GPT-4.1 and Llama 3.3 70 fail to detect clinically relevant errors in radiology reports in zero-shot evaluation.

Saved in:
Bibliographic Details
Title: GPT-4.1 and Llama 3.3 70 fail to detect clinically relevant errors in radiology reports in zero-shot evaluation.
Authors: Akinci D'Antonoli T; Division of Diagnostic and Interventional Neuroradiology, Department of Radiology, University Hospital Basel, Basel, Switzerland. tugba.akincidantonoli@unibas.ch.; Department of Pediatric Radiology, University Children's Hospital Basel, Basel, Switzerland. tugba.akincidantonoli@unibas.ch., Adams LC; Institute of Diagnostic and Interventional Radiology, TUM University Hospital, TUM School of Medicine and Health, Technical University of Munich, Munich, Germany., Lübberstedt J; Institute of Diagnostic and Interventional Radiology, TUM University Hospital, TUM School of Medicine and Health, Technical University of Munich, Munich, Germany., Graf MM; Institute of Diagnostic and Interventional Radiology, TUM University Hospital, TUM School of Medicine and Health, Technical University of Munich, Munich, Germany., Mertens CJ; Institute of Diagnostic and Interventional Radiology, TUM University Hospital, TUM School of Medicine and Health, Technical University of Munich, Munich, Germany., Busch F; Institute of Diagnostic and Interventional Radiology, TUM University Hospital, TUM School of Medicine and Health, Technical University of Munich, Munich, Germany., Ziegelmayer S; Institute of Diagnostic and Interventional Radiology, TUM University Hospital, TUM School of Medicine and Health, Technical University of Munich, Munich, Germany., Makowski MR; Institute of Diagnostic and Interventional Radiology, TUM University Hospital, TUM School of Medicine and Health, Technical University of Munich, Munich, Germany., Bressem K; Institute of Diagnostic and Interventional Radiology, TUM University Hospital, TUM School of Medicine and Health, Technical University of Munich, Munich, Germany.; Department of Cardiovascular Radiology and Nuclear Medicine, German Heart Center Munich, TUM University Hospital, Munich, Germany., Luiken I; Institute of Diagnostic and Interventional Radiology, TUM University Hospital, TUM School of Medicine and Health, Technical University of Munich, Munich, Germany.
Source: European radiology [Eur Radiol] 2026 Jun 19. Date of Electronic Publication: 2026 Jun 19.
Publication Type: Journal Article
Journal Info: Publisher: Springer International Country of Publication: Germany NLM ID: 9114774 Publication Model: Print-Electronic Cited Medium: Internet ISSN: 1432-1084 (Electronic) Linking ISSN: 09387994 NLM ISO Abbreviation: Eur Radiol Subsets: MEDLINE
Database: MEDLINE Ultimate
FullText Text:
  Availability: 0
Header DbId: mdl
DbLabel: MEDLINE Ultimate
An: 42319406
AccessLevel: 2
PubType: Academic Journal
PubTypeId: academicJournal
PreciseRelevancyScore: 0
IllustrationInfo
Items – Name: Title
  Label: Title
  Group: Ti
  Data: GPT-4.1 and Llama 3.3 70 fail to detect clinically relevant errors in radiology reports in zero-shot evaluation.
– Name: Author
  Label: Authors
  Group: Au
  Data: <searchLink fieldCode="AU" term="%22Akinci+D'Antonoli+T%22">Akinci D'Antonoli T</searchLink>; Division of Diagnostic and Interventional Neuroradiology, Department of Radiology, University Hospital Basel, Basel, Switzerland. tugba.akincidantonoli@unibas.ch.; Department of Pediatric Radiology, University Children's Hospital Basel, Basel, Switzerland. tugba.akincidantonoli@unibas.ch.<br /><searchLink fieldCode="AU" term="%22Adams+LC%22">Adams LC</searchLink>; Institute of Diagnostic and Interventional Radiology, TUM University Hospital, TUM School of Medicine and Health, Technical University of Munich, Munich, Germany.<br /><searchLink fieldCode="AU" term="%22Lübberstedt+J%22">Lübberstedt J</searchLink>; Institute of Diagnostic and Interventional Radiology, TUM University Hospital, TUM School of Medicine and Health, Technical University of Munich, Munich, Germany.<br /><searchLink fieldCode="AU" term="%22Graf+MM%22">Graf MM</searchLink>; Institute of Diagnostic and Interventional Radiology, TUM University Hospital, TUM School of Medicine and Health, Technical University of Munich, Munich, Germany.<br /><searchLink fieldCode="AU" term="%22Mertens+CJ%22">Mertens CJ</searchLink>; Institute of Diagnostic and Interventional Radiology, TUM University Hospital, TUM School of Medicine and Health, Technical University of Munich, Munich, Germany.<br /><searchLink fieldCode="AU" term="%22Busch+F%22">Busch F</searchLink>; Institute of Diagnostic and Interventional Radiology, TUM University Hospital, TUM School of Medicine and Health, Technical University of Munich, Munich, Germany.<br /><searchLink fieldCode="AU" term="%22Ziegelmayer+S%22">Ziegelmayer S</searchLink>; Institute of Diagnostic and Interventional Radiology, TUM University Hospital, TUM School of Medicine and Health, Technical University of Munich, Munich, Germany.<br /><searchLink fieldCode="AU" term="%22Makowski+MR%22">Makowski MR</searchLink>; Institute of Diagnostic and Interventional Radiology, TUM University Hospital, TUM School of Medicine and Health, Technical University of Munich, Munich, Germany.<br /><searchLink fieldCode="AU" term="%22Bressem+K%22">Bressem K</searchLink>; Institute of Diagnostic and Interventional Radiology, TUM University Hospital, TUM School of Medicine and Health, Technical University of Munich, Munich, Germany.; Department of Cardiovascular Radiology and Nuclear Medicine, German Heart Center Munich, TUM University Hospital, Munich, Germany.<br /><searchLink fieldCode="AU" term="%22Luiken+I%22">Luiken I</searchLink>; Institute of Diagnostic and Interventional Radiology, TUM University Hospital, TUM School of Medicine and Health, Technical University of Munich, Munich, Germany.
– Name: TitleSource
  Label: Source
  Group: Src
  Data: <searchLink fieldCode="JN" term="%229114774%22">European radiology</searchLink> [Eur Radiol] 2026 Jun 19. <i>Date of Electronic Publication: </i>2026 Jun 19.
– Name: TypePub
  Label: Publication Type
  Group: TypPub
  Data: Journal Article
– Name: TitleSource
  Label: Journal Info
  Group: Src
  Data: <i>Publisher: </i><searchLink fieldCode="PB" term="%22Springer+International%22">Springer International </searchLink><i>Country of Publication: </i>Germany <i>NLM ID: </i>9114774 <i>Publication Model: </i>Print-Electronic <i>Cited Medium: </i>Internet <i>ISSN: </i>1432-1084 (Electronic) <i>Linking ISSN: </i><searchLink fieldCode="IS" term="%2209387994%22">09387994 </searchLink><i>NLM ISO Abbreviation: </i>Eur Radiol <i>Subsets: </i>MEDLINE
PLink https://search.ebscohost.com/login.aspx?direct=true&site=eds-live&db=mdl&AN=42319406
RecordInfo BibRecord:
  BibEntity:
    Identifiers:
      – Type: doi
        Value: 10.1007/s00330-026-12697-z
    Languages:
      – Code: eng
        Text: English
    Titles:
      – TitleFull: GPT-4.1 and Llama 3.3 70 fail to detect clinically relevant errors in radiology reports in zero-shot evaluation.
        Type: main
  BibRelationships:
    HasContributorRelationships:
      – PersonEntity:
          Name:
            NameFull: Akinci D'Antonoli T
      – PersonEntity:
          Name:
            NameFull: Adams LC
      – PersonEntity:
          Name:
            NameFull: Lübberstedt J
      – PersonEntity:
          Name:
            NameFull: Graf MM
      – PersonEntity:
          Name:
            NameFull: Mertens CJ
      – PersonEntity:
          Name:
            NameFull: Busch F
      – PersonEntity:
          Name:
            NameFull: Ziegelmayer S
      – PersonEntity:
          Name:
            NameFull: Makowski MR
      – PersonEntity:
          Name:
            NameFull: Bressem K
      – PersonEntity:
          Name:
            NameFull: Luiken I
    IsPartOfRelationships:
      – BibEntity:
          Dates:
            – D: 19
              M: 06
              Text: 2026 Jun 19
              Type: published
              Y: 2026
          Identifiers:
            – Type: issn-electronic
              Value: 1432-1084
          Titles:
            – TitleFull: European radiology
              Type: main
ResultId 1