Skip to main content

module cleanlab_codex.types.tlm

Types for TLM.

class TLMOptions​

Customization options for querying TLM. For details, see the TLM documentation.

class TLMOptions(TypedDict):
     custom_eval_criteria: 'typing.Iterable[object]'
     log: 'typing.List[str]'
     max_tokens: 'int'
     model: 'str'
     num_candidate_responses: 'int'
     num_consistency_samples: 'int'
     reasoning_effort: 'str'
     similarity_measure: 'str'
     use_self_reflection: 'bool'

class TLMPromptResponse​

The response from prompting TLM.

class TLMPromptResponse:
     response: 'str'
     trustworthiness_score: 'float'
     log: 'typing.Optional[object]'

class TLMScoreResponse​

The result of scoring a response with TLM.

class TLMScoreResponse:
     trustworthiness_score: 'float'
     log: 'typing.Optional[object]'

class TLMConfig​

Advanced configuration options for TLM.

class TLMConfig(BaseModel):
     quality_preset: 'typing.Optional[typing.Literal['best', 'high', 'medium', 'low', 'base']]'
     task: 'typing.Optional[str]'
     options: 'typing.Optional[cleanlab_codex.types.tlm.TLMOptions]'

Args: