Struct Faithfulness 

pub struct Faithfulness<M: BaseChatModel> { /* private fields */ }

Faithfulness evaluator (hallucination detection): how faithful an answer is to the reference context.

Splits prediction into atomic claims and asks the judge per claim whether it can be derived from reference; pass rate = derivable claims / total claims.

Implementations

impl<M: BaseChatModel> Faithfulness<M>

pub fn new(judge: M) -> Self

Creates a faithfulness evaluator.

pub fn with_llm_split(self, v: bool) -> Self

Splits claims with the LLM (default: rule-based on punctuation; LLM split handles comma compound sentences)

pub fn with_empty_score(self, score: f64) -> Self

Score for an empty prediction: default 0.0 (no answer = not faithful); can be set to 1.0 for “not fabricating is faithful”

pub fn with_max_context_chars(self, max: usize) -> Self

Per-claim reference-context transmission cap (chars). P2-5: default 2000, preventing a long reference from being fully stuffed into the prompt once per claim.

Trait Implementations

impl<M: BaseChatModel> Evaluator for Faithfulness<M>

fn eval<'life0, 'life1, 'life2, 'life3, 'async_trait>( &'life0 self, _input: &'life1 str, prediction: &'life2 str, reference: &'life3 str, ) -> Pin<Box<dyn Future<Output = Result<Score, EvalError>> + Send + 'async_trait>>

where Self: 'async_trait, 'life0: 'async_trait, 'life1: 'async_trait, 'life2: 'async_trait, 'life3: 'async_trait,

Scores a single prediction

fn name(&self) -> &str

Evaluator name (used in report summaries)