feat: impklent actual score

README.md

@@ -82,15 +82,15 @@ rails generate leva:eval sentiment_accuracy
 class SentimentAccuracyEval < Leva::BaseEval
   def evaluate(prediction, record)
     score = prediction == record.expected_label ? 1.0 : 0.0
-    score
+    [score, record.expected_label]
   end
 end
 
 class SentimentF1Eval < Leva::BaseEval
   def evaluate(prediction, record)
     # Calculate F1 score
     # ...
-    f1_score
+    [f1_score, record.f1_score]
   end
 end
 ```

app/models/leva/runner_result.rb

@@ -32,7 +32,6 @@ class RunnerResult < ApplicationRecord
     has_many :evaluation_results, dependent: :destroy
 
     validates :prediction, presence: true
-    validates :actual_result, presence: true
     validates :prompt, presence: true
   end
 end

lib/generators/leva/templates/eval.rb.erb

@@ -3,10 +3,10 @@
 class <%= class_name %>Eval < Leva::BaseEval
   # @param prediction [String] The prediction to evaluate
   # @param record [YourRecordClass] The record to evaluate
-  # @return [Float] The score of the evaluation
+  # @return [Array<(Float, Object)>] An array containing the evaluation score and the actual result.
   def evaluate(prediction, record)
     # Implement your evaluation logic here
 
-    1.0
+    [1.0, record.actual_result]
   end
 end

lib/leva.rb

@@ -57,7 +57,6 @@ def execute_and_store(experiment, dataset_record, prompt)
         dataset_record: dataset_record,
         prompt: prompt,
         prediction: result,
-        actual_result: dataset_record.actual_result
       )
     end
   end
@@ -71,7 +70,7 @@ class BaseEval
     #
     # @param prediction [Object] The model's prediction.
     # @param record [Object] The expected result.
-    # @return [Float] The evaluation score.
+    # @return [Array<(Float, Object)>] An array containing the evaluation score and the actual result.
     # @raise [NotImplementedError] if the method is not implemented in a subclass.
     def evaluate(prediction, record)
       raise NotImplementedError, "#{self.class} must implement #evaluate"
@@ -83,7 +82,11 @@ def evaluate(prediction, record)
     # @param runner_result [Leva::RunnerResult] The runner result to evaluate.
     # @return [Leva::EvaluationResult] The stored evaluation result.
     def evaluate_and_store(experiment, runner_result)
-      score = evaluate(runner_result.prediction, runner_result.dataset_record.recordable)
+      score, actual_result = evaluate(runner_result.prediction, runner_result.dataset_record.recordable)
+
+      if runner_result.actual_result.blank?
+        runner_result.update!(actual_result: actual_result)
+      end
 
       EvaluationResult.create!(
         experiment: experiment,

test/dummy/app/evals/sentiment_accuracy_eval.rb

@@ -6,5 +6,6 @@ class SentimentAccuracyEval < Leva::BaseEval
   # @return [Float] The score of the evaluation
   def evaluate(prediction, text_content)
     prediction == text_content.expected_label ? 1.0 : 0.0
+    [1.0, text_content.expected_label]
   end
 end

test/dummy/app/evals/sentiment_f1_eval.rb

@@ -6,6 +6,6 @@ class SentimentF1Eval < Leva::BaseEval
   # @return [Float] The score of the evaluation
   def evaluate(prediction, text_content)
     sleep 3
-    rand
+    [rand(0.0..1.0), 2]
   end
 end