# Coverage-Controlled Preference Mining from Noisy Claim Verification for Evidence-Grounded Generation

Source: [Coverage-Controlled Preference Mining from Noisy Claim Verification for Evidence-Grounded Generation](<https://www.sota2.com/research/paper/69b248a9d030084c22e2f9f6>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-07-08T05:00:13.057Z

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

Published: 2026-03-11T07:41:58.000Z

## Abstract

Evidence-grounded generation produces summaries whose claims should be supported by supplied evidence, but claim-level verifiers provide noisy feedback and can reward models that simply say less. We study this problem in clinical Brief Hospital Course summarization, where outputs must remain grounded in patient-specific EHR evidence. We introduce VERI-DPO, a preference-mining framework that converts noisy claim verification into coverage-controlled summary-level preferences. For each evidence-window prompt, VERI-DPO samples multiple candidate summaries, decomposes them into claims, verifies each claim against patient evidence, and forms a preference pair only when the chosen summary has better aggregate verifier-estimated support while retaining comparable verifiable content. Standard Direct Preference Optimization then distills these pairs into a single-sample policy, avoiding inference-time reranking. On patient-disjoint MIMIC-III-Ext-VeriFact-BHC test data, VERI-DPO reduces Not Supported rates from 10.7% to 1.9% under the mining verifier and from 11.6% to 6.4% under a separately prompted GPT-4o judge. In 100 blinded pairwise assessments by two domain researchers, VERI-DPO is preferred over the base model 56 times versus 18 times for factual faithfulness. In a locked zero-shot MIMIC-IV-Ext-BHC transfer test with 1,000 patients and no model adaptation, VERI-DPO lowers Not Supported rates with nearly unchanged scored-claim counts. Multi-seed ablations show that verifier-guided pair construction drives the gains, while coverage and anti-degeneration controls prevent apparent factuality improvements from coming from shorter or less checkable outputs.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [Hospital Course Summarization on MIMIC-III-ext-VeriFact-BHC 120 prompts (test)](<https://www.sota2.com/research/sota/hospital-course-summarization-on-mimic-iii-ext-verifact-bhc-120-prompts-test.md>): Hospital Course Summarization; MIMIC-III-ext-VeriFact-BHC 120 prompts (test); Local Verifier NS-rate: 1.9.

- [Clinical Claim Verification on Clinical Claim Verification Dataset (test)](<https://www.sota2.com/research/sota/clinical-claim-verification-on-clinical-claim-verification-dataset-test.md>): Clinical claim verification; Clinical Claim Verification Dataset (test); Macro-F1: not recorded.

- [Clinical Claim Verification on Clinical Claim Verification Dataset (dev)](<https://www.sota2.com/research/sota/clinical-claim-verification-on-clinical-claim-verification-dataset-dev.md>): Clinical claim verification; Clinical Claim Verification Dataset (dev); Macro-F1: not recorded.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2603.10494>)

[PDF](<https://arxiv.org/pdf/2603.10494.pdf>)
