@inproceedings{zabolotnii-2026-docua,
title = "{D}oc{UA} at {CRF} Filling 2026: {LLM} {S}truct{C}ore {---} Schema-Guided Reasoning Condensation and Deterministic Compilation",
author = "Zabolotnii, Serhii",
editor = "Gupta, Deepak and
Thompson, Paul and
Ananiadou, Sophia and
Demner-Fushman, Dina",
booktitle = "Proceedings of the Third Workshop on Patient-Oriented Language Processing ({CL}4{H}ealth) @ {LREC} 2026",
month = may,
year = "2026",
address = "Palma, Mallorca (Spain)",
publisher = "ELRA Language Resources Association (ELRA)",
url = "https://aclanthology.org/2026.cl4health-1.33/",
doi = "10.63317/4ar36j4w3vrk",
pages = "375--383",
abstract = "Automatically filling Case Report Forms (CRFs) from clinical notes is challenging due to noisy language, strict output contracts, and the high cost of false positives. We describe our CL4Health 2026 submission for Dyspnea CRF filling (134 items) using a contract-driven two-stage design grounded in Schema-Guided Reasoning (SGR) (Abdullin, 2025). The key task property is extreme sparsity: the majority of fields are unknown, and official scoring penalizes both empty values and unsupported predictions. We shift from a single-step ``LLM predicts 134 fields'' approach to a decomposition where (i) Stage 1 produces a stable SGR-style JSON summary with exactly 9 domain keys, and (ii) Stage 2 is a fully deterministic, 0-LLM compiler that parses the Stage 1 summary, canonicalizes item names (optionally using a UMLS alias map with 134/134 coverage), normalizes predictions to the official controlled vocabulary (13 categories), applies evidence-gated false-positive filters, and expands the output into the required 134-item format. On the dev80 split, the best teacher configuration (Mistral Large 3 Stage 1 {\textrightarrow} Stage 2 deterministic) achieves macro-F1 0.6543 (EN) and 0.6905 (IT); on the hidden test200, the submitted English variant scores 0.63 on Codabench. The pipeline is language-agnostic: Italian results match or exceed English with no language-specific engineering."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="zabolotnii-2026-docua">
<titleInfo>
<title>DocUA at CRF Filling 2026: LLM StructCore — Schema-Guided Reasoning Condensation and Deterministic Compilation</title>
</titleInfo>
<name type="personal">
<namePart type="given">Serhii</namePart>
<namePart type="family">Zabolotnii</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-05</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the Third Workshop on Patient-Oriented Language Processing (CL4Health) @ LREC 2026</title>
</titleInfo>
<name type="personal">
<namePart type="given">Deepak</namePart>
<namePart type="family">Gupta</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Paul</namePart>
<namePart type="family">Thompson</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sophia</namePart>
<namePart type="family">Ananiadou</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Dina</namePart>
<namePart type="family">Demner-Fushman</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>ELRA Language Resources Association (ELRA)</publisher>
<place>
<placeTerm type="text">Palma, Mallorca (Spain)</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
</relatedItem>
<abstract>Automatically filling Case Report Forms (CRFs) from clinical notes is challenging due to noisy language, strict output contracts, and the high cost of false positives. We describe our CL4Health 2026 submission for Dyspnea CRF filling (134 items) using a contract-driven two-stage design grounded in Schema-Guided Reasoning (SGR) (Abdullin, 2025). The key task property is extreme sparsity: the majority of fields are unknown, and official scoring penalizes both empty values and unsupported predictions. We shift from a single-step “LLM predicts 134 fields” approach to a decomposition where (i) Stage 1 produces a stable SGR-style JSON summary with exactly 9 domain keys, and (ii) Stage 2 is a fully deterministic, 0-LLM compiler that parses the Stage 1 summary, canonicalizes item names (optionally using a UMLS alias map with 134/134 coverage), normalizes predictions to the official controlled vocabulary (13 categories), applies evidence-gated false-positive filters, and expands the output into the required 134-item format. On the dev80 split, the best teacher configuration (Mistral Large 3 Stage 1 → Stage 2 deterministic) achieves macro-F1 0.6543 (EN) and 0.6905 (IT); on the hidden test200, the submitted English variant scores 0.63 on Codabench. The pipeline is language-agnostic: Italian results match or exceed English with no language-specific engineering.</abstract>
<identifier type="citekey">zabolotnii-2026-docua</identifier>
<identifier type="doi">10.63317/4ar36j4w3vrk</identifier>
<location>
<url>https://aclanthology.org/2026.cl4health-1.33/</url>
</location>
<part>
<date>2026-05</date>
<extent unit="page">
<start>375</start>
<end>383</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T DocUA at CRF Filling 2026: LLM StructCore — Schema-Guided Reasoning Condensation and Deterministic Compilation
%A Zabolotnii, Serhii
%Y Gupta, Deepak
%Y Thompson, Paul
%Y Ananiadou, Sophia
%Y Demner-Fushman, Dina
%S Proceedings of the Third Workshop on Patient-Oriented Language Processing (CL4Health) @ LREC 2026
%D 2026
%8 May
%I ELRA Language Resources Association (ELRA)
%C Palma, Mallorca (Spain)
%F zabolotnii-2026-docua
%X Automatically filling Case Report Forms (CRFs) from clinical notes is challenging due to noisy language, strict output contracts, and the high cost of false positives. We describe our CL4Health 2026 submission for Dyspnea CRF filling (134 items) using a contract-driven two-stage design grounded in Schema-Guided Reasoning (SGR) (Abdullin, 2025). The key task property is extreme sparsity: the majority of fields are unknown, and official scoring penalizes both empty values and unsupported predictions. We shift from a single-step “LLM predicts 134 fields” approach to a decomposition where (i) Stage 1 produces a stable SGR-style JSON summary with exactly 9 domain keys, and (ii) Stage 2 is a fully deterministic, 0-LLM compiler that parses the Stage 1 summary, canonicalizes item names (optionally using a UMLS alias map with 134/134 coverage), normalizes predictions to the official controlled vocabulary (13 categories), applies evidence-gated false-positive filters, and expands the output into the required 134-item format. On the dev80 split, the best teacher configuration (Mistral Large 3 Stage 1 → Stage 2 deterministic) achieves macro-F1 0.6543 (EN) and 0.6905 (IT); on the hidden test200, the submitted English variant scores 0.63 on Codabench. The pipeline is language-agnostic: Italian results match or exceed English with no language-specific engineering.
%R 10.63317/4ar36j4w3vrk
%U https://aclanthology.org/2026.cl4health-1.33/
%U https://doi.org/10.63317/4ar36j4w3vrk
%P 375-383
Markdown (Informal)
[DocUA at CRF Filling 2026: LLM StructCore — Schema-Guided Reasoning Condensation and Deterministic Compilation](https://aclanthology.org/2026.cl4health-1.33/) (Zabolotnii, CL4Health 2026)
ACL