pyrit.score.RefusalScorerPaths

pyrit.score.RefusalScorerPaths#

class RefusalScorerPaths(value, names=None, *, module=None, qualname=None, type=None, start=1, boundary=None)[source]#

Bases: Enum

Paths to refusal scorer system prompt YAML files.

Each enum value represents a different refusal detection strategy:

  • DEFAULT: Standard refusal detection that works with or without an explicit objective.

    If an objective is provided, evaluates refusal against it; if not, evaluates against the implied objective. Safe completions (including partial information, redirections, asking questions, or excessive caveats) are NOT considered refusals.

  • STRICT: Strict refusal detection that treats “safe completions” as refusals. Works best when

    an explicit objective is provided.

__init__(*args, **kwds)#

Attributes

DEFAULT

STRICT

DEFAULT = PosixPath('/home/runner/work/PyRIT/PyRIT/pyrit/datasets/score/refusal/refusal_default.yaml')#
STRICT = PosixPath('/home/runner/work/PyRIT/PyRIT/pyrit/datasets/score/refusal/refusal_strict.yaml')#
Contents