diff --git a/src/lerobot/datasets/lerobot_dataset.py b/src/lerobot/datasets/lerobot_dataset.py
index a1aeae0ce..9484a4342 100644
--- a/src/lerobot/datasets/lerobot_dataset.py
+++ b/src/lerobot/datasets/lerobot_dataset.py
@@ -863,17 +863,16 @@ class LeRobotDataset(torch.utils.data.Dataset):
 
         features = get_hf_features_from_features(self.features)
 
-        # This is the v2.1 logic that forces an efficient, pre-decoded cache build.
-        # This is the key to performance for dtype="image" datasets.
-
         # 1. Check if specific episodes are requested by the user.
+        #    This is the "data_files" path, which may be slow, but is
+        #    necessary for visualization or evaluation on a subset.
         if self.episodes is not None:
             # Get the unique set of parquet files for the requested episodes
             fpaths = set()
             for ep_idx in self.episodes:
                 # Need to read metadata to find the file path for this episode
-                # We use self.meta.episodes (the loaded dataset) here
-                ep_meta = self.meta.episodes[ep_idx]
+                # Use the pre-loaded metadata list
+                ep_meta = self.episodes_metadata_list[ep_idx]
                 chunk_idx = ep_meta["data/chunk_index"]
                 file_idx = ep_meta["data/file_index"]
                 fpath_str = self.meta.data_path.format(chunk_index=chunk_idx, file_index=file_idx)
@@ -886,17 +885,15 @@ class LeRobotDataset(torch.utils.data.Dataset):
             )
 
             # Filter the loaded dataset to *only* include the requested episodes
-            # This is necessary because the v3 files contain multiple episodes.
+            # This is necessary because v3 files can contain multiple episodes.
             requested_episodes_set = set(self.episodes)
             hf_dataset = hf_dataset.filter(
-                lambda x: x["episode_index"] in requested_episodes_set,
-                batched=True,  # Use batched=True for faster filtering
-                batch_size=1000,
+                lambda x: x["episode_index"] in requested_episodes_set, batched=True, batch_size=1000
             )
 
         else:
             # THIS IS THE FAST PATH FOR TRAINING (self.episodes is None)
-            # Load all data files using data_dir, which is the most efficient.
+            # We must use `data_dir` to trigger the v2.1-style efficient cache.
             data_dir = str(self.root / "data")
             hf_dataset = datasets.load_dataset("parquet", data_dir=data_dir, features=features, split="train")