docs: de-duplicate sidebar navigation under section captions

The left-sidebar globaltoc was showing each Dataset/IL/RL child page
twice — once flat under the caption and again nested inside an
"Overview" item that pointed at the section index, because the
section index's own ``toctree`` re-listed the same children.

Two structural changes, no content removed:

1. ``docs/source/index.md``: drop the redundant ``Overview <section/index>``
   line from the captioned root toctrees for Dataset & Benchmark and
   Imitation Learning. The flat listing of leaf pages stays — that's
   what the sidebar surfaces.

2. ``docs/source/dataset_benchmark/index.md`` and
   ``docs/source/roboverse_learn/index.md``: mark as ``:orphan:`` and
   remove their internal captioned toctrees. The pages stay accessible
   at their URLs (with all their body content) but no longer act as
   structural parents in the global toctree, which was the source of
   the duplication.

The four sidebar captions remain (Dataset & Benchmark / Imitation
Learning / Reinforcement Learning / FAQ); each now lists its leaves
exactly once.

Author: geng-haoran

docs/source/dataset_benchmark/index.md

@@ -1,3 +1,7 @@
+---
+orphan: true
+---
+
 # Dataset and Benchmark
 
 This section documents the RoboVerse task suite, asset library, and evaluation protocols.
@@ -63,31 +67,3 @@ See [Benchmark Overview](benchmark/overview.md) for detailed evaluation protocol
 
 ---
 
-## Table of Contents
-
-```{toctree}
-:caption: RoboVerse Tasks
-:titlesonly:
-
-tasks/overview
-tasks/descriptions
-tasks/task_groups
-```
-
-```{toctree}
-:caption: RoboVerse Dataset
-:titlesonly:
-
-dataset/robots
-dataset/objects
-dataset/scenes
-```
-
-```{toctree}
-:caption: RoboVerse Benchmark
-:titlesonly:
-
-benchmark/overview
-benchmark/results
-benchmark/usage
-```

docs/source/index.md

@@ -141,7 +141,6 @@ If you find this work useful in your research, please consider citing:
 :titlesonly:
 :hidden:
 
-Overview <dataset_benchmark/index>
 Tasks Overview <dataset_benchmark/tasks/overview>
 Task Descriptions <dataset_benchmark/tasks/descriptions>
 Task Groups <dataset_benchmark/tasks/task_groups>
@@ -159,7 +158,6 @@ Benchmark Usage <dataset_benchmark/benchmark/usage>
 :titlesonly:
 :hidden:
 
-Overview <roboverse_learn/index>
 Diffusion Policy <roboverse_learn/imitation_learning/diffusion_policy>
 ACT <roboverse_learn/imitation_learning/ACT>
 OpenVLA <roboverse_learn/imitation_learning/openvla>

docs/source/roboverse_learn/index.md

@@ -1,3 +1,7 @@
+---
+orphan: true
+---
+
 # RoboVerse Learn
 
 **RoboVerse Learn** provides a comprehensive suite of learning algorithms for robot policy training. It integrates seamlessly with MetaSim environments, enabling end-to-end training pipelines for both imitation learning and reinforcement learning.
@@ -115,30 +119,3 @@ pip install -e ".[vla]"
 ## Contributing
 
 Want to add a new algorithm? See our [Contributing Guide](imitation_learning/contributing.md) for instructions on integrating new methods.
-
----
-
-```{toctree}
-:caption: Imitation Learning
-:titlesonly:
-
-imitation_learning/diffusion_policy
-imitation_learning/ACT
-imitation_learning/openvla
-imitation_learning/smolvla
-imitation_learning/rdt
-imitation_learning/octo
-imitation_learning/contributing
-```
-
-```{toctree}
-:caption: Reinforcement Learning
-:titlesonly:
-
-reinforcement_learning/ppo
-reinforcement_learning/fast_td3
-reinforcement_learning/sac
-reinforcement_learning/td3
-reinforcement_learning/skillblender_rl
-reinforcement_learning/humanoid
-```