TransFool: An Adversarial Attack against Neural Machine Translation Models

This upload contains the language models utilized in the paper "TransFool: An Adversarial Attack against Neural Machine Translation Models". TransFool is an adversarial attack against NMT models. By generating adversarial examples, TransFool aims to reduce translation quality while maintaining similarity to the original sentences.

These language models, along with their respective fully connected layers, are trained for the NMT models discussed in the paper. For further details on the implementation and usage of TransFool, please refer to the official GitHub repository.