@inproceedings{kostova-gievska-2026-role,
title = "Role-Diverse Agent Forests for Neuroimaging Triage: When Does Prompt-Induced Role Diversity Help?",
author = "Kostova, Tamara and
Gievska, Sonja",
editor = "Kamalloo, Ehsan and
Gontier, Nicolas and
Rozanov, Nikolai and
Hashemzadeh, Maryam and
Arora, Akhil",
booktitle = "Proceedings of the Second Workshop for {RE}search on Agent Language Models ({REALM} 2026)",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.realm-1.60/",
pages = "881--892",
ISBN = "979-8-89176-485-9",
abstract = "Majority-vote LLM ensembles improve accuracy monotonically with ensemble size when agents are homogeneous samples of the same prompt (Li et al., 2024). We test whether this holds for four heterogeneous clinical-role prompts on one frozen MedGemma backbone, across binary tumor detection, tumor subtyping, MS, and stroke triage. Our comparison is against single-agent triage: we ask whether role-diverse voting clears that bar, not whether it matches sampling diversity, since we do not run a paired homogeneous-sampling forest. On a paired 500-image evaluation per task, a four-agent forest does not reliably beat single-agent triage: significantly worse on binary tumor (both stages) and MS end-to-end, indistinguishable on tumor subtyping and stroke. Pairwise error correlation between roles is far from independent on every task ({\ensuremath{\varphi}} = 0.43{--}0.75 vs. {\ensuremath{\varphi}} = 0 under independence), explaining why voting fails to average errors out. Roles diverge behaviorally on the easiest task but collapse onto a shared ``normal'' bias on the two hardest, so voting cannot recover sensitivity no member has. Dissent rate, a candidate escalation signal, is valid on two tasks, uninformative on MS, and inverted on stroke, where agreement is less reliable than disagreement. These results argue against assuming role diversity on a shared backbone helps by default, and for task-conditional validation before deployment."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="kostova-gievska-2026-role">
<titleInfo>
<title>Role-Diverse Agent Forests for Neuroimaging Triage: When Does Prompt-Induced Role Diversity Help?</title>
</titleInfo>
<name type="personal">
<namePart type="given">Tamara</namePart>
<namePart type="family">Kostova</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sonja</namePart>
<namePart type="family">Gievska</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-10</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the Second Workshop for REsearch on Agent Language Models (REALM 2026)</title>
</titleInfo>
<name type="personal">
<namePart type="given">Ehsan</namePart>
<namePart type="family">Kamalloo</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Nicolas</namePart>
<namePart type="family">Gontier</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Nikolai</namePart>
<namePart type="family">Rozanov</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Maryam</namePart>
<namePart type="family">Hashemzadeh</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Akhil</namePart>
<namePart type="family">Arora</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Budapest, Hungary</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-485-9</identifier>
</relatedItem>
<abstract>Majority-vote LLM ensembles improve accuracy monotonically with ensemble size when agents are homogeneous samples of the same prompt (Li et al., 2024). We test whether this holds for four heterogeneous clinical-role prompts on one frozen MedGemma backbone, across binary tumor detection, tumor subtyping, MS, and stroke triage. Our comparison is against single-agent triage: we ask whether role-diverse voting clears that bar, not whether it matches sampling diversity, since we do not run a paired homogeneous-sampling forest. On a paired 500-image evaluation per task, a four-agent forest does not reliably beat single-agent triage: significantly worse on binary tumor (both stages) and MS end-to-end, indistinguishable on tumor subtyping and stroke. Pairwise error correlation between roles is far from independent on every task (\ensuremathǎrphi = 0.43–0.75 vs. \ensuremathǎrphi = 0 under independence), explaining why voting fails to average errors out. Roles diverge behaviorally on the easiest task but collapse onto a shared “normal” bias on the two hardest, so voting cannot recover sensitivity no member has. Dissent rate, a candidate escalation signal, is valid on two tasks, uninformative on MS, and inverted on stroke, where agreement is less reliable than disagreement. These results argue against assuming role diversity on a shared backbone helps by default, and for task-conditional validation before deployment.</abstract>
<identifier type="citekey">kostova-gievska-2026-role</identifier>
<location>
<url>https://aclanthology.org/2026.realm-1.60/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>881</start>
<end>892</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Role-Diverse Agent Forests for Neuroimaging Triage: When Does Prompt-Induced Role Diversity Help?
%A Kostova, Tamara
%A Gievska, Sonja
%Y Kamalloo, Ehsan
%Y Gontier, Nicolas
%Y Rozanov, Nikolai
%Y Hashemzadeh, Maryam
%Y Arora, Akhil
%S Proceedings of the Second Workshop for REsearch on Agent Language Models (REALM 2026)
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-485-9
%F kostova-gievska-2026-role
%X Majority-vote LLM ensembles improve accuracy monotonically with ensemble size when agents are homogeneous samples of the same prompt (Li et al., 2024). We test whether this holds for four heterogeneous clinical-role prompts on one frozen MedGemma backbone, across binary tumor detection, tumor subtyping, MS, and stroke triage. Our comparison is against single-agent triage: we ask whether role-diverse voting clears that bar, not whether it matches sampling diversity, since we do not run a paired homogeneous-sampling forest. On a paired 500-image evaluation per task, a four-agent forest does not reliably beat single-agent triage: significantly worse on binary tumor (both stages) and MS end-to-end, indistinguishable on tumor subtyping and stroke. Pairwise error correlation between roles is far from independent on every task (\ensuremathǎrphi = 0.43–0.75 vs. \ensuremathǎrphi = 0 under independence), explaining why voting fails to average errors out. Roles diverge behaviorally on the easiest task but collapse onto a shared “normal” bias on the two hardest, so voting cannot recover sensitivity no member has. Dissent rate, a candidate escalation signal, is valid on two tasks, uninformative on MS, and inverted on stroke, where agreement is less reliable than disagreement. These results argue against assuming role diversity on a shared backbone helps by default, and for task-conditional validation before deployment.
%U https://aclanthology.org/2026.realm-1.60/
%P 881-892
Markdown (Informal)
[Role-Diverse Agent Forests for Neuroimaging Triage: When Does Prompt-Induced Role Diversity Help?](https://aclanthology.org/2026.realm-1.60/) (Kostova & Gievska, REALM 2026)
ACL