ASMIL: Attention-Stabilized Multiple Instance Learning for Whole Slide Imaging

2026-03-01

AI Summary

This work addresses key limitations of attention mechanisms in multiple instance learning for whole-slide image diagnosis, where unstable attention distributions, excessive concentration, and overfitting often degrade performance. To jointly mitigate these issues, the authors propose ASMIL, a novel framework that systematically integrates three complementary components: an anchor-based model to stabilize attention distributions, a normalized Sigmoid activation replacing Softmax to alleviate over-concentration, and token-wise stochastic dropout to suppress overfitting. These modules are designed to be generic and can be seamlessly incorporated into existing MIL architectures. Experimental results on public datasets demonstrate that ASMIL alone improves F1 score by up to 6.49%, and when integrated into baseline models, yields further gains of up to 10.73% in F1 score.

Technology Category

• Machine Learning: Multi-instance/Multi-view Learning
• Computer Vision: Multi-modal Vision
• Multiagent Systems: Multiagent Learning

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

Attention-based multiple instance learning (MIL) has emerged as a powerful framework for whole slide image (WSI) diagnosis, leveraging attention to aggregate instance-level features into bag-level predictions. Despite this success, we find that such methods exhibit a new failure mode: unstable attention dynamics. Across four representative attention-based MIL methods and two public WSI datasets, we observe that attention distributions oscillate across epochs rather than converging to a consistent pattern, degrading performance. This instability adds to two previously reported challenges: overfitting and over-concentrated attention distribution. To simultaneously overcome these three limitations, we introduce attention-stabilized multiple instance learning (ASMIL), a novel unified framework. ASMIL uses an anchor model to stabilize attention, replaces softmax with a normalized sigmoid function in the anchor to prevent over-concentration, and applies token random dropping to mitigate overfitting. Extensive experiments demonstrate that ASMIL achieves up to a 6.49\% F1 score improvement over state-of-the-art methods. Moreover, integrating the anchor model and normalized sigmoid into existing attention-based MIL methods consistently boosts their performance, with F1 score gains up to 10.73\%. All code and data are publicly available at https://github.com/Linfeng-Ye/ASMIL.

Problem

• attention instability
• multiple instance learning
• whole slide imaging
• overfitting
• attention concentration

Innovation

• attention stabilization
• multiple instance learning
• whole slide imaging
• normalized sigmoid
• token dropping