A graph-based evaluation harness transforms clinical guidelines into a queryable knowledge graph to dynamically generate contamination-resistant multiple-choice questions, revealing LLM performance gaps on symptom recognition versus treatment decisions.
Radqa: A question answering dataset to improve comprehension of radiology reports
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.AI 1years
2025 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
From Guidelines to Guarantees: A Graph-Based Evaluation Harness for Domain-Specific Evaluation of LLMs
A graph-based evaluation harness transforms clinical guidelines into a queryable knowledge graph to dynamically generate contamination-resistant multiple-choice questions, revealing LLM performance gaps on symptom recognition versus treatment decisions.