From eb7f19a9834c66d6a263a0a66b916559a01a8003 Mon Sep 17 00:00:00 2001
From: Yuhang Zhou <86864241+Ralph-Zhou@users.noreply.github.com>
Date: Sun, 18 May 2025 17:29:25 +0800
Subject: [PATCH] Initial update README.md

---
 README.md | 47 +++++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 47 insertions(+)
 create mode 100644 README.md

diff --git a/README.md b/README.md
new file mode 100644
index 0000000..f41852a
--- /dev/null
+++ b/README.md
@@ -0,0 +1,47 @@
+# OWL: Optimized Workforce Learning for General Multi-Agent Assistance for Real-World Task Automation
+
+We present Workforce, a hierarchical multi-agent framework that decouples planning from execution through a modular
+architecture with a domain-agnostic Planner, Coordinator, and specialized Workers. This enables cross-domain transfer by
+allowing worker modification without full system retraining. On the GAIA benchmark, Workforce achieves state-of-the-art
+69.70% accuracy, outperforming commercial systems.
+
+This repository contains inference part code for the OWL framework (Workforce).
+
+## Inference
+
+The camel version we use is `0.2.46`. To reproduce Workforce inference performance (69.70% - Claude-3.7 accuracy on GAIA benchmark and 60.61% - GPT-4o
+accuracy on GAIA benchmark), follow the steps below:
+
+### Installation and Setup
+
+1. Create a Python 3.11 Conda environment:
+
+```bash
+conda create -n owl python=3.11
+```
+
+2. Install the required packages:
+
+```bash
+pip install -r requirements.txt
+```
+
+3. Set up envionment variables:
+
+copy `.env.example` to `.env` and set the environment variables, and set the keys in `.env` file.
+
+4. Run the inference:
+
+- For reproducing results using GPT-4o, run:
+
+```bash
+python run_gaia_workforce.py
+```
+
+- For reproducing results using Claude-3.7, run:
+
+```bash
+python run_gaia_workforce_claude.py
+```
+
+