Parameter-Efficient Subspace Decoupling ViT for Mitigating Multi-Task Negative Transfer in Histological Scoring

2026-05-28

AI Summary

This work addresses the challenges of negative transfer and high annotation costs in histological multi-task assessment of non-alcoholic fatty liver disease (NAFLD), which arise due to strong inter-task correlations. To mitigate these issues, the authors propose a subspace-decoupled multi-task Vision Transformer that introduces lightweight task-specific Adapters with orthogonality constraints. This design enables the construction of independent feature subspaces for steatosis, ballooning, and inflammation, effectively reducing inter-task interference while preserving shared representations. Experiments on a newly curated murine NAFLD multi-task dataset demonstrate that the proposed method significantly enhances model stability and generalization, alleviates negative transfer, and achieves efficient learning with substantially lower computational overhead compared to training separate single-task models.

Abstract

Histological scoring is essential for diagnosing Non-Alcoholic Fatty Liver Disease (NAFLD), yet its automation remains challenging due to the high annotation cost and negative transfer among the strongly correlated NAFLD Activity Score (NAS) indicators in multi-task learning. To address this issue, we propose a subspace-decoupled multi-task Vision Transformer (ViT) that integrates lightweight task-specific Adapters with orthogonality-based constraints. This design constructs independent feature subspaces for steatosis, ballooning, and inflammation, effectively reducing task interference while retaining shared representations. We further construct a curated multi-task mouse NAFLD histology dataset with expert annotations for all NAS components. Experimental results demonstrate that the proposed method improves multi-task stability and generalization with substantially reduced computational cost compared to training separate single-task models. The code and the curated dataset have been prepared and will be made publicly available upon acceptance to support reproducibility.

Problem

• multi-task negative transfer
• histological scoring
• NAFLD Activity Score
• task interference
• multi-task learning

Innovation

• Subspace Decoupling
• Parameter-Efficient Adapter
• Multi-Task Negative Transfer
• Vision Transformer
• Orthogonality Constraint