Evaluating Prompt Engineering Techniques for Accuracy and Confidence Elicitation in Medical LLMs

2025/05/29 by Nariman Naderi, Zahra Atf, Naderi, Nariman +9
Health Professions · #Artificial Intelligence (cs.AI) #Computation and Language (cs.CL) #Computers and Society (cs.CY) #FOS: Computer and information sciences #Machine Learning (cs.LG) #Quality and Safety in Healthcare

paper · pdf · doi:10.48550/arxiv.2506.00072

openalex publication_date 2025/05/29 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28

Abstract

This paper investigates how prompt engineering techniques impact both accuracy and confidence elicitation in Large Language Models (LLMs) applied to medical contexts. Using a stratified dataset of Persian board exam questions across multiple specialties, we evaluated five LLMs - GPT-4o, o3-mini, Llama-3.3-70b, Llama-3.1-8b, and DeepSeek-v3 - across 156 configurations. These configurations varied in temperature settings (0.3, 0.7, 1.0), prompt styles (Chain-of-Thought, Few-Shot, Emotional, Expert Mimicry), and confidence scales (1-10, 1-100). We used AUC-ROC, Brier Score, and Expected Calibration Error (ECE) to evaluate alignment between confidence and actual performance. Chain-of-Thought prompts improved accuracy but also led to overconfidence, highlighting the need for calibration. Emotional prompting further inflated confidence, risking poor decisions. Smaller models like Llama-3.1-8b underperformed across all metrics, while proprietary models showed higher accuracy but still lacked calibrated confidence. These results suggest prompt engineering must address both accuracy and uncertainty to be effective in high-stakes medical tasks.

Citations

• The challenge of uncertainty quantification of large language models in medicine
• Accuracy and Consistency of LLMs in the Registered Dietitian Exam: The Impact of Prompt Engineering and Knowledge Retrieval
• What Did I Do Wrong? Quantifying LLMs' Sensitivity and Consistency to Prompt Engineering
• Cycles of Thought: Measuring LLM Confidence through Stable Explanations
• To Believe or Not to Believe Your LLM
• SaySelf: Teaching LLMs to Express Confidence with Self-Reflective Rationales
• Just rephrase it! Uncertainty estimation in closed-source language models via multiple rephrased queries
• Examining the robustness of LLM evaluation to the distributional assumptions of benchmarks
• Uncertainty Quantification for In-Context Learning of Large Language Models
• Can LLMs Express Their Uncertainty? An Empirical Evaluation of Confidence Elicitation in LLMs
• Do Language Models Know When They're Hallucinating References?
• Language Models (Mostly) Know What They Know
• What Disease does this Patient Have? A Large-scale Open Domain Question Answering Dataset from Medical Exams

Related

• Methods to Estimate Large Language Model Confidence
• Do Large Language Models have Shared Weaknesses in Medical Question Answering?
• Beyond Overall Accuracy: A Psychometric Deep Dive into the Topic-Specific Medical Capabilities of 80 Large Language Models
• Susceptibility of Large Language Models to User-Driven Factors in Medical Queries
• ConfidenceBench: Evaluating Confidence Calibration in Large Language Models