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

arXiv: https://arxiv.org/abs/2603.06658 · HTML: https://mlsanity.com/paper/2603.06658/

## Key facts

- Cited by 6 papers in the map (data through 14 Oct 2026).
- Builds on 45 of its 77 references that are in the map.
- First posted on arXiv 1 Mar 2026.

## Builds on (most cited first, 30 of 45)

- [Deep Residual Learning for Image Recognition](https://mlsanity.com/paper/1512.03385/) (2015)
- [A Simple Framework for Contrastive Learning of Visual Representations](https://mlsanity.com/paper/2002.05709/) (2020)
- [Distilling the Knowledge in a Neural Network](https://mlsanity.com/paper/1503.02531/) (2015)
- [ImageNet Large Scale Visual Recognition Challenge](https://mlsanity.com/paper/1409.0575/) (2014)
- [Emerging Properties in Self-Supervised Vision Transformers](https://mlsanity.com/paper/2104.14294/) (2021)
- [Bootstrap your own latent: A new approach to self-supervised Learning](https://mlsanity.com/paper/2006.07733/) (2020)
- [DINOv3](https://mlsanity.com/paper/2508.10104/) (2025)
- [Attention-based Deep Multiple Instance Learning](https://mlsanity.com/paper/1802.04712/) (2018)
- [Data Efficient and Weakly Supervised Computational Pathology on Whole Slide Images](https://mlsanity.com/paper/2004.09666/) (2020)
- [TransMIL: Transformer based Correlated Multiple Instance Learning for Whole Slide Image Classification](https://mlsanity.com/paper/2106.00908/) (2021)
- [Dual-stream Multiple Instance Learning Network for Whole Slide Image Classification with Self-supervised Contrastive Learning](https://mlsanity.com/paper/2011.08939/) (2020)
- [Nystr"omformer: A Nystr"om-Based Algorithm for Approximating Self-Attention](https://mlsanity.com/paper/2102.03902/) (2021)
- [From Softmax to Sparsemax: A Sparse Model of Attention and Multi-Label Classification](https://mlsanity.com/paper/1602.02068/) (2016)
- [DTFD-MIL: Double-Tier Feature Distillation Multiple Instance Learning for Histopathology Whole Slide Image Classification](https://mlsanity.com/paper/2203.12081/) (2022)
- [Whole Slide Images based Cancer Survival Prediction using Attention Guided Deep Multiple Instance Learning Networks](https://mlsanity.com/paper/2009.11169/) (2020)
- [Benchmarking Self-Supervised Learning on Diverse Pathology Datasets](https://mlsanity.com/paper/2212.04690/) (2022)
- [Visual Language Pretrained Multiple Instance Zero-Shot Transfer for Histopathology Images](https://mlsanity.com/paper/2306.07831/) (2023)
- [BRACS: A Dataset for BReAst Carcinoma Subtyping in H&E Histology Images](https://mlsanity.com/paper/2111.04740/) (2021)
- [Whole Slide Images are 2D Point Clouds: Context-Aware Survival Prediction using Patch-based Graph Convolutional Networks](https://mlsanity.com/paper/2107.13048/) (2021)
- [Interventional Bag Multi-Instance Learning On Whole-Slide Pathological Images](https://mlsanity.com/paper/2303.06873/) (2023)
- [Morphological Prototyping for Unsupervised Slide Representation Learning in Computational Pathology](https://mlsanity.com/paper/2405.11643/) (2024)
- [Feature Re-Embedding: Towards Foundation Model-Level Performance in Computational Pathology](https://mlsanity.com/paper/2402.17228/) (2024)
- [Attention-Challenging Multiple Instance Learning for Whole Slide Image Classification](https://mlsanity.com/paper/2311.07125/) (2023)
- [Multiple Instance Learning Framework with Masked Hard Instance Mining for Whole Slide Image Classification](https://mlsanity.com/paper/2307.15254/) (2023)
- [MetaGCD: Learning to Continually Learn in Generalized Category Discovery](https://mlsanity.com/paper/2308.11063/) (2023)
- [Classifying and Segmenting Microscopy Images Using Convolutional Multiple Instance Learning](https://mlsanity.com/paper/1511.05286/) (2015)
- [Test-Time Domain Adaptation by Learning Domain-Aware Batch Normalization](https://mlsanity.com/paper/2312.10165/) (2023)
- [Do Multiple Instance Learning Models Transfer?](https://mlsanity.com/paper/2506.09022/) (2025)
- [PathGen-1.6M: 1.6 Million Pathology Image-text Pairs Generation through Multi-agent Collaboration](https://mlsanity.com/paper/2407.00203/) (2024)
- [Bayes Conditional Distribution Estimation for Knowledge Distillation Based on Conditional Mutual Information](https://mlsanity.com/paper/2401.08732/) (2024)

## Built on by (most cited first, 6 of 6)

- [Are Compact Rationales Free? Measuring Tile Selection Headroom in Frozen WSI-MIL](https://mlsanity.com/paper/2605.12575/) (2026)
- [From Visual Widgets to UI Code: Efficient Tool-Grounded Generation](https://mlsanity.com/paper/2608.12611/) (2026)
- [Learning through Creation: A Hash-Free Framework for On-the-Fly Category Discovery](https://mlsanity.com/paper/2603.13858/) (2026)
- [GCE-MIL: Faithful and Recoverable Evidence for Multiple Instance Learning in Whole-Slide Imaging](https://mlsanity.com/paper/2605.17456/) (2026)
- [CARA: Collision-Aware Resolution Adaptation for Multiresolution Hash Encoding Based Image Fitting](https://mlsanity.com/paper/2609.18554/) (2026)
- [One Tile, Multiple Instances: Rethinking MIL for Sparse Diagnostic Evidence](https://mlsanity.com/paper/2610.04853/) (2026)

## Abstract (arXiv)

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.