Improve dataset card for OceanGym: add paper abstract, code link, and update metadata & citation

This PR significantly enhances the dataset card for OceanGym by:

- Adding `task_categories: ['robotics']` to the metadata for better discoverability.
- Enriching the `tags` with `['robotics', 'benchmark', 'environment', 'underwater', 'multi-modal', 'mllm', 'large-language-models']` to provide more descriptive context.
- Updating the paper link to the official Hugging Face Papers page (`https://huggingface.co/papers/2509.26536`) and changing its label to "Paper".
- Adding a direct link to the GitHub repository (`https://github.com/OceanGPT/OceanGym`).
- Cleaning up the existing Google Drive link by removing the `?usp=sharing` suffix.
- Adding the Baidu Drive link from the GitHub README.
- Including the comprehensive paper abstract in the dataset card for a clearer overview.
- Updating the reference to the paper within the "Decision Task" section to point to the new Hugging Face paper link.
- Adding a new "Datasets" section, mirroring the GitHub README, to provide specific links and structure for accessing the evaluation data. The Table of Contents is also updated.
- Replacing the placeholder BibTeX citation with the complete and correct entry from the GitHub README.

These updates aim to provide a more complete, accurate, and user-friendly dataset card for `OceanGym`.

README.md

@@ -1,9 +1,18 @@
 ---
-license: mit
 language:
 - en
+license: mit
+task_categories:
+- robotics
 tags:
 - agent
+- robotics
+- benchmark
+- environment
+- underwater
+- multi-modal
+- mllm
+- large-language-models
 ---
 
 <h1 align="center"> 🌊 OceanGym 🦾 </h1>
@@ -11,15 +20,19 @@ tags:
 
 <p align="center">
   🌐 <a href="https://oceangpt.github.io/OceanGym" target="_blank">Home Page</a>
-  📄 <a href="https://arxiv.org/abs/123" target="_blank">ArXiv Paper</a>
+  📄 <a href="https://huggingface.co/papers/2509.26536" target="_blank">Paper</a>
+  💻 <a href="https://github.com/OceanGPT/OceanGym" target="_blank">Code</a>
   🤗 <a href="https://huggingface.co/datasets/zjunlp/OceanGym" target="_blank">Hugging Face</a>
-  ☁️ <a href="https://drive.google.com/drive/folders/1H7FTbtOCKTIEGp3R5RNsWvmxZ1oZxQih?usp=sharing" target="_blank">Google Drive</a>
+  ☁️ <a href="https://drive.google.com/drive/folders/1H7FTbtOCKTIEGp3R5RNsWvmxZ1oZxQih" target="_blank">Google Drive</a>
+  ☁️ <a href="https://pan.baidu.com/s/19c-BeIpAG1EjMjXZHCAqPA?pwd=sgjs" target="_blank">Baidu Drive</a>
 </p>
 
 <img src="asset/img/o1.png" align=center>
 
 **OceanGym** is a high-fidelity embodied underwater environment that simulates a realistic ocean setting with diverse scenes. As illustrated in figure, OceanGym establishes a robust benchmark for evaluating autonomous agents through a series of challenging tasks, encompassing various perception analyses and decision-making navigation. The platform facilitates these evaluations by supporting multi-modal perception and providing action spaces for continuous control.
 
+We introduce OceanGym, the first comprehensive benchmark for ocean underwater embodied agents, designed to advance AI in one of the most demanding real-world environments. Unlike terrestrial or aerial domains, underwater settings present extreme perceptual and decision-making challenges, including low visibility, dynamic ocean currents, making effective agent deployment exceptionally difficult. OceanGym encompasses eight realistic task domains and a unified agent framework driven by Multi-modal Large Language Models (MLLMs), which integrates perception, memory, and sequential decision-making. Agents are required to comprehend optical and sonar data, autonomously explore complex environments, and accomplish long-horizon objectives under these harsh conditions. Extensive experiments reveal substantial gaps between state-of-the-art MLLM-driven agents and human experts, highlighting the persistent difficulty of perception, planning, and adaptability in ocean underwater environments. By providing a high-fidelity, rigorously designed platform, OceanGym establishes a testbed for developing robust embodied AI and transferring these capabilities to real-world autonomous ocean underwater vehicles, marking a decisive step toward intelligent agents capable of operating in one of Earth's last unexplored frontiers. The code and data are available at this https URL .
+
 # 💐 Acknowledgement
 
 OceanGym environment is based on Unreal Engine (UE) 5.3.
@@ -63,6 +76,7 @@ Thanks for their great contributions!
 - [⏱️ Results](#️-results)
   - [Decision Task](#decision-task-1)
   - [Perception Task](#perception-task-1)
+- [📚 Datasets](#-datasets)
 - [🚩 Citation](#-citation)
 
 # 📺 Quick Start
@@ -292,11 +306,11 @@ C:\Users\Windows\AppData\Local\holoocean\2.0.0\worlds\Ocean
 
 > All commands are applicable to **Windows** only, because it requires full support from the `UE5 Engine`.
 
-The decision experiment can be run with reference to the [Quick Start](#️-quick-start).
+The decision experiment can be run with reference to the [Quick Start](#-quick-start).
 
 ## Target Object Locations
 
-We have provided eight tasks. For specific task descriptions, please refer to the paper.
+We have provided eight tasks. For specific task descriptions, please refer to the [paper](https://huggingface.co/papers/2509.26536).
 
 The following are the coordinates for each target object in the environment (in meters):
 
@@ -591,6 +605,8 @@ python perception/task/init_map_with_sonar.py \
 
 # ⏱️ Results
 
+**We provide the trajectory data of OceanGym’s various task evaluations at the [next section](#-datasets), enabling readers to analyze and reproduce the results.**
+
 ## Decision Task
 
 <img src="asset/img/t1.png" align=center>
@@ -605,14 +621,66 @@ python perception/task/init_map_with_sonar.py \
 - Values represent accuracy percentages.
 - Adding sonar means using both RGB and sonar images.
 
+# 📚 Datasets
+**The link to the dataset is as follows**\
+ ☁️ <a href="https://drive.google.com/drive/folders/1VhrvhvbWvnaS4EyeyaV1fmTQ6gPo8GCN?usp=drive_link" target="_blank">Google Drive</a>
+- Decision Task
+
+```python
+decision_dataset
+├── main
+│ ├── gpt4omini
+│ │ ├── task1
+│ │ │ ├── point1
+│ │ │ │ ├── llm_output_...log
+│ │ │ │ ├── memory_...json
+│ │ │ │ └── important_memory_...json
+│ │ │ └── ... (other data points like point2, point3...)
+│ │ └── ... (other tasks like task2, task3...)
+│ ├── gemini
+│ │ └── ... (structure is the same as gpt4omini)
+│ └── qwen
+│ └── ... (structure is the same as gpt4omini)
+│
+├── migration
+│ ├── gpt4o
+│ │ └── ... (structure is the same as above)
+│ └── qwen
+│ └── ... (structure is the same as above)
+│
+└── scale
+  ├── qwen
+  └── gpt4omini
+```
+
+
+- Perception Task
+
+```python
+perception_dataset
+├── data
+│ ├── highLight
+│ ├── highLightContext
+│ ├── lowLight
+│ ├── lowLightContext
+│
+└── result
+
+```
+
 # 🚩 Citation
 
 If this OceanGym paper or benchmark is helpful, please kindly cite as this:
 
 ```bibtex
-@inproceedings{xxx,
-  title={OceanGym: A Benchmark Environment for Underwater Embodied Agents},
-  ...
+@misc{xue2025oceangymbenchmarkenvironmentunderwater,
+      title={OceanGym: A Benchmark Environment for Underwater Embodied Agents}, 
+      author={Yida Xue and Mingjun Mao and Xiangyuan Ru and Yuqi Zhu and Baochang Ren and Shuofei Qiao and Mengru Wang and Shumin Deng and Xinyu An and Ningyu Zhang and Ying Chen and Huajun Chen},
+      year={2025},
+      eprint={2509.26536},
+      archivePrefix={arXiv},
+      primaryClass={cs.CL},
+      url={https://arxiv.org/abs/2509.26536}, 
 }
 ```