Update README

README.md

@@ -7,6 +7,14 @@ base_model:
 pipeline_tag: image-to-image
 ---
 
+> **DeepGen 1.0 Checkpoints**
+>
+> | Stage | Repository | Description |
+> | :--- | :--- | :--- |
+> | Pretrain | [deepgenteam/DeepGen-1.0-Pretrain](https://huggingface.co/deepgenteam/DeepGen-1.0-Pretrain) | Alignment pre-training checkpoint |
+> | SFT | [deepgenteam/DeepGen-1.0-SFT](https://huggingface.co/deepgenteam/DeepGen-1.0-SFT) | Supervised fine-tuning checkpoint |
+> | **RL** | **[deepgenteam/DeepGen-1.0](https://huggingface.co/deepgenteam/DeepGen-1.0)** | Reinforcement learning checkpoint (MR-GDPO) *(this repo)* |
+
 # 💡 DeepGen 1.0: A Lightweight Unified Multimodal Model for Advancing Image Generation and Editing
 <p align="left">
   <a href="http://arxiv.org/abs/2602.12205">
@@ -29,14 +37,14 @@ pipeline_tag: image-to-image
   </a>
 </p>
 DeepGen 1.0 is a lightweight unified multimodal model with only 5B parameters (3B VLM + 2B DiT). It integrates five core capabilities—general image generation, general image editing, reasoning image generation, reasoning image editing, and text rendering—within a single model. Across multiple authoritative benchmarks, DeepGen 1.0 is competitive with competitive with or surpassing the state-of-the-art unified multimodal models that are 3× to 16× larger, achieving comprehensive performance, demonstrating that massive scaling is not the sole path to high-performance multimodal generation.
-<p align="left"><img src="bubble_chart.png" width="80%"></p>
+<p align="left"><img src="docs/bubble_chart.png" width="80%"></p>
 
 ## 🧠 Method
 Our core observation is that a lightweight model, when empowered by synergistic architecture design and data-centric training strategies, can achieve comprehensive capabilities competitive with or even surpassing much larger counterparts.
 To overcome the limitations of lightweight models in semantic understanding and fine-grained control, we introduce **Stacked Channel Bridging (SCB)**, a deep alignment framework that extracts hierarchical features from multiple VLM layers and fuses them with learnable ``think tokens'' to provide the generative backbone with structured, reasoning-rich guidance. 
 We further design a data-centric training strategy spanning three progressive stages: (1) **Alignment Pre-training** on large-scale image-text pairs and editing triplets to synchronize VLM and DiT representations, (2) **Joint Supervised Fine-tuning** on a high-quality mixture of generation, editing, and reasoning tasks to foster omni-capabilities, and (3) **Reinforcement Learning with MR-GRPO**, which leverages a mixture of reward functions and supervision signals, resulting in substantial gains in generation quality and alignment with human preferences, while maintaining stable training progress and avoiding visual artifacts.
 
-<p align="left"><img src="arch.png" width="80%"></p>
+<p align="left"><img src="docs/arch.png" width="80%"></p>
 
 ## 📊 Benchmarks
 
@@ -91,16 +99,18 @@ We further design a data-centric training strategy spanning three progressive st
 | **DeepGen 1.0 (RL)** | **3B + 2B** | 10.8 🥉 | 75.7 🥈 |
 
 ## 🎨 Quantitative results
-<p align="left"><img src="teaser.png" width="80%"></p>
+<p align="left"><img src="docs/teaser.png" width="80%"></p>
 
 ## 🛠️ Usage
 
-### Merge ZIP Files
-To use the DeepGen checkpoints, please merge the sharded model files first. We release Pre-traning, Supervised Fine-Tuning and Reinforcement Learning checkpoints.
+### Download Checkpoint
+This repository contains the **Reinforcement Learning (MR-GDPO)** checkpoint — the final release model.
 
 ```bash
-# Merge zip
-cat DeepGen_CKPT.zip.part-* > DeepGen_CKPT.zip
-# Unzip DeepGen checkpoints 
-unzip DeepGen_CKPT.zip
-```
+# Using hf CLI
+hf download deepgenteam/DeepGen-1.0 model.pt --local-dir .
+
+# Or using Python
+from huggingface_hub import hf_hub_download
+hf_hub_download("deepgenteam/DeepGen-1.0", "model.pt", local_dir=".")
+```