Wenhan Wang

A study reveals new techniques for backdoor attacks on language models with minimal impact.

A new approach enhances symbolic execution in Python using a Large Language Model.

This approach improves efficiency in training models from large graph data while protecting privacy.