Multi-Task LLM with LoRA Fine-Tuning for Automated Cancer Staging and Biomarker Extraction

arXiv cs.LG / 4/16/2026

💬 OpinionSignals & Early TrendsIdeas & Deep AnalysisModels & Research

共有:

Key Points

The paper proposes a parameter-efficient multi-task LLM framework that automates breast cancer TNM staging, histologic grade, and biomarker extraction from unstructured pathology reports.
It fine-tunes a Llama-3-8B-Instruct encoder with LoRA using an expert-verified dataset of 10,677 reports, aiming to reduce compute cost and mitigate hallucination issues common in generative extraction.
Instead of purely generative prompting, the model uses parallel classification heads to enforce schema adherence for more consistent structured outputs.
The approach reports strong performance (Macro F1 = 0.976), particularly for resolving contextual ambiguities and heterogeneous report formats that often break rule-based NLP, zero-shot LLMs, and single-task baselines.
The authors argue the adapter-efficient multi-task design can scale pathology-derived cancer staging and biomarker profiling, supporting clinical decision support and oncology research.

Abstract

Pathology reports serve as the definitive record for breast cancer staging, yet their unstructured format impedes large-scale data curation. While Large Language Models (LLMs) offer semantic reasoning, their deployment is often limited by high computational costs and hallucination risks. This study introduces a parameter-efficient, multi-task framework for automating the extraction of Tumor-Node-Metastasis (TNM) staging, histologic grade, and biomarkers. We fine-tune a Llama-3-8B-Instruct encoder using Low-Rank Adaptation (LoRA) on a curated, expert-verified dataset of 10,677 reports. Unlike generative approaches, our architecture utilizes parallel classification heads to enforce consistent schema adherence. Experimental results demonstrate that the model achieves a Macro F1 score of 0.976, successfully resolving complex contextual ambiguities and heterogeneous reporting formats that challenge traditional extraction methods including rule-based natural language processing (NLP) pipelines, zero-shot LLMs, and single-task LLM baselines. The proposed adapter-efficient, multi-task architecture enables reliable, scalable pathology-derived cancer staging and biomarker profiling, with the potential to enhance clinical decision support and accelerate data-driven oncology research.