Add metadata and link to paper/code

Hi, I'm Niels from the Hugging Face community team.

This PR improves the model card for X-Coder-RL-Qwen3-8B by:
- Adding `library_name: transformers` to the metadata.
- Adding `pipeline_tag: text-generation` to categorize the model.
- Adding the Arxiv ID (`2601.06953`) to link the model to the paper [X-Coder: Advancing Competitive Programming with Fully Synthetic Tasks, Solutions, and Tests](https://huggingface.co/papers/2601.06953).
- Ensuring direct links to the GitHub repository are present.

These additions will enable the inference snippet button and improve the model's discoverability on the Hub.

Best regards,
Niels

README.md

@@ -1,23 +1,30 @@
 ---
-license: apache-2.0
 base_model:
-  - IIGroup/X-Coder-SFT-Qwen3-8B
+- IIGroup/X-Coder-SFT-Qwen3-8B
 datasets:
-  - IIGroup/X-Coder-RL-40k
+- IIGroup/X-Coder-RL-40k
 language:
-  - en
+- en
+license: apache-2.0
+library_name: transformers
+pipeline_tag: text-generation
 tags:
-  - code
-  - rl
-  - competitive-programming
+- code
+- rl
+- competitive-programming
+- arxiv:2601.06953
 ---
 
 # X-Coder-RL-Qwen3-8B
 
-X-Coder-RL-Qwen3-8B is a code reasoning foundation model trained with RLVR on fully synthetic rl data, achieving strong performance on competitive programming.
+X-Coder-RL-Qwen3-8B is a code reasoning foundation model trained with RLVR on fully synthetic reinforcement learning data, achieving strong performance on competitive programming. It was introduced in the paper [X-Coder: Advancing Competitive Programming with Fully Synthetic Tasks, Solutions, and Tests](https://huggingface.co/papers/2601.06953).
+
+The model leverages the **SynthSmith** pipeline, which generates diverse and challenging tasks, verified solutions, and tests to advance code reasoning without relying on real-world data.
 
 ## Model Description
 
+- **Paper**: [X-Coder: Advancing Competitive Programming with Fully Synthetic Tasks, Solutions, and Tests](https://arxiv.org/abs/2601.06953)
+- **Code**: [X-Coder GitHub repository](https://github.com/JieWu02/X-Coder)
 - **Base Model**: [IIGroup/X-Coder-SFT-Qwen3-8B](https://huggingface.co/IIGroup/X-Coder-SFT-Qwen3-8B)
 - **Training Method**: GRPO
 - **Training Data**: [IIGroup/X-Coder-RL-40k](https://huggingface.co/datasets/IIGroup/X-Coder-RL-40k)
@@ -25,11 +32,11 @@ X-Coder-RL-Qwen3-8B is a code reasoning foundation model trained with RLVR on fu
 
 ## Training
 
-This model was trained using the X-Coder RLVR recipe. For training details and code, please refer to the [X-Coder GitHub repository](https://github.com/JieWu02/X-Coder).
+This model was trained using the X-Coder RLVR recipe. For training details and environment setup, please refer to the [official GitHub repository](https://github.com/JieWu02/X-Coder).
 
 ## Performance
 
-**Average Performance on LiveCodeBench v5 & v6.**
+X-Coder-RL-Qwen3-8B achieves significant performance gains on competitive programming using fully synthetic data, as shown in results for LiveCodeBench v5 and v6.
 
 ![Results](results.png)
 
@@ -80,4 +87,4 @@ print(tokenizer.decode(outputs[0], skip_special_tokens=True))
 
 ## License
 
-This project is licensed under the Apache License 2.0.
+This project is licensed under the Apache License 2.0.