Doublespeak: In-Context Representation Hijacking

Research project that introduces Doublespeak, an in-context representation hijacking attack against large language models where harmful keywords are systematically replaced with benign tokens across multiple in-context examples so that the benign token’s internal representation acquires the harmful semantics and bypasses standard safety alignment checks.

Endorsements support MentaLeap.

No endorsements yet.

Discussion

Start Date

-

End Date

-

Expected Duration

-

Funding Raised to Date

-