Table 1 Overview of LLM techniques for diagnostic tasks

From: Large language models for disease diagnosis: a scoping review

Techniques

Types

Representative studies

Prompting

Zero-shot

Text196,197, image65,198, audio70,72, text-image52, text-time series73,199, text-tabular200

 

Few-shot

Text25,187, image58, text-image41,201, text-image-tabular153

 

CoT

Text51,202, audio203, time series155, text-image44,204

 

Self-consistency

Text89, audio205, text-image-tabular-time series45

 

Soft prompt

Text206, image207, tabular-time series47,208, text-image-graph59

RAG

Knowledge graph

Text81, text-time series94

 

Corpus

Text85,87, text-image64,86, text-time series83

 

Database

Text80,93, text-image90

Fine-tuning

SFT

Text98,209,210, text-image133,211,212, text-video102,112, text-audio111,213, text-tabular42,200

 

RLHF

Text116,117,214, text-image115

 

PEFT

Text98,124,215, text-image104

Pre-training

-

Text124,129,131, text-image109,133,137, text-tabular135,200, text-video213, text-omics109

  1. SFT supervised fine-tuning, RLHF reinforcement learning from human feedback, PEFT parameter-efficient fine-tuning.
Back to article page