Struct RewardModel 

pub struct RewardModel { /* private fields */ }

Running weighted-average reward per (model, prompt_template) key.

Each new reward is blended with the running average using an exponential decay factor: avg = decay * avg + (1 - decay) * new_reward.

Implementations

impl RewardModel

pub fn new(decay: f64) -> Self

Create a new model with the given exponential decay factor.

decay must be in (0, 1). A value of 0.9 means older observations retain 90% of their weight after each update.

pub fn update(&self, model: &str, template: &str, reward: f64)

Update the running average for (model, template) with a new reward.

pub fn average(&self, model: &str, template: &str) -> Option<f64>

Return the current weighted-average reward for (model, template).

pub fn history(&self, model: &str, template: &str) -> Vec<f64>

Return the full chronological reward history for (model, template).

Trait Implementations

impl Clone for RewardModel

fn clone(&self) -> RewardModel

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for RewardModel

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.