Improve model card: add paper link, repository link, and model description

README.md

@@ -1,34 +1,51 @@
 ---
+base_model: Qwen/Qwen2.5-3B-Instruct
 library_name: transformers
 license: other
-base_model: Qwen/Qwen2.5-3B-Instruct
+pipeline_tag: text-generation
 tags:
 - llama-factory
 - full
 - generated_from_trainer
+- agent
+- tool-use
+- long-horizon-planning
 model-index:
 - name: TravelPlanner_Instruct_3B_1128_1K_4Epoch
   results: []
 ---
 
-<!-- This model card has been generated automatically according to the information the Trainer had access to. You
-should probably proofread and complete it, then remove this comment. -->
-
-# TravelPlanner_Instruct_3B_1128_1K_4Epoch
+# TravelPlanner_Instruct_3B_1128_1K_4Epoch (Agent-STAR-SFT-3B)
 
-This model is a fine-tuned version of [Qwen/Qwen2.5-3B-Instruct](https://huggingface.co/Qwen/Qwen2.5-3B-Instruct) on the travelplanner_1108_5k_filter_1128_1k dataset.
+This model is a fine-tuned version of [Qwen/Qwen2.5-3B-Instruct](https://huggingface.co/Qwen/Qwen2.5-3B-Instruct) and is part of the **Agent-STAR** project. It corresponds to the **STAR-SFT-3B** stage described in the paper: [Demystifying Reinforcement Learning for Long-Horizon Tool-Using Agents: A Comprehensive Recipe](https://huggingface.co/papers/2603.21972).
 
 ## Model description
 
-More information needed
+**Agent-STAR** is a systematic framework designed to evolve Large Language Models into autonomous agents capable of long-horizon planning and tool orchestration. This specific checkpoint represents the Supervised Fine-Tuning (SFT) stage of the **STAR [Data Synthesis → SFT → RL]** pipeline.
+
+It was fine-tuned from the Qwen2.5-Instruct base using **1K successful trajectories** generated by stronger models (DeepSeek-V3.2-Exp-Thinking) on synthetic travel planning queries. The model is optimized to satisfy multifaceted constraints (commonsense and hard constraints) in multi-turn environments like the TravelPlanner benchmark.
+
+- **Paper:** [Demystifying Reinforcement Learning for Long-Horizon Tool-Using Agents: A Comprehensive Recipe](https://huggingface.co/papers/2603.21972)
+- **Repository:** [https://github.com/WxxShirley/Agent-STAR](https://github.com/WxxShirley/Agent-STAR)
+- **Point of Contact:** [Xixi Wu](mailto:xxwu@se.cuhk.edu.hk)
 
 ## Intended uses & limitations
 
-More information needed
+The model is intended for research on agentic LLMs, particularly those requiring tool use and multi-step planning. While it achieves high performance on travel planning tasks, users should be aware of typical LLM limitations regarding factual consistency and potential biases.
 
-## Training and evaluation data
+## Usage
 
-More information needed
+To run ReAct inference with this model, you can use the following command structure from the official repository:
+
+```bash
+python3 -u main.py \
+  --model xxwu/TravelPlanner_Instruct_3B_1128_1K_4Epoch \
+  --save_suffix your_experiment_name \
+  --max_workers 20 \
+  --split validation \
+  --max_context 32768 \
+  --max_turns 60 
+```
 
 ## Training procedure
 
@@ -43,19 +60,30 @@ The following hyperparameters were used during training:
 - num_devices: 8
 - gradient_accumulation_steps: 4
 - total_train_batch_size: 32
-- total_eval_batch_size: 64
-- optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
+- optimizer: AdamW (fused)
 - lr_scheduler_type: cosine
 - lr_scheduler_warmup_ratio: 0.1
 - num_epochs: 4.0
 
-### Training results
-
-
-
 ### Framework versions
 
 - Transformers 4.57.1
 - Pytorch 2.9.0+cu128
 - Datasets 4.0.0
 - Tokenizers 0.22.1
+
+## Citation
+
+If you find Agent-STAR helpful to your work, please cite:
+
+```bibtex
+@misc{wu2026agentstar,
+      title={Demystifying Reinforcement Learning for Long-Horizon Tool-Using Agents: A Comprehensive Recipe}, 
+      author={Xixi Wu and Qianguo Sun and Ruiyang Zhang and Chao Song and Junlong Wu and Yiyan Qi and Hong Cheng},
+      year={2026},
+      eprint={2603.21972},
+      archivePrefix={arXiv},
+      primaryClass={cs.LG},
+      url={https://arxiv.org/abs/2603.21972}, 
+}
+```