concept
active
concept:adversarial-suffix-attackAdversarial Suffix Attack
Optimization-based jailbreak method appending strings to prompts to elicit harmful outputs.
Neighborhood — ranked by edge-count
Papers (1)
paper
Concepts (1)
concept
- Jailbreak Attackassociated_withSecurity attack that bypasses LLM safety alignment by suppressing deliberation or exploiting reflection inhibition.
Related by similarity (8)
cosine ≥ 0.65 · no typed edgeEntities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.
- Reduces superposition but needs 80% L2 norm.
- Competitive multi-agent setting with conflicting incentives and direct opposition via bidding and bluffing.
- Eight instruction variants appended to prompts to attempt to break superficial role-play and test depth of character
- Technique used in VPD to enforce mechanistic faithfulness of parameter decompositions.
- Risk that multiple truth directions enable attacks that shift outputs without triggering the primary truth direction
- Linked to superposition interference.
- Intervention used to control robustness.
- Adversarial multi-turn experiment where first turn uses pre-finetuning model to test if follow-up maintains character