MergeSlide: Continual Model Merging and Task-to-Class Prompt-Aligned Inference for Lifelong Learning on Whole Slide Images

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Bui, Doanh C., Ngo, Ba Hung, Pham, Hoai Luan, Nguyen, Khang, Nguyen, Maï K., Nakashima, Yasuhiko
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908659181682688
author Bui, Doanh C.
Ngo, Ba Hung
Pham, Hoai Luan
Nguyen, Khang
Nguyen, Maï K.
Nakashima, Yasuhiko
author_facet Bui, Doanh C.
Ngo, Ba Hung
Pham, Hoai Luan
Nguyen, Khang
Nguyen, Maï K.
Nakashima, Yasuhiko
contents Lifelong learning on Whole Slide Images (WSIs) aims to train or fine-tune a unified model sequentially on cancer-related tasks, reducing the resources and effort required for data transfer and processing, especially given the gigabyte-scale size of WSIs. In this paper, we introduce MergeSlide, a simple yet effective framework that treats lifelong learning as a model merging problem by leveraging a vision-language pathology foundation model. When a new task arrives, it is: 1) defined with class-aware prompts, 2) fine-tuned for a few epochs using an MLP-free backbone, and 3) merged into a unified model using an orthogonal continual merging strategy that preserves performance and mitigates catastrophic forgetting. For inference under the class-incremental learning (CLASS-IL) setting, where task identity is unknown, we introduce Task-to-Class Prompt-aligned (TCP) inference. Specifically, TCP first identifies the most relevant task using task-level prompts and then applies the corresponding class-aware prompts to generate predictions. To evaluate MergeSlide, we conduct experiments on a stream of six TCGA datasets. The results show that MergeSlide outperforms both rehearsal-based continual learning and vision-language zero-shot baselines. Code and data are available at https://github.com/caodoanh2001/MergeSlide.
format Preprint
id arxiv_https___arxiv_org_abs_2511_13099
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle MergeSlide: Continual Model Merging and Task-to-Class Prompt-Aligned Inference for Lifelong Learning on Whole Slide Images
Bui, Doanh C.
Ngo, Ba Hung
Pham, Hoai Luan
Nguyen, Khang
Nguyen, Maï K.
Nakashima, Yasuhiko
Computer Vision and Pattern Recognition
Lifelong learning on Whole Slide Images (WSIs) aims to train or fine-tune a unified model sequentially on cancer-related tasks, reducing the resources and effort required for data transfer and processing, especially given the gigabyte-scale size of WSIs. In this paper, we introduce MergeSlide, a simple yet effective framework that treats lifelong learning as a model merging problem by leveraging a vision-language pathology foundation model. When a new task arrives, it is: 1) defined with class-aware prompts, 2) fine-tuned for a few epochs using an MLP-free backbone, and 3) merged into a unified model using an orthogonal continual merging strategy that preserves performance and mitigates catastrophic forgetting. For inference under the class-incremental learning (CLASS-IL) setting, where task identity is unknown, we introduce Task-to-Class Prompt-aligned (TCP) inference. Specifically, TCP first identifies the most relevant task using task-level prompts and then applies the corresponding class-aware prompts to generate predictions. To evaluate MergeSlide, we conduct experiments on a stream of six TCGA datasets. The results show that MergeSlide outperforms both rehearsal-based continual learning and vision-language zero-shot baselines. Code and data are available at https://github.com/caodoanh2001/MergeSlide.
title MergeSlide: Continual Model Merging and Task-to-Class Prompt-Aligned Inference for Lifelong Learning on Whole Slide Images
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2511.13099