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

2026-02-26

AI Summary

This work addresses a critical safety issue in clinical decision-making with large language models (LLMs): their tendency to render premature judgments or excessive abstentions due to an inability to assess whether available information suffices to support a conclusion. To tackle this, the authors introduce ClinDet-Bench, the first benchmark specifically designed to evaluate LLMs’ capacity for judging determinability—shifting beyond conventional evaluation paradigms that focus solely on answer correctness. Grounded in clinical scoring systems, the benchmark categorizes incomplete-information scenarios into determinable and indeterminable cases and evaluates model robustness by testing conclusions against multiple hypotheses about missing data, including low-probability ones. Experiments reveal that while leading models perform well under complete information, they consistently fail to accurately discern when a reliable judgment is feasible, exposing a fundamental limitation in current evaluation frameworks.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: Safety and Robustness
• Reasoning under Uncertainty: Other Foundations of Reasoning under Uncertainty

Application Category

• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems

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.

Problem

• clinical decision-making
• incomplete information
• judgment determinability
• large language models
• abstention

Innovation

• determinability
• incomplete information
• clinical decision-making
• abstention
• LLM evaluation