JudgeResult#

class scikitplot.corpus.JudgeResult(metric, score, verdict=None, completions=(), error=None, undefined=None, details=<factory>)[source]#

One judged score.

Parameters:

metricstr

Metric name.

scorefloat

In [0, 1], or NaN when error or undefined is set.

verdictmapping or None, default=None

The parsed verdict (None when the judge failed). For preference, {"forward": ..., "reverse": ...}.

completionstuple of Completion, default=()

Every model response the score used, in call order.

errorstr or None, default=None

Why the judge failed: "parse: ...", "incomplete: ..." or "provider: <ExceptionType>: ...".

undefinedstr or None, default=None

Why a valid verdict has no score (for example, no claims).

detailsmapping, default={}

Counts behind the score (n_claims, n_supported …); for preference, winner and position_consistent.

Parameters:

• metric (str)

• score (float)

• verdict (Mapping[str, Any] | None)

• completions (tuple)

• error (str | None)

• undefined (str | None)

• details (Mapping[str, Any])

completions: tuple = ()#

details: Mapping[str, Any][source]#

error: str | None = None#

classmethod from_dict(data)[source]#

Rebuild from to_dict output (None score is NaN).

Parameters:

data (Mapping[str, Any])

Return type:

JudgeResult

property input_tokens: int | None#

Prompt tokens over completions (None if any is unreported).

metric: str[source]#

property output_tokens: int | None#

Output tokens over completions (None if any is unreported).

score: float[source]#

to_dict()[source]#

Return a JSON-compatible mapping.

Return type:

dict[str, Any]

undefined: str | None = None#

verdict: Mapping[str, Any] | None = None#