@inproceedings{wang-etal-2026-clintracebench,
title = "{C}lin{T}race{B}ench: Source-Verifiable Longitudinal Clinical Reasoning over {EHR}-Derived Dialogues",
author = "Wang, Huimin and
Zhao, Zhengyi and
Zhao, Yutian",
editor = "Lee, Hung-Yi and
Dev, Sunipa and
Li, Jessy and
Elliott, Desmond",
booktitle = "Findings of the {A}ssociation for {C}omputational {L}inguistics: {EMNLP} 2026",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.findings-emnlp.1480/",
pages = "31524--31549",
ISBN = "979-8-89176-462-0",
abstract = "Clinical LLM assistants must reason over multi-visit patient trajectories, yet whether the compact history representations used to scale them{---}retrieval, structured timelines, LLM summaries, agentic memory{---}preserve the longitudinal signal clinical reasoning needs has not been measured. We introduce ClinTraceBench: 385 MIMIC-IV-derived verified dialogues with event-ID provenance, a nine-task taxonomy (T1{--}T9), and L0{--}L4 deterministic + L5 human-audit validation (98.92{\%} agreement). We evaluate eight history representation strategies{---}a no-context floor, last-visit-only, full-context, BGE-M3 dense-retrieval, two compression schemes, and two agentic-memory systems (Mem0, A-Mem){---}across four backbones (DeepSeek-V3, GPT-4o-mini, Haiku 4.5, Sonnet 4.6) on 6,271 questions: 32 cells, 200,672 predictions. Four findings emerge: (F1) compressed strategies pay an aggregation tax on multi-visit trends and cross-patient comparisons; (F2) the blind-to-full gap spans +29.8 percentage points (GPT-4o-mini) to +62.7 (Haiku); (F3) abstention scales non-monotonically with context length; (F4) a controlled T3 injection probe isolates compression-induced relation loss{---}with the attribution sentence present before construction, Mem0, A-Mem, and llm-summary still recover only 0{--}5.3{\%} of the injected positives. On the Pareto frontier Haiku dominates Sonnet under full-context ($25.76 vs. $106.21), inverting the ``biggest backbone wins'' heuristic."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="wang-etal-2026-clintracebench">
<titleInfo>
<title>ClinTraceBench: Source-Verifiable Longitudinal Clinical Reasoning over EHR-Derived Dialogues</title>
</titleInfo>
<name type="personal">
<namePart type="given">Huimin</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Zhengyi</namePart>
<namePart type="family">Zhao</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yutian</namePart>
<namePart type="family">Zhao</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>Findings of the Association for Computational Linguistics: EMNLP 2026</title>
</titleInfo>
<name type="personal">
<namePart type="given">Hung-Yi</namePart>
<namePart type="family">Lee</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sunipa</namePart>
<namePart type="family">Dev</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jessy</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Desmond</namePart>
<namePart type="family">Elliott</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">979-8-89176-462-0</identifier>
</relatedItem>
<abstract>Clinical LLM assistants must reason over multi-visit patient trajectories, yet whether the compact history representations used to scale them—retrieval, structured timelines, LLM summaries, agentic memory—preserve the longitudinal signal clinical reasoning needs has not been measured. We introduce ClinTraceBench: 385 MIMIC-IV-derived verified dialogues with event-ID provenance, a nine-task taxonomy (T1–T9), and L0–L4 deterministic + L5 human-audit validation (98.92% agreement). We evaluate eight history representation strategies—a no-context floor, last-visit-only, full-context, BGE-M3 dense-retrieval, two compression schemes, and two agentic-memory systems (Mem0, A-Mem)—across four backbones (DeepSeek-V3, GPT-4o-mini, Haiku 4.5, Sonnet 4.6) on 6,271 questions: 32 cells, 200,672 predictions. Four findings emerge: (F1) compressed strategies pay an aggregation tax on multi-visit trends and cross-patient comparisons; (F2) the blind-to-full gap spans +29.8 percentage points (GPT-4o-mini) to +62.7 (Haiku); (F3) abstention scales non-monotonically with context length; (F4) a controlled T3 injection probe isolates compression-induced relation loss—with the attribution sentence present before construction, Mem0, A-Mem, and llm-summary still recover only 0–5.3% of the injected positives. On the Pareto frontier Haiku dominates Sonnet under full-context (25.76 vs. 106.21), inverting the “biggest backbone wins” heuristic.</abstract>
<identifier type="citekey">wang-etal-2026-clintracebench</identifier>
<location>
<url>https://aclanthology.org/2026.findings-emnlp.1480/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>31524</start>
<end>31549</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T ClinTraceBench: Source-Verifiable Longitudinal Clinical Reasoning over EHR-Derived Dialogues
%A Wang, Huimin
%A Zhao, Zhengyi
%A Zhao, Yutian
%Y Lee, Hung-Yi
%Y Dev, Sunipa
%Y Li, Jessy
%Y Elliott, Desmond
%S Findings of the Association for Computational Linguistics: EMNLP 2026
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-462-0
%F wang-etal-2026-clintracebench
%X Clinical LLM assistants must reason over multi-visit patient trajectories, yet whether the compact history representations used to scale them—retrieval, structured timelines, LLM summaries, agentic memory—preserve the longitudinal signal clinical reasoning needs has not been measured. We introduce ClinTraceBench: 385 MIMIC-IV-derived verified dialogues with event-ID provenance, a nine-task taxonomy (T1–T9), and L0–L4 deterministic + L5 human-audit validation (98.92% agreement). We evaluate eight history representation strategies—a no-context floor, last-visit-only, full-context, BGE-M3 dense-retrieval, two compression schemes, and two agentic-memory systems (Mem0, A-Mem)—across four backbones (DeepSeek-V3, GPT-4o-mini, Haiku 4.5, Sonnet 4.6) on 6,271 questions: 32 cells, 200,672 predictions. Four findings emerge: (F1) compressed strategies pay an aggregation tax on multi-visit trends and cross-patient comparisons; (F2) the blind-to-full gap spans +29.8 percentage points (GPT-4o-mini) to +62.7 (Haiku); (F3) abstention scales non-monotonically with context length; (F4) a controlled T3 injection probe isolates compression-induced relation loss—with the attribution sentence present before construction, Mem0, A-Mem, and llm-summary still recover only 0–5.3% of the injected positives. On the Pareto frontier Haiku dominates Sonnet under full-context (25.76 vs. 106.21), inverting the “biggest backbone wins” heuristic.
%U https://aclanthology.org/2026.findings-emnlp.1480/
%P 31524-31549
Markdown (Informal)
[ClinTraceBench: Source-Verifiable Longitudinal Clinical Reasoning over EHR-Derived Dialogues](https://aclanthology.org/2026.findings-emnlp.1480/) (Wang et al., Findings 2026)
ACL