diff --git a/src/alignrl/grpo.py b/src/alignrl/grpo.py
index 3e1407b..3eaf7da 100644
--- a/src/alignrl/grpo.py
+++ b/src/alignrl/grpo.py
@@ -120,7 +120,7 @@ def train(self) -> TrainResult:
             save_strategy="steps",
             save_steps=50,
         )
-        if self.config.reward_weights:
+        if self.config.reward_weights is not None:
             grpo_args.reward_weights = self.config.reward_weights
 
         trainer = GRPOTrainer(