Paper

TRACE: Data-Free Text Reconstruction Attacks against Approximate Unlearning in LLMs · ByteTrail