Add pipeline tag, library metadata, and research paper link

README.md

@@ -1,15 +1,19 @@
 ---
-license: apache-2.0
-datasets:
-- HuggingFaceH4/ultrachat_200k
 base_model:
 - Qwen/Qwen2.5-32B-Instruct
+datasets:
+- HuggingFaceH4/ultrachat_200k
+license: apache-2.0
+pipeline_tag: text-generation
+library_name: vllm
 ---
+
 # Qwen2.5-32B-Instruct_EAGLE3_UltraChat
 
+This model is an EAGLE-3 drafter for [Qwen2.5-32B-Instruct](https://huggingface.co/Qwen/Qwen2.5-32B-Instruct), introduced in the paper [Fail Fast, Win Big: Rethinking the Drafting Strategy in Speculative Decoding via Diffusion LLMs](https://huggingface.co/papers/2512.20573).
+
 ### Introduction
-**Qwen2.5-32B-Instruct_EAGLE3_UltraChat** is trained based on the open-source Qwen2.5-32B-Instruct model using the [SpecForge](https://github.com/sgl-project/SpecForge) framework, 
-and can be used for the Eagle-3 speculative decoding algorithm to speed up the inference of large language models during the decoding stage.
+**Qwen2.5-32B-Instruct_EAGLE3_UltraChat** is trained based on the open-source Qwen2.5-32B-Instruct model using the [SpecForge](https://github.com/sgl-project/SpecForge) framework, and can be used for the Eagle-3 speculative decoding algorithm to speed up the inference of large language models during the decoding stage.
 
 
 ### Training Configuration
@@ -17,8 +21,8 @@ We adopted the default training hyperparameters in SpecForge and trained EAGLE-3
 
 This model checkpoint is obtained after five epochs of training ($\sim$260k training steps with bs=4). We find that even though further training improves training-time accuracy, they have a negligible impact on the end-to-end speedup of EAGLE-3.
 
-- Dataset: Utilized the UltraChat-200K dataset.
-- Training environment: The training was conducted on 4 NVIDIA H100 GPUs with 80 GB VRAM each, leveraging the DeepSpeed framework. Each training epoch took approximately 3.5 hours.
+- **Dataset**: Utilized the [UltraChat-200K](https://huggingface.co/datasets/HuggingFaceH4/ultrachat_200k) dataset.
+- **Training environment**: The training was conducted on 4 NVIDIA H100 GPUs with 80 GB VRAM each, leveraging the DeepSpeed framework. Each training epoch took approximately 3.5 hours.
 
 ### Model Inference Launch Command
 
@@ -27,7 +31,7 @@ To launch the EAGLE-3 algorithm service using vLLM, here is the instruction:
 vllm serve Qwen/Qwen2.5-32B-Instruct \
   --dtype auto -tp 2 --max_model_len 2048 \
   --gpu-memory-utilization 0.8 --port 30000 \
-  --speculative_config '{"model": "/PATH/TO/EAGLE/WEIGHTS", "draft_tensor_parallel_size": 1, "num_speculative_tokens": 5, "method": "eagle3"}'
+  --speculative_config '{"model": "ruipeterpan/Qwen2.5-32B-Instruct_EAGLE3_UltraChat", "draft_tensor_parallel_size": 1, "num_speculative_tokens": 5, "method": "eagle3"}'
 ```
 
 To launch vanilla decoding, our performance baseline, here is the instruction:
@@ -48,10 +52,19 @@ We run our evaluations on two NVIDIA A6000-48GB GPUs connected via PCIe 4.0 x16.
 | **Qwen2.5-7B-Instruct**  | 2.19x | 2.05x | 2.02x | 1.78x | 2.25x | **2.06x** |
 
 
-### Relevant Link
+### Citation
 
-Qwen2.5-32B-Instruct Open-source Weights: https://huggingface.co/Qwen/Qwen2.5-32B-Instruct
+```bibtex
+@article{pan2025failfast,
+  title={Fail Fast, Win Big: Rethinking the Drafting Strategy in Speculative Decoding via Diffusion LLMs},
+  author={Pan, Rui and Chen, Zhuofu and Liu, Hongyi and Krishnamurthy, Arvind and Netravali, Ravi},
+  journal={arXiv preprint arXiv:2512.20573},
+  year={2025}
+}
+```
 
-"Fail Fast, Win Big: Rethinking the Drafting Strategy in Speculative Decoding via Diffusion LLMs" [arXiv '25]: https://arxiv.org/pdf/2512.20573
+### Relevant Links
 
-Artifact of FailFast: https://github.com/ruipeterpan/failfast
+- **GitHub Repository**: [ruipeterpan/failfast](https://github.com/ruipeterpan/failfast)
+- **Paper**: [arXiv:2512.20573](https://arxiv.org/abs/2512.20573)
+- **Qwen2.5-32B-Instruct Open-source Weights**: [Qwen/Qwen2.5-32B-Instruct](https://huggingface.co/Qwen/Qwen2.5-32B-Instruct)