Add pipeline tag and repository links

README.md

@@ -1,19 +1,26 @@
 ---
-license: mit
 base_model:
 - openai-community/gpt2
+license: mit
+pipeline_tag: text-generation
+library_name: transformers
 ---
+
 # COCONUT Model
 
 <div align="center">
 
 [![HuggingFace](https://img.shields.io/badge/🤗%20HuggingFace-Model-fcc21b?style=for-the-badge&logo=huggingface&logoColor=white)](https://huggingface.co/ModalityDance/latent-tts-coconut)
+[![Paper](https://img.shields.io/badge/Paper-arXiv-b31b1b?style=for-the-badge&logo=arxiv)](https://arxiv.org/abs/2510.07745)
+[![GitHub](https://img.shields.io/badge/GitHub-Code-blue?style=for-the-badge&logo=github)](https://github.com/ModalityDance/LatentTTS)
 
 </div>
 
 ## Overview
 
-**COCONUT** (Chain of Continuous Thought) is a latent reasoning model based on GPT-2 that enables continuous thought generation in latent space. This model is part of the [Parallel Test-Time Scaling for Latent Reasoning Models](https://arxiv.org/abs/2510.07745) framework.
+**COCONUT** (Chain of Continuous Thought) is a latent reasoning model based on GPT-2 that enables continuous thought generation in latent space. This model is part of the research presented in the paper [Parallel Test-Time Scaling for Latent Reasoning Models](https://huggingface.co/papers/2510.07745).
+
+Official Code: [https://github.com/ModalityDance/LatentTTS](https://github.com/ModalityDance/LatentTTS)
 
 ## Model Details
 
@@ -40,6 +47,8 @@ huggingface-cli download ModalityDance/latent-tts-coconut --local-dir checkpoint
 
 ### Basic Usage
 
+Note: Inference requires the `src` directory and custom implementation files from the [official GitHub repository](https://github.com/ModalityDance/LatentTTS).
+
 ```python
 from transformers import AutoTokenizer
 from src.generation_mixin import LatentGenerationMixin, LatentGenerationConfig
@@ -71,7 +80,8 @@ model = LatentCOCONUT.from_pretrained(
 )
 
 # Prepare input (note: newline before <|start-latent|>)
-question = "What is 2 + 2?\n<|start-latent|>"
+question = "What is 2 + 2?
+<|start-latent|>"
 inputs = tokenizer(question, return_tensors="pt").to(model.device)
 
 # Configure generation
@@ -104,9 +114,12 @@ The model fully supports batch processing:
 ```python
 # Prepare batch inputs
 questions = [
-    "What is 2 + 2?\n<|start-latent|>",
-    "What is 5 * 3?\n<|start-latent|>",
-    "What is 10 - 4?\n<|start-latent|>",
+    "What is 2 + 2?
+<|start-latent|>",
+    "What is 5 * 3?
+<|start-latent|>",
+    "What is 10 - 4?
+<|start-latent|>",
 ]
 inputs = tokenizer(questions, return_tensors="pt", padding=True).to(model.device)
 
@@ -169,7 +182,7 @@ print(f"Answer: {answer}")
 
 ## Evaluation
 
-Run evaluation using the provided scripts:
+Run evaluation using the provided scripts in the official repository:
 
 ```bash
 # For COCONUT (GPT-2 based models)