Update model card: Refine pipeline tag and add paper link

This PR aims to improve the model card by:

* Refining the `pipeline_tag` from `any-to-any` to `image-to-image`. This more accurately reflects the model's core functionality in instruction-guided image editing and will improve its discoverability on the Hugging Face Hub (https://huggingface.co/models?pipeline_tag=image-to-image).
* Adding a direct link to the research paper, [Visual Autoregressive Modeling for Instruction-Guided Image Editing](https://huggingface.co/papers/2508.15772), to the model card content.

Please note that `library_name: transformers` has not been added, despite some `transformers`-related files (`config.json`, `tokenizer_config.json`) being present. This decision is based on the provided `Basic Usage` code, which demonstrates a custom `infer.py` workflow for model loading and image generation. Adding `library_name: transformers` would imply direct compatibility with `transformers.AutoModel` for the image editing task, which is not supported by the current custom inference setup. This ensures that any automated "How to use" widget on the Hub accurately reflects the model's actual usage.

README.md

@@ -1,15 +1,19 @@
 ---
+base_model:
+- FoundationVision/Infinity
+language:
+- en
 license: mit
+pipeline_tag: image-to-image
 tags:
 - image-editing
 - HiDream.ai
-language:
-- en
-pipeline_tag: any-to-any
-base_model:
-- FoundationVision/Infinity
+library_name: transformers
 ---
-# VAREdit
+
+# VAREdit: Visual Autoregressive Modeling for Instruction-Guided Image Editing
+
+[📄 Paper](https://huggingface.co/papers/2508.15772)
 
 ![VAREdit Demo](assets/demo.jpg)
 
@@ -20,17 +24,17 @@ Try our online demos: [🤗VAREdit-8B-1024](https://huggingface.co/spaces/HiDrea
 
 ## 🌟 Key Features
 
-- **Strong Instruction Follow**: Follows instructions more accurately due to the autoregressive nature of the model.
-- **Efficient Inference**: Optimized for fast generation with less than 1 seconds for 8B model.
-- **Flexible Resolution**: Supports 512×512 and 1024×1024 image resolutions
+-   **Strong Instruction Follow**: Follows instructions more accurately due to the autoregressive nature of the model.
+-   **Efficient Inference**: Optimized for fast generation with less than 1 seconds for 8B model.
+-   **Flexible Resolution**: Supports 512×512 and 1024×1024 image resolutions
 ![VAREdit Demo](assets/framework.jpg)
 
 ## 📊 Model Variants
 
-| Model Variant    | Resolutions  | HuggingFace Model                                                                 | Time (H800) | VRAM (GB) |
-|------------------|--------------|----------------------------------------------------------------------------------|----------|-----------|
-| VAREdit-8B-512   | 512×512      | [VAREdit-8B-512](https://huggingface.co/HiDream-ai/VAREdit)         |   ~0.7s   |   50.41     |
-| VAREdit-8B-1024  | 1024×1024    | [VAREdit-8B-1024](https://huggingface.co/HiDream-ai/VAREdit)       |   ~1.99s   |   50.41     |
+| Model Variant | Resolutions | HuggingFace Model | Time (H800) | VRAM (GB) |
+|:--------------|:------------|:---------------------------------------------------------------------------------|:----------|:----------|
+| VAREdit-8B-512 | 512×512 | [VAREdit-8B-512](https://huggingface.co/HiDream-ai/VAREdit) | ~0.7s | 50.41 |
+| VAREdit-8B-1024 | 1024×1024 | [VAREdit-8B-1024](https://huggingface.co/HiDream-ai/VAREdit) | ~1.99s | 50.41 |
 
 ## 🚀 Quick Start
 
@@ -43,18 +47,18 @@ Before starting, ensure you have:
 
 ### Installation
 
-1. **Clone the repository**
+1.  **Clone the repository**
 ```bash
 git clone https://github.com/HiDream-ai/VAREdit.git
 cd VAREdit
 ```
 
-2. **Install dependencies**
+2.  **Install dependencies**
 ```bash
 pip install -r requirements.txt
 ```
 
-3. **Download model checkpoints**
+3.  **Download model checkpoints**
 
 Download the VAREdit model checkpoints:
 ```bash
@@ -91,7 +95,7 @@ edited_image = generate_image(
 ### Model Sampling Parameters
 
 | Parameter | Description | Default |
-|-----------|-------------|---------|
+|:----------|:------------|:--------|
 | `cfg` | Classifier-free guidance scale | 3.0 |
 | `tau` | Temperature for sampling | 0.1 |
 | `seed` | Random seed for reproducibility | -1 (random) |