LLMs trained with Span-Centric Learning improve ICD coding accuracy and efficiency

Researchers have developed a new training framework called Span-Centric Learning (SCL) to improve the accuracy of Large Language Models (LLMs) in assigning International Classification of Diseases (ICD) codes to clinical documents. This method focuses on training LLMs to recognize evidence from local text spans, which is more scalable than annotating entire documents. SCL enhances LLMs' reasoning at the span level and transfers this capability to document-level coding, leading to significant improvements in accuracy with reduced training costs. AI

Summary written by gemini-2.5-flash-lite from 1 source. How we write summaries →

IMPACT Introduces a more scalable method for training LLMs on clinical data, potentially improving diagnostic coding accuracy and auditability.

RANK_REASON This is a research paper detailing a new training framework for LLMs. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CL →

• Xu Zhang
• Llama 3.1
• Span-Centric Learning
• International Classification of Diseases
• LLMs

• paper
• model release