Add library_name, pipeline_tag and arxiv metadata

This PR improves the model card for ASTRA-14B-Thinking-v1 by adding relevant metadata to the YAML header. Specifically, it adds:
- `library_name: transformers`: To ensure the model is correctly identified and to enable the "Use in Transformers" button.
- `pipeline_tag: text-generation`: To categorize the model correctly on the Hub.
- `arxiv: 2601.21558`: To link the model repository to its research paper.
- Relevant task tags like `agent` and `tool-use`.

The description has also been slightly updated to reference the research paper.

README.md

@@ -1,7 +1,14 @@
 ---
-license: apache-2.0
 base_model:
 - Qwen/Qwen3-14B
+license: apache-2.0
+library_name: transformers
+pipeline_tag: text-generation
+arxiv: 2601.21558
+tags:
+- tool-use
+- agent
+- reinforcement-learning
 ---
 
 [![GitHub](https://img.shields.io/badge/GitHub-Astra-blue?logo=github)](https://github.com/LianjiaTech/astra)
@@ -12,7 +19,9 @@ base_model:
 
 ## Model Description
 
-The **ASTRA-14B-Thinking-v1** model is derived from [**Qwen3-14B**](https://huggingface.co/Qwen/Qwen3-14B) and specifically optimized for multi-step, tool-augmented tasks, with enhanced agentic capabilities in complex tool use and structured reasoning.
+The **ASTRA-14B-Thinking-v1** model is derived from [**Qwen3-14B**](https://huggingface.co/Qwen/Qwen3-14B) and specifically optimized for multi-step, tool-augmented tasks, with enhanced agentic capabilities in complex tool use and structured reasoning. 
+
+This model was introduced in the paper [ASTRA: Automated Synthesis of agentic Trajectories and Reinforcement Arenas](https://huggingface.co/papers/2601.21558). ASTRA is a fully automated end-to-end framework for training tool-augmented language model agents via scalable data synthesis and verifiable reinforcement learning.
 
 We also provide a **32B** variant [**ASTRA-32B-Thinking-v1**](https://huggingface.co/Emperorizzis/ASTRAL-32B-Thinking-v1).