Interface ReinforcementTuningStringMatchRewardScorer

Scores parsed responses for string matching use cases.

interface ReinforcementTuningStringMatchRewardScorer {
    correctAnswerReward?: number;
    jsonMatchExpression?: ReinforcementTuningStringMatchRewardScorerJsonMatchExpression;
    stringMatchExpression?: ReinforcementTuningStringMatchRewardScorerStringMatchExpression;
    wrongAnswerReward?: number;
}

Index

Properties

correctAnswerReward? jsonMatchExpression? stringMatchExpression? wrongAnswerReward?

Properties

`Optional`correctAnswerReward

correctAnswerReward?: number

Correct answer reward is returned if evaluator evaluates to true. All correct answers get the same reward.

`Optional`jsonMatchExpression

jsonMatchExpression?: ReinforcementTuningStringMatchRewardScorerJsonMatchExpression

Uses json match expression to evaluate parsed response.

`Optional`stringMatchExpression

stringMatchExpression?: ReinforcementTuningStringMatchRewardScorerStringMatchExpression

Uses string match expression to evaluate parsed response.

`Optional`wrongAnswerReward

wrongAnswerReward?: number

Wrong answer reward is returned if evaluator evaluates to false. All wrong answers get the same reward.