diff --git a/src/google/adk/evaluation/local_eval_service.py b/src/google/adk/evaluation/local_eval_service.py
index 1a032bad64..9f759979c9 100644
--- a/src/google/adk/evaluation/local_eval_service.py
+++ b/src/google/adk/evaluation/local_eval_service.py
@@ -534,4 +534,5 @@ async def _perform_inference_single_eval_item(
       )
       inference_result.status = InferenceStatus.FAILURE
       inference_result.error_message = str(e)
+      inference_result.inferences = []  # Ensure inferences is not None on failure
       return inference_result
diff --git a/src/google/adk/optimization/gepa_root_agent_prompt_optimizer.py b/src/google/adk/optimization/gepa_root_agent_prompt_optimizer.py
index e9b82cdd50..5f0eb92d17 100644
--- a/src/google/adk/optimization/gepa_root_agent_prompt_optimizer.py
+++ b/src/google/adk/optimization/gepa_root_agent_prompt_optimizer.py
@@ -147,7 +147,7 @@ def evaluate(
       trajectories = []
 
       for example_id in batch:
-        score = result.scores[example_id]
+        score = result.scores.get(example_id, 0.0)
         scores.append(score)
 
         eval_data = result.data.get(example_id, {}) if result.data else {}
diff --git a/src/google/adk/optimization/local_eval_sampler.py b/src/google/adk/optimization/local_eval_sampler.py
index b00c34280f..e5d9a41f6c 100644
--- a/src/google/adk/optimization/local_eval_sampler.py
+++ b/src/google/adk/optimization/local_eval_sampler.py
@@ -289,7 +289,7 @@ def _extract_eval_data(
         for eval_metric_result in per_invocation_result.eval_metric_results:
           eval_metric_results.append({
               "metric_name": eval_metric_result.metric_name,
-              "score": round(eval_metric_result.score, 2),  # accurate enough
+              "score": round(eval_metric_result.score, 2) if eval_metric_result.score is not None else 0.0,
               "eval_status": eval_metric_result.eval_status.name,
           })
         per_invocation_result_dict = {