RASC+: Retrieval-Constrained LLM Adjudication for Clinical Value Set Authoring
Quick Answer
The study introduces a retrieval-constrained LLM adjudication method that significantly enhances clinical value set authoring, achieving a macro F1 score of 0.549 using GPT-5.
Quick Take
This approach optimizes candidate selection from a recall-enhanced pool, improving recall from 0.553 to 0.730 on the RASC benchmark. The results indicate that effective retrieval strategies can lead to better performance in clinical coding tasks.
Key Points
- RASC benchmark shows struggle with clinical code generation.
- Recall improved from 0.553 to 0.730 with Qwen3-based retrieval.
- Macro F1 score increased to 0.549 using GPT-5 for candidate selection.
- Original SAPBert cross-encoder achieved only 0.287 macro F1.
- All returned codes must come from an auditable candidate pool.
Paper Resources
📖 Reader Mode
~2 min readAbstract:Clinical value sets define the standardized terminology codes used in quality measurement, phenotyping, cohort construction, and clinical decision support. The recently introduced Retrieval-Augmented Set Completion (RASC) benchmark showed that direct zero-shot large language model (LLM) generation is poorly suited to this task: clinical code systems are large, version-controlled, and not reliably memorized by language models. We study a stage-wise alternative in which candidate-pool construction is optimized for recall and a constrained LLM adjudicator is optimized for candidate selection. On the full 3,744-value-set RASC test split, Qwen3-based retrieval with vocabulary-aware expansion and code-display rescue retrieval increases candidate-pool recall from the original RASC retrieval baseline of 0.553 to 0.730; on the held-out-publisher stratum, pool recall is 0.655. The higher-recall pool alone is not sufficient: applying the original SAPBert cross-encoder to this expanded pool gives full-test macro F1 of 0.287 and held-out-publisher macro F1 of 0.233. Replacing the stage-2 selector with blinded GPT-5 adjudication over the same pool increases full-test macro F1 to 0.549 and held-out-publisher macro F1 to 0.533. These results show that retrieval-constrained LLM adjudication can substantially improve value set completion while preserving the safety constraint that all returned codes must come from an auditable candidate pool.
| Subjects: | Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG) |
| Cite as: | arXiv:2606.23992 [cs.CL] |
| (or arXiv:2606.23992v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2606.23992 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Sumit Mukherjee [view email]
[v1]
Mon, 22 Jun 2026 22:51:22 UTC (13 KB)
— Originally published at arxiv.org
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from arXiv cs.CL
See more →TriAgent: Divergence-Aware Committees for Cost-Efficient Financial Sentiment Analysis
TriAgent introduces a cost-efficient multi-agent system for financial sentiment analysis, combining VADER, FinBERT, and Qwen2.5. It achieves an F1 score of ~0.87 with significant savings of $9.3M/year at a 10M-user scale compared to GPT-4o-mini, while also detecting hallucinations with an AUC of 0.90.