Research

Paper

AI LLM March 11, 2026

VERI-DPO: Evidence-Aware Alignment for Clinical Summarization via Claim Verification and Direct Preference Optimization

Authors

Weixin Liu, Congning Ni, Qingyuan Song, Susannah L. Rose, Christopher Symons, Murat Kantarcioglu, Bradley A. Malin, Zhijun Yin

Abstract

Brief Hospital Course (BHC) narratives must be clinically useful yet faithful to fragmented EHR evidence. LLM-based clinical summarizers still introduce unsupported statements, and alignment can encourage omissions ("say-less" degeneration). We introduce VERI-DPO, which uses claim verification to mine preferences and distill them into the summarizer with Direct Preference Optimization (DPO). On MIMIC-III-Ext-VeriFact-BHC (100 ICU patients; patient-level splits), we train a retrieval-augmented verifier to label claim-evidence pairs as Supported, Not Supported, or Not Addressed via a single-token format. The verifier scores sentence-level claims from sampled BHC candidates and aggregates margins into a coverage-aware utility to mine length-controlled, contradiction-anchored preference pairs. On held-out patients, verifier-mined preferences separate candidates by contradiction density, and VERI-DPO reduces Not Supported claim rates from 10.7% to 1.9% (local verifier judge) and from 11.6% to 6.4% (GPT-4o judge), while improving validity from 76.7% to 82.5% and maintaining informative length.

Metadata

arXiv ID: 2603.10494
Provider: ARXIV
Primary Category: cs.CL
Published: 2026-03-11
Fetched: 2026-03-12 04:21

Related papers

Raw Data (Debug)
{
  "raw_xml": "<entry>\n    <id>http://arxiv.org/abs/2603.10494v1</id>\n    <title>VERI-DPO: Evidence-Aware Alignment for Clinical Summarization via Claim Verification and Direct Preference Optimization</title>\n    <updated>2026-03-11T07:41:58Z</updated>\n    <link href='https://arxiv.org/abs/2603.10494v1' rel='alternate' type='text/html'/>\n    <link href='https://arxiv.org/pdf/2603.10494v1' rel='related' title='pdf' type='application/pdf'/>\n    <summary>Brief Hospital Course (BHC) narratives must be clinically useful yet faithful to fragmented EHR evidence. LLM-based clinical summarizers still introduce unsupported statements, and alignment can encourage omissions (\"say-less\" degeneration). We introduce VERI-DPO, which uses claim verification to mine preferences and distill them into the summarizer with Direct Preference Optimization (DPO). On MIMIC-III-Ext-VeriFact-BHC (100 ICU patients; patient-level splits), we train a retrieval-augmented verifier to label claim-evidence pairs as Supported, Not Supported, or Not Addressed via a single-token format. The verifier scores sentence-level claims from sampled BHC candidates and aggregates margins into a coverage-aware utility to mine length-controlled, contradiction-anchored preference pairs. On held-out patients, verifier-mined preferences separate candidates by contradiction density, and VERI-DPO reduces Not Supported claim rates from 10.7% to 1.9% (local verifier judge) and from 11.6% to 6.4% (GPT-4o judge), while improving validity from 76.7% to 82.5% and maintaining informative length.</summary>\n    <category scheme='http://arxiv.org/schemas/atom' term='cs.CL'/>\n    <category scheme='http://arxiv.org/schemas/atom' term='cs.LG'/>\n    <published>2026-03-11T07:41:58Z</published>\n    <arxiv:comment>Paper submitted to AMIA 2026 Annual Symposium</arxiv:comment>\n    <arxiv:primary_category term='cs.CL'/>\n    <author>\n      <name>Weixin Liu</name>\n    </author>\n    <author>\n      <name>Congning Ni</name>\n    </author>\n    <author>\n      <name>Qingyuan Song</name>\n    </author>\n    <author>\n      <name>Susannah L. Rose</name>\n    </author>\n    <author>\n      <name>Christopher Symons</name>\n    </author>\n    <author>\n      <name>Murat Kantarcioglu</name>\n    </author>\n    <author>\n      <name>Bradley A. Malin</name>\n    </author>\n    <author>\n      <name>Zhijun Yin</name>\n    </author>\n  </entry>"
}