Update README.md

README.md

@@ -14,7 +14,7 @@ tags:
 
 ## Description:
 
-Arena-Dexsuite-Lift-RL-Newton-Task is a model trained for the object lifting task in one of IsaacLab-Arena Newton example workflows. The Kuka Allergo robot learns to pick up a primitive shape on the table and lift it to target position through reinforcement learning. The task uses a command-based goal specification, where the RL agent learns to reach sampled target poses.<br><br>
+Arena-Dexsuite-Lift-RL-Newton-Task is a model trained for the object lifting task in one of IsaacLab-Arena Newton example workflows. The Kuka Allegro robot learns to pick up primitive shapes (e.g., cube-shaped objects) on a table and lift them to target positions through reinforcement learning. The task uses a command-based goal specification, where the RL agent learns to reach sampled target poses.<br><br>
 This model is ready for non-commercial use only.
 
 ## Governing Terms
@@ -25,8 +25,8 @@ Global
 
 ### Release Date:
 
-GitHub: 03/30/2026 via [isaac-sim/IsaacLab-Arena](https://github.com/isaac-sim/IsaacLab-Arena/)
-Hugging Face: 03/30/2026 via [nvidia/Arena-Dexsuite-Lift-RL-Newton-Task](https://huggingface.co/nvidia/Arena-Dexsuite-Lift-RL-Newton-Task)
+GitHub: 04/09/2026 via [isaac-sim/IsaacLab-Arena](https://github.com/isaac-sim/IsaacLab-Arena/)
+Hugging Face: 04/09/2026 via [nvidia/Arena-Dexsuite-Lift-RL-Newton-Task](https://huggingface.co/nvidia/Arena-Dexsuite-Lift-RL-Newton-Task)
 
 ### Use Case:
 Researchers, Academics, Open-Source Community: AI-driven robotics research and algorithm development.
@@ -34,14 +34,14 @@ Developers: Integrate and customize AI for various robotic applications.
 Startups & Companies: Accelerate robotics development and reduce training costs.
 
 ## Reference(s):
-RSL‑RL: A Learning Library for Robotics Research, arXiv preprint (2025), arXiv:2509.10771
+[RSL‑RL: A Learning Library for Robotics Research, arXiv preprint (2025)](https://arxiv.org/abs/2509.10771)
 
 ## Model Architecture:
 
 **Architecture Type:**  Multi-layer Perceptron
 
 **Network Architecture:** 
-The model is trained to move the Kuka end effector lifting the object to commanded target position. The model itself is a MLP network trained with on-policy PPO using RSL-RL PPO framework using Newton physics engine.
+MLP network trained with on-policy PPO using RSL-RL PPO framework.
 
 **Number of Model Parameters:**
 105,871
@@ -56,43 +56,54 @@ The model is trained to move the Kuka end effector lifting the object to command
 **Input Parameters:**
 * Robot State: One-Dimensional (1D) - Floating number vector<br>
 
+**Other Properties Related to Input:** Continuous-value vectors correspond to different motor controls on a robot, which depends on Degrees of Freedom of the robot embodiment.
+
 ## Output:
 **Output Type(s):** End-effector target pose<br>
 **Output Format:** Continuous-value vectors<br>
 **Output Parameters:** [One-Dimensional (1D)] <br>
+**Other Properties Related to Output:** Continuous-value vectors correspond to different motor controls on a robot, which depends on Degrees of Freedom of the robot embodiment.
+
+Our AI models are designed and/or optimized to run on NVIDIA GPU-accelerated systems. By leveraging NVIDIA's hardware (e.g. GPU cores) and software frameworks (e.g., CUDA libraries), the model achieves faster training and inference times compared to CPU-only solutions. <br>
 
-Our AI models are designed and/or optimized to run on NVIDIA GPU-accelerated systems. By leveraging NVIDIA’s hardware (e.g. GPU cores) and software frameworks (e.g., CUDA libraries), the model achieves faster training and inference times compared to CPU-only solutions. <br>
+The integration of foundation and fine-tuned models into AI systems requires additional testing using use-case-specific data to ensure safe and effective deployment. Following the V-model methodology, iterative testing and validation at both unit and system levels are essential to mitigate risks, meet technical and functional requirements, and ensure compliance with safety and ethical standards before deployment.
 
 ## Model Version(s):
 Version 0.1.
 
 # Training, Testing, and Evaluation:
-**Training & Testing:** The model was trained completely in simulation (IsaacLab-Arena) using a on-policy Reinforcement Learning method.
 
-**Data Collection Method:** N/A (Reinforcement Learning Only)
+## Training & Testing
+
+The model was trained completely in simulation (IsaacLab-Arena) using a on-policy Reinforcement Learning method.
+
+**Data Modality:** Other: Robot simulation
+
+**Training Data Size:** Not Applicable
+
+**Data Collection Method:** Simulated Kuka Allegro - Reinforcement Learning
+
+**Labeling Method:** Not Applicable
+
+**Properties:** The model was trained completely in simulation (IsaacLab-Arena) using a on-policy Reinforcement Learning method. In this case, the data was collected for a couple of episodes, then the algorithm ran gradient descent on the collected data to update the model weights. Once this step was completed, the collected data was discarded and a new batch of data was collected.
+
+The simulation environment contained the robot itself, a primitive sitting on a table, and a ground plane.
 
-**Labeling Method:** N/A
+## Evaluation
 
-**Properties:** There was no data collected or used for training. The model was trained completely in simulation (IsaacLab-Arena) using a on-policy Reinforcement Learning method. In this case, the data will be collected for a couple of episodes, then the algorithm will run gradient descent on these collected data to update the model weights. Once this step is done, the collected data will be discarded and a new batch of data will be collected and so on.
+**Data Modality:** Other: Robot simulation
 
-As for the simulation environment, it contains the robot itself, a primitive sitting on a table, and a ground plane.
+**Data Collection Method:** Simulated Kuka Allegro - Reinforcement Learning
 
-All the training, testing, and evaluation was done in the same simulation environment. No data has been collected in any of these procedures.
+**Labeling Method:** Not Applicable
 
+**Properties:** The evaluation was performed in simulation using the IsaacLab-Arena (Newton physics engine). The evaluation data consists of dynamically generated episodes of lifting tasks.
 
 ## Inference:
 **Acceleration Engine(s):** PyTorch
 
 **Test Hardwares** 
-All of the below:
-* NVIDIA Ampere
-* NVIDIA Blackwell
-* NVIDIA Jetson
-* NVIDIA Hopper
-* NVIDIA Lovelace
-
-**Supported Operating System(s):**
-* Linux
+* NVIDIA RTX 6000 Ada
 
 ## Software Integration
 
@@ -108,8 +119,6 @@ All of the below:
 
 **Preferred/Supported Operating System(s):** Linux
 
-The integration of foundation and fine-tuned models into AI systems requires additional testing using use-case-specific data to ensure safe and effective deployment. Following the V-model methodology, iterative testing and validation at both unit and system levels are essential to mitigate risks, meet technical and functional requirements, and ensure compliance with safety and ethical standards before deployment.
-
 ## Ethical Considerations:
 NVIDIA believes Trustworthy AI is a shared responsibility and we have established policies and practices to enable development for a wide array of AI applications.  When downloaded or used in accordance with our terms of service, developers should work with their internal model team to ensure this model meets requirements for the relevant industry and use case and addresses unforeseen product misuse.