GeneralAI / InformaticsResearch

Richer prompts increase LLM confidence but not severity scores in salivary gland imaging

European archives of oto-rhino-laryngology : official journal of the European Federation of Oto-Rhino-Laryngological Societies (EUFOS) : affiliated with the German Society for Oto-Rhino-Laryngology - Head and Neck Surgery3w ago

Three large language models rated salivary gland severity similarly, but self-reported confidence rose with richer prompts (odds ratio 5.43, 95% CI 3.00-9.81). Output detail varied; the text-only design did not assess image interpretation.

  • Severity scores did not significantly differ among ChatGPT, Gemini, and Claude (χ²=2.31, P=0.31).
  • Claude reported more additional structures than ChatGPT (mean 5.67 vs 0.00); AI system significantly affected reporting verbosity (P<0.001).
  • The authors note the design does not evaluate image interpretation, so increased confidence may reflect richer input rather than improved diagnostic reasoning.
Read the source

Automated summary

RadPigeon summaries are original and for information only. They are not clinical advice.