@inproceedings{ma-zhang-2026-retrieval,
title = "Retrieval-Induced Mirage: When Similar Cases Become Fabricated Evidence in Medical {VLM}s",
author = "Ma, Xiaoxu and
Zhang, Xiangbo",
editor = "Zhang, Yue and
Xu, Ruifeng and
Purwarianti, Ayu and
Rong, Tong and
Suzuki, Jun",
booktitle = "Findings of the {A}ssociation for {C}omputational {L}inguistics: {IJCNLP}-{AACL} 2026",
month = nov,
year = "2026",
address = "Hengqin, China",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.findings-aacl.108/",
pages = "2127--2144",
ISBN = "979-8-89176-444-6",
abstract = "Retrieval-augmented medical vision-language models are increasingly used to support clinical reasoning when patient information is incomplete. However, retrieved cases are observations from other patients, not direct evidence for the target patient. In this paper, we study a source-attribution failure that we call retrieval-induced mirage: under missing target modalities, a model may copy findings from retrieved complete cases and present them as if they were observed in the target patient. We introduce RIM-BENCH, a counterfactual evaluation benchmark that fixes a controlled missing-modality setting while varying retrieval consistency and source-awareness. Unlike standard diagnostic accuracy evaluations, RIM-BENCH measures whether generated clinical claims are attributed to the correct evidence source. We define metrics for unavailable-modality claims, retrieved-to-target misattribution, evidence-source accuracy, retrieval-following under conflicting evidence, and missing-modality awareness. Across 1,200 target instances and three main models, implicit RAG raises unavailable-modality claims from 20.0{\%} to 36.3{\%} and induces a 26.2{\%} RIMR, while source-attribution prompting lowers RIMR to 11.4{\%} but does not eliminate the failure. Our framework highlights a safety risk for retrieval-augmented multimodal diagnosis: retrieval can be helpful as statistical context, but unsafe when models verbalize retrieved evidence as patient-specific observations."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="ma-zhang-2026-retrieval">
<titleInfo>
<title>Retrieval-Induced Mirage: When Similar Cases Become Fabricated Evidence in Medical VLMs</title>
</titleInfo>
<name type="personal">
<namePart type="given">Xiaoxu</namePart>
<namePart type="family">Ma</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Xiangbo</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-11</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Findings of the Association for Computational Linguistics: IJCNLP-AACL 2026</title>
</titleInfo>
<name type="personal">
<namePart type="given">Yue</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ruifeng</namePart>
<namePart type="family">Xu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ayu</namePart>
<namePart type="family">Purwarianti</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Tong</namePart>
<namePart type="family">Rong</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jun</namePart>
<namePart type="family">Suzuki</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Hengqin, China</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-444-6</identifier>
</relatedItem>
<abstract>Retrieval-augmented medical vision-language models are increasingly used to support clinical reasoning when patient information is incomplete. However, retrieved cases are observations from other patients, not direct evidence for the target patient. In this paper, we study a source-attribution failure that we call retrieval-induced mirage: under missing target modalities, a model may copy findings from retrieved complete cases and present them as if they were observed in the target patient. We introduce RIM-BENCH, a counterfactual evaluation benchmark that fixes a controlled missing-modality setting while varying retrieval consistency and source-awareness. Unlike standard diagnostic accuracy evaluations, RIM-BENCH measures whether generated clinical claims are attributed to the correct evidence source. We define metrics for unavailable-modality claims, retrieved-to-target misattribution, evidence-source accuracy, retrieval-following under conflicting evidence, and missing-modality awareness. Across 1,200 target instances and three main models, implicit RAG raises unavailable-modality claims from 20.0% to 36.3% and induces a 26.2% RIMR, while source-attribution prompting lowers RIMR to 11.4% but does not eliminate the failure. Our framework highlights a safety risk for retrieval-augmented multimodal diagnosis: retrieval can be helpful as statistical context, but unsafe when models verbalize retrieved evidence as patient-specific observations.</abstract>
<identifier type="citekey">ma-zhang-2026-retrieval</identifier>
<location>
<url>https://aclanthology.org/2026.findings-aacl.108/</url>
</location>
<part>
<date>2026-11</date>
<extent unit="page">
<start>2127</start>
<end>2144</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Retrieval-Induced Mirage: When Similar Cases Become Fabricated Evidence in Medical VLMs
%A Ma, Xiaoxu
%A Zhang, Xiangbo
%Y Zhang, Yue
%Y Xu, Ruifeng
%Y Purwarianti, Ayu
%Y Rong, Tong
%Y Suzuki, Jun
%S Findings of the Association for Computational Linguistics: IJCNLP-AACL 2026
%D 2026
%8 November
%I Association for Computational Linguistics
%C Hengqin, China
%@ 979-8-89176-444-6
%F ma-zhang-2026-retrieval
%X Retrieval-augmented medical vision-language models are increasingly used to support clinical reasoning when patient information is incomplete. However, retrieved cases are observations from other patients, not direct evidence for the target patient. In this paper, we study a source-attribution failure that we call retrieval-induced mirage: under missing target modalities, a model may copy findings from retrieved complete cases and present them as if they were observed in the target patient. We introduce RIM-BENCH, a counterfactual evaluation benchmark that fixes a controlled missing-modality setting while varying retrieval consistency and source-awareness. Unlike standard diagnostic accuracy evaluations, RIM-BENCH measures whether generated clinical claims are attributed to the correct evidence source. We define metrics for unavailable-modality claims, retrieved-to-target misattribution, evidence-source accuracy, retrieval-following under conflicting evidence, and missing-modality awareness. Across 1,200 target instances and three main models, implicit RAG raises unavailable-modality claims from 20.0% to 36.3% and induces a 26.2% RIMR, while source-attribution prompting lowers RIMR to 11.4% but does not eliminate the failure. Our framework highlights a safety risk for retrieval-augmented multimodal diagnosis: retrieval can be helpful as statistical context, but unsafe when models verbalize retrieved evidence as patient-specific observations.
%U https://aclanthology.org/2026.findings-aacl.108/
%P 2127-2144
Markdown (Informal)
[Retrieval-Induced Mirage: When Similar Cases Become Fabricated Evidence in Medical VLMs](https://aclanthology.org/2026.findings-aacl.108/) (Ma & Zhang, Findings 2026)
ACL