remove wrong information in doc

geng-haoran · geng-haoran · commit 485e7bf0a556 · 2026-01-31T23:51:09.000+08:00
diff --git a/docs/source/FAQ/index.md b/docs/source/FAQ/index.md
@@ -8,36 +8,15 @@ This page answers common questions about RoboVerse. For simulator-specific issue
 
 ### What is RoboVerse?
 
-RoboVerse is a unified platform for scalable and generalizable robot learning. It provides:
-
-- A multi-simulator abstraction layer (MetaSim) supporting 11+ physics engines
-- A large-scale dataset with 200+ tasks and 50+ robot embodiments
-- Integrated learning algorithms for imitation learning and reinforcement learning
+RoboVerse is a unified platform for scalable and generalizable robot learning. See the [project page](https://roboverseorg.github.io) and [paper](https://arxiv.org/abs/2504.18904) for details.
 
 ### What simulators are supported?
 
-| Simulator | GPU Parallel | Rendering | Best For |
-|-----------|--------------|-----------|----------|
-| MuJoCo | Via MJX | Native | Fast prototyping, RL |
-| Isaac Sim | Native | RTX | High-fidelity visuals, sim2real |
-| SAPIEN 3 | Native | Ray Tracing | Manipulation tasks |
-| PyBullet | - | OpenGL | Debugging, education |
-| Genesis | Native | Native | Large-scale training |
-| MJX | Native | - | Massively parallel RL |
-
-See the full [Support Matrix](../metasim/features/support_matrix.rst) for details.
+See the [Support Matrix](../metasim/features/support_matrix.rst) for the full list of supported simulators and features.
 
 ### What robots are available?
 
-RoboVerse includes 50+ robot configurations across categories:
-
-- **Manipulators**: Franka Panda, UR5/UR10, Kinova, xArm, Realman, Flexiv, etc.
-- **Mobile Manipulators**: Fetch, Tiago, Stretch
-- **Humanoids**: Unitree H1/G1, Fourier GR1, Digit
-- **Dexterous Hands**: Allegro, Shadow, LEAP Hand
-- **Quadrupeds**: Unitree Go1/A1, ANYmal
-
-See [Robots Dataset](../dataset_benchmark/dataset/robots.md) for the complete list.
+See [Robots Dataset](../dataset_benchmark/dataset/robots.md) for the complete list of available robots.
 
 ---
 
diff --git a/docs/source/dataset_benchmark/index.md b/docs/source/dataset_benchmark/index.md
@@ -1,45 +1,6 @@
 # Dataset and Benchmark
 
-RoboVerse provides a large-scale dataset and standardized benchmarks for robot learning research. This section documents our task suite, asset library, and evaluation protocols.
-
----
-
-## Overview
-
-::::{grid} 3
-:gutter: 3
-
-:::{grid-item-card} 200+ Tasks
-:text-align: center
-
-Manipulation, locomotion, and navigation tasks across difficulty levels
-:::
-
-:::{grid-item-card} 50+ Robots
-:text-align: center
-
-Manipulators, mobile robots, humanoids, and dexterous hands
-:::
-
-:::{grid-item-card} 1000+ Objects
-:text-align: center
-
-Diverse object assets from household items to industrial parts
-:::
-
-::::
-
----
-
-## Dataset Statistics
-
-| Category | Count | Description |
-|----------|-------|-------------|
-| **Tasks** | 200+ | Manipulation, locomotion, navigation |
-| **Robots** | 50+ | Arms, mobile bases, humanoids, hands |
-| **Objects** | 1000+ | Household, industrial, articulated |
-| **Scenes** | 50+ | Tables, shelves, kitchens, factories |
-| **Demonstrations** | 100K+ | Expert trajectories for IL training |
+This section documents the RoboVerse task suite, asset library, and evaluation protocols.
 
 ---
 
diff --git a/docs/source/index.md b/docs/source/index.md
@@ -16,12 +16,7 @@
 
 ## What is RoboVerse?
 
-**RoboVerse** is a unified platform for scalable and generalizable robot learning, providing:
-
-- **Multi-Simulator Support**: A unified interface across 11 simulators including MuJoCo, Isaac Sim, SAPIEN, PyBullet, Genesis, and more
-- **Large-Scale Dataset**: 200+ manipulation and locomotion tasks with diverse robots and objects  
-- **Cross-Embodiment**: Seamlessly transfer skills across 50+ robot embodiments
-- **Learning Algorithms**: Integrated imitation learning (ACT, Diffusion Policy, OpenVLA) and reinforcement learning (PPO, TD3, SAC)
+**RoboVerse** is a unified platform for scalable and generalizable robot learning. See the [project page](https://roboverseorg.github.io) and [paper](https://arxiv.org/abs/2504.18904) for more details.
 
 ---
 
@@ -80,7 +75,7 @@ Core simulation framework documentation including installation, tutorials, conce
 :link: dataset_benchmark/index
 :link-type: doc
 
-Explore 200+ tasks, robot configurations, object assets, and benchmark results.
+Explore tasks, robot configurations, object assets, and benchmark results.
 :::
 
 :::{grid-item-card} RoboVerse Learn
@@ -101,40 +96,6 @@ Complete API documentation for MetaSim modules.
 
 ---
 
-## Key Features
-
-### Cross-Simulator Compatibility
-
-Write once, run anywhere. RoboVerse provides a unified API that works across multiple physics simulators:
-
-| Simulator | GPU Parallel | Rendering | Status |
-|-----------|--------------|-----------|--------|
-| MuJoCo | Via MJX | Native | Stable |
-| Isaac Sim | Native | RTX | Stable |
-| SAPIEN 3 | Native | Ray Tracing | Stable |
-| PyBullet | - | OpenGL | Stable |
-| Genesis | Native | Native | Beta |
-
-### Cross-Embodiment Transfer
-
-Train policies that generalize across different robot embodiments:
-
-- **Manipulators**: Franka Panda, UR5/UR10, Kinova, xArm, and more
-- **Mobile Robots**: Fetch, Tiago, Stretch
-- **Humanoids**: H1, G1, GR1
-- **Dexterous Hands**: Allegro, Shadow, LEAP
-
-### Domain Randomization
-
-Built-in support for comprehensive domain randomization:
-
-- Physics parameters (mass, friction, damping)
-- Visual properties (textures, lighting, colors)
-- Sensor noise and latency
-- Object poses and configurations
-
----
-
 ## System Architecture
 
 <p align="center">
diff --git a/docs/source/metasim/features/cross_embodiment.md b/docs/source/metasim/features/cross_embodiment.md
@@ -1,22 +1,8 @@
 # Cross-Embodiment Transfer
 
-Cross-embodiment is a key feature of RoboVerse that enables policies and demonstrations to be transferred across different robot morphologies. This allows you to:
+Cross-embodiment enables policies and demonstrations to be transferred across different robot morphologies.
 
-- Train on one robot and deploy on another
-- Leverage demonstrations from multiple robot types
-- Test generalization across embodiments
-
----
-
-## Supported Robot Categories
-
-| Category | Robots | Gripper Type |
-|----------|--------|--------------|
-| **7-DOF Arms** | Franka Panda, KUKA iiwa, Kinova Gen3 | Parallel gripper |
-| **6-DOF Arms** | UR5, UR10, xArm6 | Parallel gripper |
-| **Mobile Manipulators** | Fetch, Tiago, Stretch | Various |
-| **Dexterous Hands** | Allegro, Shadow, LEAP | Multi-finger |
-| **Humanoids** | H1, G1, GR1 | Anthropomorphic |
+See [Robots Dataset](../../dataset_benchmark/dataset/robots.md) for the full list of supported robots.
 
 ---
 
diff --git a/docs/source/metasim/features/cross_sim.md b/docs/source/metasim/features/cross_sim.md
@@ -65,57 +65,3 @@ python scripts/advanced/replay_demo.py --sim=sapien3 --task=PickCube
 ### Feature Parity
 
 Not all features are available across all simulators. Check the [Support Matrix](support_matrix.rst) for detailed compatibility information.
-
-### Performance Differences
-
-| Simulator | Single Env Speed | Parallel Scaling | Rendering Quality |
-|-----------|------------------|------------------|-------------------|
-| MuJoCo | Fast | Good (via MJX) | Medium |
-| Isaac Sim | Medium | Excellent | Excellent (RTX) |
-| SAPIEN 3 | Fast | Good | Good |
-| Genesis | Fast | Excellent | Good |
-
-### Asset Compatibility
-
-Different simulators may require different asset formats:
-
-| Simulator | Preferred Format |
-|-----------|------------------|
-| MuJoCo | MJCF (.xml) |
-| Isaac Sim | USD (.usd, .usda) |
-| SAPIEN | URDF (.urdf) |
-| PyBullet | URDF (.urdf) |
-
-MetaSim handles format conversion automatically when possible.
-
----
-
-## Example: Multi-Simulator Evaluation
-
-Evaluate the same policy across different simulators to test sim2sim transfer:
-
-```python
-from metasim.scenario.scenario import ScenarioCfg
-from metasim.task.registry import get_task_class
-
-task_cls = get_task_class("StackCube")
-
-simulators = ["mujoco", "sapien3", "isaacsim"]
-results = {}
-
-for sim in simulators:
-    scenario = task_cls.scenario.update(
-        simulator=sim,
-        num_envs=10,
-        headless=True,
-    )
-    env = task_cls(scenario=scenario)
-    
-    # Run evaluation
-    success_rate = evaluate_policy(env, policy)
-    results[sim] = success_rate
-    
-    env.close()
-
-print("Cross-simulator results:", results)
-```
diff --git a/docs/source/metasim/index.md b/docs/source/metasim/index.md
@@ -8,10 +8,10 @@
 
 MetaSim enables you to:
 
-- **Write Once, Run Anywhere**: Develop simulation code that works seamlessly across MuJoCo, Isaac Sim, SAPIEN, PyBullet, Genesis, and more
-- **Configure Declaratively**: Define robots, scenes, cameras, and physics parameters through a type-safe configuration system
-- **Scale Efficiently**: Run thousands of parallel environments with GPU acceleration
-- **Extend Easily**: Add new robots, tasks, and simulators through well-defined interfaces
+- **Write Once, Run Anywhere**: Develop simulation code that works across multiple simulators
+- **Configure Declaratively**: Define robots, scenes, cameras, and physics parameters through a configuration system
+- **Scale**: Run parallel environments
+- **Extend**: Add new robots, tasks, and simulators through defined interfaces
 
 ### Core Concepts
 
diff --git a/docs/source/roboverse_learn/index.md b/docs/source/roboverse_learn/index.md
@@ -27,29 +27,6 @@ Train policies through trial and error with PPO, TD3, SAC, and specialized algor
 
 ---
 
-## Algorithm Comparison
-
-### Imitation Learning
-
-| Algorithm | Input | Architecture | Best For |
-|-----------|-------|--------------|----------|
-| [Diffusion Policy](imitation_learning/diffusion_policy.md) | Image/State | DiT, UNet | Multi-modal actions |
-| [ACT](imitation_learning/ACT.md) | Image/State | Transformer | Bimanual tasks |
-| [OpenVLA](imitation_learning/openvla.md) | Image+Language | VLM | Language-conditioned |
-| [SmolVLA](imitation_learning/smolvla.md) | Image+Language | Compact VLM | Edge deployment |
-| [RDT](imitation_learning/rdt.md) | Image/State | Transformer | Dexterous manipulation |
-| [Octo](imitation_learning/octo.md) | Image | Transformer | Cross-embodiment |
-
-### Reinforcement Learning
-
-| Algorithm | Type | GPU Parallel | Best For |
-|-----------|------|--------------|----------|
-| [PPO](reinforcement_learning/ppo.md) | On-policy | Yes | Stable training |
-| [FastTD3](reinforcement_learning/fast_td3.md) | Off-policy | Yes | Sample efficiency |
-| [SAC](reinforcement_learning/sac.md) | Off-policy | Yes | Continuous control |
-| [TD3](reinforcement_learning/td3.md) | Off-policy | Yes | Deterministic policy |
-| [SkillBlender](reinforcement_learning/skillblender_rl.md) | Hierarchical | Yes | Complex skills |
-
 ---
 
 ## Quick Start
