uqlm.nli.cluster.SemanticClusterer

class uqlm.nli.cluster.SemanticClusterer(nli=None)

Bases: object

__init__(nli=None)

Methods

__init__([nli])
cluster_responses(responses[, prompt])	This method create clusters from a list of responses based on the semantic meaning of each response.
evaluate(responses[, prompt, ...])	Evaluate the cluster of responses.

cluster_responses(responses, prompt=None)

This method create clusters from a list of responses based on the semantic meaning of each response.

Return type:

Any

Parameters:

• responses (list of str, default=None) – A list of model responses

• prompt (str, default=None) – A prompt for the responses.

Return type:

A list of lists, where each list represents a cluster.

evaluate(responses, prompt=None, response_probabilities=None)

Evaluate the cluster of responses.

Return type:

Tuple[str, List[List[str]], List[float], Dict[Tuple[str, str], float]]

References