This paper addresses the problem of Multiple Instance Learning (MIL), where a single class label is assigned to a bag of instances. The authors propose a neural network-based approach that models the Bernoulli distribution of the bag label, parameterized by neural networks. They introduce a permutation-invariant aggregation operator, inspired by the attention mechanism, which allows for the identification of key instances contributing to the bag label. The proposed method is evaluated on various benchmark MIL datasets and real-life histopathology datasets, demonstrating comparable or superior performance to existing methods while providing interpretability through the attention weights. The attention-based MIL pooling is flexible and can be adapted to different tasks and data, making it a promising approach for MIL problems, especially in medical imaging applications.This paper addresses the problem of Multiple Instance Learning (MIL), where a single class label is assigned to a bag of instances. The authors propose a neural network-based approach that models the Bernoulli distribution of the bag label, parameterized by neural networks. They introduce a permutation-invariant aggregation operator, inspired by the attention mechanism, which allows for the identification of key instances contributing to the bag label. The proposed method is evaluated on various benchmark MIL datasets and real-life histopathology datasets, demonstrating comparable or superior performance to existing methods while providing interpretability through the attention weights. The attention-based MIL pooling is flexible and can be adapted to different tasks and data, making it a promising approach for MIL problems, especially in medical imaging applications.