Add annotated language instructions

Upload meta/tasks_annotated.parquet and document the per-episode language annotations in the dataset card.

README.md

@@ -22,6 +22,12 @@ This dataset was created using [LeRobot](https://github.com/huggingface/lerobot)
 - **Paper:** [More Information Needed]
 - **License:** apache-2.0
 
+## Language Annotations
+
+This dataset includes annotated language instructions in [`meta/tasks_annotated.parquet`](meta/tasks_annotated.parquet). The file is indexed by `episode_index` and has a `task` column containing our per-episode annotated instruction.
+
+The standard LeRobot loader resolves a frame's language instruction through `task_index`: each data row stores a `task_index`, which is looked up in [`meta/tasks.parquet`](meta/tasks.parquet). In our training code, setting `--use_annotated_task true` loads `meta/tasks_annotated.parquet` and replaces the frame's task by looking up the current `episode_index`; if no valid annotated row is available, it falls back to the standard LeRobot task.
+
 ## Dataset Structure
 
 [meta/info.json](meta/info.json):

meta/tasks_annotated.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:db585b5a8f633fd6b0281df238b43ec928a1cf558653828072a6fe8c07b8ea6b
+size 3509