Soft Prompt Threats: Attacking Safety Alignment and Unlearning in Open-Source LLMs through the Embedding Space

Item #:

079017-0288

Details

DOI: https://doi.org/10.52202/079017-0288
Author(s): David Dobre, Gauthier Gidel, Stephan Günnemann, Leo Schwinn, Sophie Xhonneux
Pages: 9086-9116 (31 pages)
Format: PDF Paper Download
Conference: Advances in Neural Information Processing Systems 37
Date/Location: Held 10-15 December 2024, Vancouver, Canada.
Series: Advances in Neural Information Processing Systems 37
Publisher: Neural Information Processing Systems Foundation, Inc. (NeurIPS)

DOI: https://doi.org/10.52202/079017-0288
Author(s): David Dobre, Gauthier Gidel, Stephan Günnemann, Leo Schwinn, Sophie Xhonneux
Pages: 9086-9116 (31 pages)
Format: PDF Paper Download
Conference: Advances in Neural Information Processing Systems 37
Date/Location: Held 10-15 December 2024, Vancouver, Canada.
Series: Advances in Neural Information Processing Systems 37
Publisher: Neural Information Processing Systems Foundation, Inc. (NeurIPS)