformations via reinforcement learning. Augmentation selection is cast as a Markov decision process and proximal policy optimization (PPO) agents are trained to choose sample-specific transforms under a composite reward combining classifier confidence gains with a self-consistency KL-divergence penalty on the model’s own softmax outputs, thereby preserving overall belief stability. On clean CIFAR-10 (1 000 samples), the adaptive ensemble raises accuracy from 88.5% (baseline) and 87.3% (static TTA) to 90.0% (+1.5 pp). On CIFAR-10-C (15 corruptions $$\\times$$ 5 severities; 1 000 images per condition), pooled top-1 accuracy improves from 75.7% (baseline) and 74.3% (static TTA) to 76.4% (+0.7 pp), and exceeds a TENT entropy-minimization baseline (75.9%) while operating in a strictly label-free regime that updates no model weights. Per-corruption gains are consistently positive across noise, blur, weather, and compression distortions, with the adaptive policy outperforming TENT on texture and compression corruptions where input-space transforms are most effective. These findings demonstrate that learned, per-sample augmentation policies improve robustness and reliability of deep vision models under diverse image conditions, against a strong baseline classifier.",
  "abstract_shown": true,
  "affiliations": [
    "Department of Information Technology, Bharati Vidyapeeth’s College of Engineering",
    "Department of Computer Science and Engineering, Bharati Vidyapeeth’s College of Engineering",
    "School of Computer Science and Engineering, Galgotias University",
    "Department of Computer Science, Faculty of Science, University of South Bohemia in České Budějovice"
  ],
  "article_type": null,
  "arxiv": null,
  "author_total": 7,
  "authors": [
    {
      "affs": [
        0
      ],
      "name": "Tushar Mittal"
    },
    {
      "affs": [
        0
      ],
      "name": "Arun Kumar Dubey"
    },
    {
      "affs": [
        1
      ],
      "name": "Dharmender Saini"
    },
    {
      "affs": [
        0
      ],
      "name": "Sarita Yadav"
    },
    {
      "affs": [
        0
      ],
      "name": "Achin Jain"
    },
    {
      "affs": [
        2
      ],
      "name": "Arvind Panwar"
    },
    {
      "affs": [
        3
      ],
      "name": "Hamidreza Namazi"
    }
  ],
  "bibtex": "https://research.draconitic.com/paper/681b9efj5b0fh.bib",
  "cited_by": 1,
  "doi": "10.1038/s41598-026-62059-4",
  "fulltext_url": "https://doi.org/10.1038/s41598-026-62059-4",
  "id": "681b9efj5b0fh",
  "issns": [
    "2045-2322"
  ],
  "landing_url": "https://doi.org/10.1038/s41598-026-62059-4",
  "language": "en",
  "licence": "CC-BY",
  "licence_label": "CC BY",
  "licence_source": "journal",
  "licence_url": "https://creativecommons.org/licenses/by/4.0/",
  "metadata_licence": null,
  "native_subjects": [
    [
      "lcc",
      "Q"
    ],
    [
      "lcc",
      "R"
    ],
    [
      "openalex",
      "1702|Adversarial Robustness in Machine Learning"
    ]
  ],
  "openalex": "W7169716460",
  "pdf_url": "https://www.nature.com/articles/s41598-026-62059-4_reference.pdf",
  "pmcid": null,
  "pmid": null,
  "published": "2026-07-01",
  "publisher": "Nature Portfolio",
  "ris": "https://research.draconitic.com/paper/681b9efj5b0fh.ris",
  "sources": [
    {
      "first_seen": "2026-10-30T13:45:52Z",
      "native_id": "2893f7f23ac742d1a8c469102cb03250",
      "source": "doaj",
      "url": "https://doi.org/10.1038/s41598-026-62059-4"
    }
  ],
  "subjects": [
    {
      "field": "medicine",
      "origin": "lcc:R",
      "subfield": "general"
    }
  ],
  "title": "Adaptive test-time augmentation via KL-regularized reinforcement learning for robust visual inference",
  "title_text": "Adaptive test-time augmentation via KL-regularized reinforcement learning for robust visual inference",
  "type": "peer-reviewed",
  "updated_at": "2026-10-30T13:45:52Z",
  "url": "https://research.draconitic.com/paper/681b9efj5b0fh",
  "venue": "Scientific Reports",
  "versions": []
}