EUVIMEDEUVIMED entdecken
EvaluationsstudieOpen Access

Model and Task-Aware Test-Time Scaling Strategies for Large Language and Vision-Language Models in Medicine: Evaluation Study

Gyutaek Oh, Seoyeon Kim, Sangjoon Park, Byung-Hoon Kim

Journal of Medical Internet Research · 23.07.2026 · 28 · e90693-e90693

DOI: 10.2196/90693 ↗

Kurz gefasst

Evaluationsstudie in der Zeitschrift Journal of Medical Internet Research, online erschienen am 23.07.2026. Verfasst von Gyutaek Oh und 3 weiteren Autor:innen. Beteiligte Einrichtung: Department of Biomedical Systems Informatics, College of Medicine, Yonsei University, 50-1 Yonsei-ro, Seodaemun-gu, Seoul, 03722, Republic of Korea, …. Hauptthemen laut MEDLINE: Large Language Models, Medical Informatics. Die Arbeit wurde bisher 2-mal zitiert (Crossref) und nennt 63 Literaturangaben. Der Volltext ist frei in PubMed Central verfügbar.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Background

Test-time scaling has emerged as a promising method to enhance the reasoning capabilities of large language models (LLMs) and vision-language models (VLMs) during inference without additional training. While foundational studies established scaling paradigms in general domains, their applicability to the unique complexities of medical AI remains underexplored.

Objective

This study aims to conduct a comprehensive investigation of test-time scaling in the medical domain. We evaluate the impact of scaling across different model sizes and task complexities. Furthermore, we seek to identify domain-specific bottlenecks and assess model robustness against user-driven perturbations, such as misleading clinical authority.

Methods

This study evaluated a diverse set of general and medical-specific LLMs and VLMs. Experiments used five textual medical benchmarks comprising over 5500 questions and two multimodal benchmarks comprising 7000 samples. Performance was measured under three scaling conditions: increasing token budgets, iterative sequential scaling, and parallel scaling. Robustness was tested by embedding misleading hints with varying tones and levels of simulated clinical expertise into prompts.

Results

For nonreasoning LLMs, accuracy saturated quickly, with token usage often remaining under 500 tokens regardless of budget increases. Reasoning models demonstrated significant performance gains on complex tasks as token budgets increased. Notably, we identified distinct domain-specific behaviors. First, current VLMs showed a structural bottleneck in integrating visual clues and experienced limited benefit from token expansion. Second, medically fine-tuned LLMs excelled in clinical question answering but exhibited degraded scaling efficiency on calculation tasks compared to general-domain models. This reflects a disparity between qualitative clinical alignment and procedural logic. Third, while optimal scaling improved robustness, models exhibited a cognitive vulnerability by readily abandoning correct reasoning when confronted with misleading expert physician hints. Regarding scaling strategies, parallel scaling outperformed sequential scaling on easier tasks. Conversely, extended sequential scaling or increased budgets proved essential for complex problem-solving.

Conclusions

Test-time scaling rules from general domains do not perfectly translate to medical AI. Longer reasoning is not universally beneficial. Concise reasoning with parallel scaling is optimal for simpler tasks. An extended chain of thought via sequential scaling or increased budgets is required for complex problems. Furthermore, safe clinical deployment requires addressing fundamental vision-language alignment, balancing clinical and procedural reasoning, and mitigating vulnerabilities to perceived clinical authority.

Schlagwörter

MeSH-Schlagwörter (MEDLINE)

Large Language ModelsMedical InformaticsHumans

Stichwörter der Autor:innen

clincal reasoninglarge language modelsmedical AIreasoning modelstest-time scalingvision-language models

Autor:innen und Institutionen

  • Department of Biomedical Systems Informatics, College of Medicine, Yonsei University, 50-1 Yonsei-ro, Seodaemun-gu, Seoul, 03722, Republic of Korea, 82 2-2228-2484.
  • Yonsei Institute for Digital Health, Seoul, Republic of Korea.
  • College of Medicine, Yonsei University, Seoul, Republic of Korea.
  • Department of Radiation Oncology, College of Medicine, Yonsei University, Seoul, Republic of Korea.
  • Heavy Ion Therapy Research Institute, Yonsei University College of Medicine, Seoul, Republic of Korea.
  • Yonsei Cancer Center, Yonsei University College of Medicine, Seoul, Republic of Korea.

Zitieren

APA

Oh, G., Kim, S., Park, S., & Kim, B. H. (2026). Model and Task-Aware Test-Time Scaling Strategies for Large Language and Vision-Language Models in Medicine: Evaluation Study. Journal of Medical Internet Research, 28, e90693-e90693. https://doi.org/10.2196/90693

Vancouver

Oh G, Kim S, Park S, Kim BH. Model and Task-Aware Test-Time Scaling Strategies for Large Language and Vision-Language Models in Medicine: Evaluation Study. Journal of Medical Internet Research. 2026;28:e90693-e90693. doi:10.2196/90693

BibTeX
@article{Oh2026,
  author = {Oh, Gyutaek and Kim, Seoyeon and Park, Sangjoon and Kim, Byung-Hoon},
  title = {Model and Task-Aware Test-Time Scaling Strategies for Large Language and Vision-Language Models in Medicine: Evaluation Study},
  journal = {Journal of Medical Internet Research},
  year = {2026},
  volume = {28},
  pages = {e90693--e90693},
  doi = {10.2196/90693}
}
Kostenlos starten →