BoN

Best-of-N augmentation search with inline judge evaluation.

Adaptive. Searches: it reads the target's replies and refines its prompts, which is why it needs an attacker model.

A stochastic search over augmentations of the goal.

Helper models

None. This attack drives no model of its own, so roles can be left out.

Settings for the algorithm itself, written under the attack's parameters. Anything left out keeps its default.

Field	Type	Default	Description
`steps`	integer (>= 1)	`4`	Sequential search steps. Each one sends `candidates` requests.
`candidates`	integer (>= 1)	`5`	Independently seeded candidates generated per step.
`sigma`	number (> 0.0, <= 1.0)	`0.4`	Augmentation strength, from untouched (0) to heavily perturbed (1).
`word_scrambling`	boolean	`true`	Shuffle the middle letters of words.
`random_capitalization`	boolean	`true`	Randomly change the case of letters.
`ascii_perturbation`	boolean	`true`	Randomly shift some characters to nearby ASCII characters.

attacks:
- name: bon