Improve model card: Add license metadata and paper abstract

This PR enhances the model card for `TeleEmbodied/humanoidgen_model` by:

* Adding the `license: cc-by-nc-4.0` to the metadata section, as explicitly stated in the GitHub repository.
* Integrating the paper's abstract into the model card content for a more informative overview.
* Adjusting the title formatting for consistency with Markdown best practices.
* Adding a brief introductory sentence linking the repository content to the paper.

All existing links to the paper, code, homepage, dataset, and other models are correctly maintained.

Please review and merge this PR if it looks good.

README.md

@@ -1,17 +1,20 @@
 ---
+pipeline_tag: robotics
 task_categories:
 - robotics
-pipeline_tag: robotics
+license: cc-by-nc-4.0
 ---
 
-<div align="center">
-<h1> HumanoidGen-Model </h1>
-</div>
+# HumanoidGen-Model
+
 <h5 align="center">
     <a href="https://openhumanoidgen.github.io/"> 🔥 Homepage</a> | <a href="https://huggingface.co/datasets/TeleEmbodied/humanoidgen_dataset/tree/main/task_datasets">⛁ Dataset</a> | <a href="https://huggingface.co/TeleEmbodied/humanoidgen_model/tree/main">🤗 Models</a> | <a href="https://arxiv.org/abs/2507.00833">📑 Paper</a> | <a href="https://github.com/TeleHuman/HumanoidGen">💻 Code</a>
 </h5>
 
-This repository contains models trained using data generated by the HumanoidGen framework.
+This repository contains models trained using data generated by the HumanoidGen framework, as presented in the paper [HumanoidGen: Data Generation for Bimanual Dexterous Manipulation via LLM Reasoning](https://arxiv.org/abs/2507.00833).
+
+## Abstract
+For robotic manipulation, existing robotics datasets and simulation benchmarks predominantly cater to robot-arm platforms. However, for humanoid robots equipped with dual arms and dexterous hands, simulation tasks and high-quality demonstrations are notably lacking. Bimanual dexterous manipulation is inherently more complex, as it requires coordinated arm movements and hand operations, making autonomous data collection challenging. This paper presents HumanoidGen, an automated task creation and demonstration collection framework that leverages atomic dexterous operations and LLM reasoning to generate relational constraints. Specifically, we provide spatial annotations for both assets and dexterous hands based on the atomic operations, and perform an LLM planner to generate a chain of actionable spatial constraints for arm movements based on object affordances and scenes. To further improve planning ability, we employ a variant of Monte Carlo tree search to enhance LLM reasoning for long-horizon tasks and insufficient annotation. In experiments, we create a novel benchmark with augmented scenarios to evaluate the quality of the collected data. The results show that the performance of the 2D and 3D diffusion policies can scale with the generated dataset.
 
 # HumanoidGen Framework
 
@@ -28,5 +31,4 @@ If you find our work helpful, please cite us:
   journal={arXiv preprint arXiv:2507.00833},
   year={2025}
 }
-```
-
+```