diff --git a/src/lerobot/configs/rewards.py b/src/lerobot/configs/rewards.py
index a53d5a417..7e99e7f71 100644
--- a/src/lerobot/configs/rewards.py
+++ b/src/lerobot/configs/rewards.py
@@ -90,9 +90,9 @@ class RewardModelConfig(draccus.ChoiceRegistry, HubMixin, abc.ABC):
     def reward_delta_indices(self) -> list | None:  # type: ignore[type-arg]
         return None
 
-    @abc.abstractmethod
-    def get_optimizer_preset(self) -> OptimizerConfig:
-        raise NotImplementedError
+    def get_optimizer_preset(self) -> OptimizerConfig | None:
+        """Default optimizer for this reward model, or ``None`` for zero-shot models."""
+        return None
 
     def get_scheduler_preset(self) -> LRSchedulerConfig | None:
         return None
diff --git a/src/lerobot/templates/lerobot_modelcard_template.md b/src/lerobot/templates/lerobot_modelcard_template.md
index f0dd0da07..b93e83b6e 100644
--- a/src/lerobot/templates/lerobot_modelcard_template.md
+++ b/src/lerobot/templates/lerobot_modelcard_template.md
@@ -41,8 +41,6 @@ For more details, see the [Physical Intelligence π₀ blog post](https://www.ph
 For more details, see the [Physical Intelligence π₀.₅ blog post](https://www.physicalintelligence.company/blog/pi05).
 {% elif model_name == "gaussian_actor" %}
 This is a Gaussian Actor policy (Gaussian policy with a tanh squash) — the policy-side component used by [Soft Actor-Critic (SAC)](https://huggingface.co/papers/1801.01290) and related maximum-entropy continuous-control algorithms.
-{% elif model_name == "reward_classifier" %}
-A reward classifier is a lightweight neural network that scores observations or trajectories for task success, providing a learned reward signal or offline evaluation when explicit rewards are unavailable.
 {% else %}
 _Model type not recognized — please update this template._
 {% endif %}