Update README.md

README.md

@@ -1,3 +1,231 @@
 ---
 license: apache-2.0
+size_categories:
+- 1K<n<10K
 ---
+<div align="center">
+
+# 🎯 EPIC-Bench: Can VLMs Perceive the Embodied Real-World?
+
+[![arXiv](https://img.shields.io/badge/arXiv-coming_soon-b31b1b.svg)](https://epic-bench.github.io/EPIC-Bench/)
+[![Project Page](https://img.shields.io/badge/Project-Page-blue)](https://epic-bench.github.io/EPIC-Bench/)
+[![Dataset](https://img.shields.io/badge/🤗-Dataset-yellow)](https://epic-bench.github.io/EPIC-Bench/)
+[![Evaluation Toolkit](https://img.shields.io/badge/⚙️-Evaluation_Toolkit-6366f1.svg)](#-epic-bench-evaluation-toolkit)
+[![License](https://img.shields.io/badge/License-TBD-lightgrey.svg)](#-license)
+
+[**Homepage**](https://epic-bench.github.io/EPIC-Bench/) | [**Paper**](https://epic-bench.github.io/EPIC-Bench/) | [**Dataset**](https://epic-bench.github.io/EPIC-Bench/) | [**Leaderboard**](https://epic-bench.github.io/EPIC-Bench/)
+
+</div>
+
+## 📃 Overview
+
+> This repo contains the official evaluation code and dataset for the paper
+> **"EPIC-Bench: Can VLMs Perceive the Embodied Real-World?"**
+
+**EPIC-Bench** is a **Mask-Grounding-based** benchmark designed to evaluate a VLM’s **Visual Perception** capability in **Embodied Scenarios**.
+
+<p align="center">
+  <img src="https://github.com/wei0623kb/EPIC-Bench-Eval/blob/main/images/teaser.png?raw=true" alt="EPIC-Bench teaser" width="100%"/>
+</p>
+
+📚 EPIC-Bench covers **3 High-Level Categories** and **23 Task Types**, following the realistic **Embodied Workflow**:
+
+- 🎯 **TargetLocalization**: **Pinpoint** the right object in the scene from a natural-language instruction.
+- 🧭 **Navigation**: **Approach** the target step by step by reading key visual cues along the way.
+- 🤲 **Manipulation**: **Operate** on the target through fine-grained, action-oriented **Grounded Perception**.
+
+The goal is to measure whether models can reliably perceive the critical **Visual** information required throughout the **Embodied Process**.
+
+## ✨ Highlights
+
+-  **Embodied-Scenario** evaluation of VLM **Visual Perception** capability.
+-  Focus on **Visual Grounding / Perception** without language shortcut exploitation.
+-  **Diverse** and **Fine-Grained** task design.
+
+## 📰 News
+
+- ~~[2026.5.10] 🚀 Huggingface Dataset and evaluation code are available!~~
+- [2026.5.10] 🚀 We released the ArXiv paper.
+
+## 📋 Todo
+
+- [x] Evaluation code for EPIC-Bench
+- [ ] Support an online leaderboard
+- [ ] Make the evaluation pipeline compatible with mask outputs
+
+
+# 🧰 EPIC-Bench Evaluation Toolkit
+
+This repository provides an end-to-end evaluation pipeline for **EPIC-Bench** on both:
+
+- **Open-Source VLMs** via the **[ms-swift](https://swift.readthedocs.io/zh-cn/latest/)** inference interface
+- **Closed-Source / API-Based VLMs** via **[lmms-eval](https://github.com/EvolvingLMMs-Lab/lmms-eval)**
+
+It includes **Dataset Conversion** utilities, **Inference Launchers**, **Response Standardization**, **Scoring**, and a Streamlit-based **Visualization** tool.
+
+
+## 🚀 Evaluation guide
+
+EPIC-Bench evaluation typically consists of the following stages.
+
+### ⚙️ 0) Environment setup
+
+Create a Python environment (example):
+
+```bash
+conda create -n epicbench python==3.10
+conda activate epicbench
+```
+
+Suggested dependencies (reference; choose what matches your model stack):
+
+| Model | Environment |
+|------|-------------|
+| General environment (compatible with Qwen2.5-VL/Qwen3-VL/InternVL/LLaVA-VL/Phi-4/gemma/RynnBrain/RoboBrain2/) | `pip install uv`<br>`uv pip install 'ms-swift' --torch-backend=auto`<br>`pip install vllm==0.15.1` |
+| Qwen3.5 | `git clone https://github.com/vllm-project/vllm.git`<br>`cd vllm`<br>`pip install -e .`<br>`git clone https://github.com/modelscope/ms-swift.git`<br>`cd ms-swift`<br>`pip install -e .`<br>`pip install transformers==5.2.0 qwen-vl-utils` |
+| Step models | `pip install onnxruntime-gpu tokenizers openai-whisper funasr vllm==0.15.1`<br>`git clone https://github.com/modelscope/ms-swift.git`<br>`cd ms-swift`<br>`pip install -e .` |
+| glm4.6 | `pip install uv`<br>`uv pip install 'ms-swift' --torch-backend=auto`<br>`pip install vllm==0.15.1 transformers==5.2.0` |
+
+For the visualization tool:
+
+```bash
+pip install streamlit pillow numpy pandas pycocotools
+```
+
+### 📦 1) Data preparation
+
+#### 1.1 Download raw annotations
+
+Download EPIC-Bench raw annotation data (and the referenced images) from the official release page (e.g., Hugging Face / ModelScope) and place them under:
+
+```
+dataset/annotation/EPIC_Bench
+```
+
+#### 1.2 Build ms-swift inference data (swift format)
+
+Generate ms-swift compatible inference JSON files from raw annotations:
+
+```bash
+bash scripts/build_swift_data.sh \
+  ANN_ROOT=dataset/annotation/EPIC_Bench \
+  OUT_DIR=dataset/swift_data/EPIC_Bench
+```
+
+Outputs will be written to:
+
+```
+dataset/swift_data/EPIC_Bench
+```
+
+#### 1.3 Customize prompts (optional)
+
+You can customize prompts and response formats in:
+
+- `tools/data_gen/prompts/`
+- `tools/data_gen/converters/`
+
+For best compatibility with the scoring pipeline, we recommend starting with the default settings in this repo.
+
+### 🤖 2) Inference
+
+Run inference using either:
+
+- example per-model scripts under `scripts/infer/<MODEL_FAMILY>/`, or
+- the unified launcher `scripts/infer.sh`
+
+Recommended (unified launcher):
+
+```bash
+bash scripts/infer.sh \
+  --model Qwen3_VL \
+  --data dataset/swift_data/EPIC_Bench \
+  --out outputs/model_response/swift_format
+```
+
+By default, raw ms-swift outputs are organized under:
+
+```
+outputs/model_response/swift_format/<model_series>/<model_name>.jsonl
+```
+
+Closed-source / API inference (optional):
+
+- `scripts/infer/api/` contains an example script for `lmms-eval`.
+- You must configure API keys via environment variables and **must not commit credentials** to GitHub.
+
+### 🔄 3) Standardize responses (std_format)
+
+Convert raw ms-swift outputs into EPIC-Bench **standard format** while preserving directory structure:
+
+```bash
+bash scripts/format_response.sh \
+  --in  outputs/model_response/swift_format \
+  --out outputs/model_response/std_format
+```
+
+If you evaluate a custom model/framework outside this repo, please ensure your outputs follow the **same std-format schema produced by** `tools/formatting/format_response.py`.
+
+### 📊 4) Scoring
+
+After obtaining standardized responses, compute detailed scores:
+
+```bash
+bash scripts/evaluate.sh \
+  --in  outputs/model_response/std_format \
+  --out outputs/scores
+```
+
+The scorer produces:
+
+- overall / category / type breakdowns
+- per-sample details (unless you pass `--no-details`)
+
+Supported formats:
+
+- **bbox** (most tasks)
+- **point** (FeasiblePath tasks)
+
+Mask-based evaluation is planned (releasing soon).
+
+
+```bash
+bash scripts/evaluate.sh 
+```
+
+### 📈 5) Visualization
+
+Launch the Streamlit visualization tool and default-load results from `outputs/scores`:
+
+```bash
+bash scripts/visualization.sh
+```
+
+
+## 🏆 Leaderboard and data examples
+
+Please refer to the [EPIC-Bench Homepage](https://epic-bench.github.io/EPIC-Bench/) for:
+
+- Leaderboard
+- Full dataset downloads
+- EPIC-Bench data examples
+
+## 📚 Citation
+
+```BibTeX
+@article{EPIC-Bench,
+  title={EPIC-Bench: Can VLMs Perceive the Embodied Real-World?},
+  author={XXX, XXX, XXX},
+  journal={},
+  year={2026}
+}
+```
+
+## 📜 License
+
+Please add an explicit `LICENSE` file before open-sourcing. If EPIC-Bench annotations or images have redistribution constraints, publish them separately (e.g., Hugging Face / ModelScope) and keep this repo code-only + small examples.
+
+## 🙏 Acknowledgements
+
+- **ms-swift** for open-source VLM inference: [ms-swift](https://swift.readthedocs.io/zh-cn/latest/)
+- **lmms-eval** for API/closed-source evaluation: [lmms-eval](https://github.com/EvolvingLMMs-Lab/lmms-eval)