Improve dataset card: add task category, paper link, and usage instructions

README.md

@@ -1,78 +1,60 @@
 ---
 license: apache-2.0
+task_categories:
+- image-segmentation
 pipeline_tag: any-to-any
 library_name: bagel-mot
 tags:
-  - sgt
-  - semantic-generative-tuning
-  - unified-multimodal
-  - image-segmentation
-  - visual-understanding
-  - visual-generation
+- sgt
+- semantic-generative-tuning
+- unified-multimodal
+- visual-understanding
+- visual-generation
 ---
 
-# SGT: Semantic Generative Tuning for Unified Multimodal Models
+# SAM-SGT: Segmentation Training Data for Semantic Generative Tuning
 
-This repository hosts checkpoints fine-tuned with **Semantic Generative Tuning (SGT)** — a training
-paradigm that couples visual *understanding* and *generation* in Unified Multimodal Models (UMMs)
-by using **image segmentation as a generative proxy**.
+This repository contains the **SAM-SGT** dataset, a collection of approximately 190k segmentation samples derived from SAM. It was introduced in the paper [Semantic Generative Tuning for Unified Multimodal Models](https://huggingface.co/papers/2605.18714).
 
-> Unified multimodal models typically optimize understanding and generation with *misaligned*
-> objectives (sparse text tokens vs. dense pixel targets), which isolates the two capabilities.
-> SGT introduces segmentation — a **high-level semantic task** — as a unified generative objective
-> that aligns the two branches, improves feature linear separability, and optimizes visual-textual
-> attention allocation.
+[**Project Page**](https://song2yu.github.io/SGT/) | [**Paper**](https://huggingface.co/papers/2605.18714) | [**Code**](https://github.com/song2yu/SGT)
 
-## 🧠 Method Overview
+## Dataset Description
 
-SGT reformulates classical visual tasks as generative proxies and establishes a **hierarchical
-taxonomy** (low-/mid-/high-level). Extensive experiments show that **high-level semantic tasks
-(e.g. image segmentation) are the optimal proxy**, outperforming depth, edge, reconstruction and
-MAE/inpainting for synergizing understanding and generation.
+SGT (Semantic Generative Tuning) is a training paradigm that couples visual understanding and generation in Unified Multimodal Models (UMMs) by using image segmentation as a generative proxy. This dataset provides high-level semantic supervision used to align multimodal representation spaces.
 
-Key findings:
+- **Content**: ~190,000 segmentation samples sourced from the Segment Anything (SAM) dataset.
+- **Format**: Tar-sharded.
+- **Role**: Serves as a high-level semantic proxy task to enhance vision-centric perception and generative layout fidelity.
 
-1. **High-level > low-level**: segmentation gives larger gains in visual understanding
-   than depth / edge / pixel reconstruction.
-2. **Perception, not reasoning**: visual supervision mainly strengthens perception
-   (spatial, hallucination, vision-centric, general VQA), rather than abstract reasoning (e.g. math, chart)
-3. **Architecture-agnostic**: the gains hold for both **BAGEL** and **OmniGen2**.
+### Training Data Distribution
+| Data Source | Samples |
+|-------------|---------|
+| **SGT — Segmentation (SAM)** | **190k** |
+| General VQA | 180k |
+| Doc / Chart / Screen | 103k |
+| Math / Reasoning | 101k |
+| Language | 72k |
+| General OCR | 45k |
+| **Total** | **~691k** |
 
-## 📦 Released Artifacts
+## Usage
 
-| Repo | Type | Base Model | Content |
-|---|---|---|---|
-| [`Two-hot/SGT-BAGEL`](https://huggingface.co/Two-hot/SGT-BAGEL)   | model   | BAGEL-7B-MoT   | SGT fine-tuned BAGEL checkpoint |
-| [`Two-hot/SGT-Gen2`](https://huggingface.co/Two-hot/SGT-Gen2)     | model   | OmniGen2       | SGT fine-tuned OmniGen2 checkpoint (transformer/ only) |
-| [`Two-hot/SAM-SGT`](https://huggingface.co/datasets/Two-hot/SAM-SGT) | dataset | —            | Segmentation training data (tar-sharded) used by SGT |
+You can download the dataset using the provided script from the official repository:
 
-### Use the SAM-SGT dataset
+```bash
+# download sam subset || Chinese users can use --use-mirror
+python download_sam.py --target-dir ./data/SAM-SGT --use-mirror
+```
 
-See [`Two-hot/SAM-SGT`](https://huggingface.co/datasets/Two-hot/SAM-SGT) for the data
-layout and the extraction instructions.
+## Citation
 
-## 📊 Highlights
-
-- **+6.02%** average gain over BAGEL on the **CV-Bench** evaluation.
-- Consistent improvements in **spatial reasoning**, **hallucination resistance**, **vision-centric**, and **general VQA**.
-- Generation: gains across **GenEval** dimensions (Position / Color etc.).
-- Verified on two representative UMM architectures (**BAGEL**, **OmniGen2**).
-
-## 📝 License
-
-Apache-2.0. Base models remain under their original licenses:
-BAGEL (Apache-2.0, based on Qwen2.5-7B + SigLIP + FLUX VAE) and
-OmniGen2 (based on Qwen2.5-VL + diffusion transformer).
-
-## ✍️ Citation
-
-If you find this work useful, please cite our paper:
+If you find this work useful, please cite:
 
 ```bibtex
-@article{sgt2026,
-  title   = {Semantic Generative Tuning for Unified Multimodal Models},
-  author  = {Songsong Yu, Yuxin Chen, Ying Shan, and Yanwei Li},
-  journal = {arxiv},
-  year    = {2026}
+@article{yu2026sgt,
+  title     = {Semantic Generative Tuning for Unified Multimodal Models},
+  author    = {Yu, Songsong and Chen, Yuxin and Shan, Ying and Li, Yanwei},
+  journal   = {arXiv preprint arXiv:2605.18714},
+  year      = {2026},
 }
 ```