Introduction to MultiVec2Text

MultiVec2Text supports research in Text Embedding Inversion Security in NLP for large language models, specifically multilingual ones, extended upon Jack Morris' Vec2Text Repository.

The models and datasets used for model training and evaluation are open sourced in Huggingface 🤗 yiyic.

Until now, the inversion models are trained on LUMI (LUMI-G nodes). The instruction will be given in the wiki-pages.