Update README.md

README.md

@@ -82,144 +82,6 @@ This repository provides an end-to-end evaluation pipeline for **EPIC-Bench** on
 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: