Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

ClinDet-Bench: Beyond Abstention, Evaluating Judgment Determinability of LLMs in Clinical Decision-Making

Yusuke Watanabe, Yohei Kobashi, Takeshi Kojima, Yusuke Iwasawa, Yasushi Okuno, Yutaka Matsuo

TL;DR

ClinDet-Bench is developed, a benchmark based on clinical scoring systems that decomposes incomplete-information scenarios into determinable and undeterminable conditions, which suggests that existing benchmarks are insufficient to evaluate the safety of large language models in clinical settings.

Abstract

Clinical decisions are often required under incomplete information. Clinical experts must identify whether available information is sufficient for judgment, as both premature conclusion and unnecessary abstention can compromise patient safety. To evaluate this capability of large language models (LLMs), we developed ClinDet-Bench, a benchmark based on clinical scoring systems that decomposes incomplete-information scenarios into determinable and undeterminable conditions. Identifying determinability requires considering all hypotheses about missing information, including unlikely ones, and verifying whether the conclusion holds across them. We find that recent LLMs fail to identify determinability under incomplete information, producing both premature judgments and excessive abstention, despite correctly explaining the underlying scoring knowledge and performing well under complete information. These findings suggest that existing benchmarks are insufficient to evaluate the safety of LLMs in clinical settings. ClinDet-Bench provides a framework for evaluating determinability recognition, leading to appropriate abstention, with potential applicability to medicine and other high-stakes domains, and is publicly available.

ClinDet-Bench: Beyond Abstention, Evaluating Judgment Determinability of LLMs in Clinical Decision-Making

TL;DR

ClinDet-Bench is developed, a benchmark based on clinical scoring systems that decomposes incomplete-information scenarios into determinable and undeterminable conditions, which suggests that existing benchmarks are insufficient to evaluate the safety of large language models in clinical settings.

Abstract

Clinical decisions are often required under incomplete information. Clinical experts must identify whether available information is sufficient for judgment, as both premature conclusion and unnecessary abstention can compromise patient safety. To evaluate this capability of large language models (LLMs), we developed ClinDet-Bench, a benchmark based on clinical scoring systems that decomposes incomplete-information scenarios into determinable and undeterminable conditions. Identifying determinability requires considering all hypotheses about missing information, including unlikely ones, and verifying whether the conclusion holds across them. We find that recent LLMs fail to identify determinability under incomplete information, producing both premature judgments and excessive abstention, despite correctly explaining the underlying scoring knowledge and performing well under complete information. These findings suggest that existing benchmarks are insufficient to evaluate the safety of LLMs in clinical settings. ClinDet-Bench provides a framework for evaluating determinability recognition, leading to appropriate abstention, with potential applicability to medicine and other high-stakes domains, and is publicly available.
Paper Structure (28 sections, 5 figures, 12 tables)

This paper contains 28 sections, 5 figures, 12 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Reasoning Limitations of LLMs
  4. Uncertainty and Abstention
  5. LLM Benchmarks in Medicine
  6. Methodology
  7. Preliminary
  8. Motivation of Benchmark
  9. ClinDet-Bench
  10. Explanation Task
  11. Clinical Decision Task
  12. Scenario Construction for Clinical Decision Task
  13. Evaluation and Statistical Analysis
  14. Explanation Task
  15. Clinical Decision Task
  16. ...and 13 more sections

Figures (5)

  • Figure 1: Overview of ClinDet-Bench. The left panel illustrates the two tasks: the Explanation Task, which tests scoring system knowledge, and the Clinical Decision Task, which tests judgment under varying information conditions. The right panel shows the three information conditions for the Clinical Decision Task, classified based on whether the possible score range crosses the decision threshold; if it does, the ground truth cannot be determined.
  • Figure 2: Accuracy in the Incomplete-Determinable versus Incomplete-Undeterminable conditions. Marker shapes represent models and colors represent prompting settings. The negative correlation indicates a trade-off between premature judgment and excessive abstention.
  • Figure A.1: Distribution of model outputs under each information condition. Rows correspond to prompting settings: Base, CoT, and Safe. The leftmost bar in each panel shows the ground-truth distribution. In the Incomplete-Undeterminable condition, the ground truth is entirely "Unable to determine," yet models frequently produce definitive judgments across all settings.
  • Figure A.2: Example of imputation of missing information.
  • Figure A.3: Example of judgment based on incompleteness.