Enhancing Trustworthy Clinical Diagnosis Decision-Making in Large Language Models via Etiology-Aware Attention Supervision

2025/08/01 by Peixian Li, Li, Peixian, Yu Tian +10
Computer Science · Medicine · #Artificial Intelligence in Healthcare and Education #Clinical Reasoning and Diagnostic Skills #Computation and Language (cs.CL) #FOS: Computer and information sciences #I.2.7 #J.3 #Machine Learning in Healthcare #cs.CL

paper · pdf · doi:10.48550/arxiv.2508.00285

20 pages, 8 figures

openalex publication_date 2025/08/01 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28 · arxiv created 2026/08/05 · arxiv updated 2026/08/06

Abstract

Objective: Large Language Models (LLMs) have demonstrated strong capabilities in medical text understanding and generation. However, their trustworthiness in diagnosis-oriented medical tasks remains constrained by the lack of structured guidance on how clinically relevant diagnostic evidence is internally attended to and utilized during model learning. Method: We propose an Etiology-Aware Attention Supervision framework that introduces structured etiological information as an external supervisory signal for training large language models. Specifically, we construct Clinical Etiology Schema (CES) derived from authoritative clinical guidelines for three acute abdominal conditions: acute appendicitis, acute pancreatitis, and acute cholecystitis. Based on CES annotations, we develop an Etiology-Aware Head Identification strategy to identify attention heads that consistently align with etiological evidence. Building on this analysis, we design a structure-guided parameter-efficient fine-tuning approach that steers attention distributions toward clinically relevant evidence through an additional supervision loss, without modifying the base model architecture. Result: Experiments conducted on a Consistent Diagnosis Cohort demonstrate that the proposed framework improves average diagnostic accuracy by 15.65% compared with baseline models. Attention-based metrics, including Inference Focus Score and Inference Attention Frequency, show more concentrated attention on etiologically relevant evidence. External evaluation on a Discrepant Diagnosis Cohort further confirms the robustness of diagnostic performance improvements under real-world clinical inconsistencies.

Citations

• Spotlight Your Instructions: Instruction-following with Dynamic Attention Steering
• MedReason: Eliciting Factual Medical Reasoning Steps in LLMs via Knowledge Graphs
• CODI: Compressing Chain-of-Thought into Continuous Space via Self-Distillation
• Improving Contextual Faithfulness of Large Language Models via Retrieval Heads-Induced Optimization
• DeepSeek-R1 incentivizes reasoning in LLMs through reinforcement learning
• Qwen2.5 Technical Report
• Reasoning Beyond Words ? Exploring framework for hidden state reasoning
• DeCoRe: Decoding by Contrasting Retrieval Heads to Mitigate Hallucinations
• DuoAttention: Efficient Long-Context LLM Inference with Retrieval and Streaming Heads
• Attention Heads of Large Language Models: A Survey
• Retrieval Head Mechanistically Explains Long-Context Factuality
• Guiding Clinical Reasoning with Large Language Models via Knowledge Seeds
• Large Language Models are Clinical Reasoners: Reasoning-Aware Diagnosis Framework with Prompt-Generated Rationales
• Diagnostic Reasoning Prompts Reveal the Potential for Large Language Model Interpretability in Medicine
• QLoRA: Efficient Finetuning of Quantized LLMs
• AdaLoRA: Adaptive Budget Allocation for Parameter-Efficient Fine-Tuning
• Large Language Models Encode Clinical Knowledge
• Large language models encode clinical knowledge
• LoRA: Low-Rank Adaptation of Large Language Models
• Attention Is All You Need

Related

• Large Language Models are Clinical Reasoners: Reasoning-Aware Diagnosis Framework with Prompt-Generated Rationales
• Diagnostic Reasoning Prompts Reveal the Potential for Large Language Model Interpretability in Medicine
• DiReCT: Diagnostic Reasoning for Clinical Notes via Large Language Models
• MedCaseReasoning: Evaluating and learning diagnostic reasoning from clinical case reports
• DiagnosisArena: Benchmarking Diagnostic Reasoning for Large Language Models