FIELD NOTES TOPIC
fast multi-dimensional refusal subspaces via rfm-agop
1 note, newest first.
Fast Multi-dimensional Refusal Subspaces via RFM-AGOP
A new method called Fast Multi-dimensional Refusal Subspaces via RFM-AGOP is proposed for extracting complex behaviors, such as the refusal to answer harmful queries, from Large Language Models (LLMs). This method…
