GEPA.Adapters.Confidence.Scoring.LinearBlend (GEPA v0.3.0)

Copy Markdown View Source

Scores correct answers as 1.0 above a probability threshold, otherwise blends linearly from min_score_on_correct to 1.0. Incorrect answers always score 0.

Summary

Types

t()

Functions

describe(strategy)
new(opts \\ [])
score(strategy, correct, logprob)

Types

t()

@type t() :: %GEPA.Adapters.Confidence.Scoring.LinearBlend{
  low_confidence_threshold: float(),
  min_score_on_correct: float()
}

Functions

describe(strategy)

@spec describe(t()) :: String.t()

new(opts \\ [])

@spec new(keyword() | map()) :: t()

score(strategy, correct, logprob)

@spec score(t() | module(), boolean(), number() | nil) :: float()