Stealing Training Data from Large Language Models in Decentralized Training through Activation Inversion Attack.

Chenxi Dai, Lin Lu, Pan Zhou

Browse the full ACL paper archive.