RD-Agent

microsoft/RD-Agent

Automating high-value generic R&D processes through AI.

14k
Stars
1.8k
Forks
192
Open issues
87
Watchers
Python MITLast pushed Jun 15, 2026

Overview

R&D-Agent is a Python-based tool automating data and model-centric research and development tasks, enhancing productivity in the AI era. It supports LLM fine-tuning, quant trading, and integrates with various LLM providers via LiteLLM backend.

Categories

Tags

Similar tools

Install

pip install RD-Agent

README

🖥️ Live Demo | 🎥 Demo Video ▶️YouTube | 📖 Documentation | 📄 Tech Report | 📃 Papers

📰 News

🗞️ News📝 Description
ICML 2026 AcceptanceWe are thrilled to announce that our paper FT-Dojo: Towards Autonomous LLM Fine-Tuning with Language Agents has been accepted to ICML 2026. The FT-Agent implementation is available in the LLM fine-tuning guide.
ACL 2026 Findings AcceptanceWe are thrilled to announce that our paper Reasoning as Gradient has been accepted to ACL 2026 Findings. Execution traces are available at Gome GPT-5 Traces
Web UI ReleaseWe release a new frontend that can be built and served by rdagent server_ui for real-time interaction and trace viewing, currently excluding the data_science scenario.
NeurIPS 2025 AcceptanceWe are thrilled to announce that our paper R&D-Agent-Quant has been accepted to NeurIPS 2025
Technical Report ReleaseOverall framework description and results on MLE-bench
R&D-Agent-Quant ReleaseApply R&D-Agent to quant trading
MLE-Bench Results ReleasedR&D-Agent currently leads as the top-performing machine learning engineering agent on MLE-bench
Support LiteLLM BackendWe now fully support LiteLLM as our default backend for integration with multiple LLM providers.
General Data Science AgentData Science Agent
Kaggle Scenario releaseWe release Kaggle Agent, try the new features!
Official WeChat group releaseWe created a WeChat group, welcome to join! (🗪QR Code)
Official Discord releaseWe launch our first chatting channel in Discord (🗪)
First releaseR&D-Agent is released on GitHub

🏆 The Best Machine Learning Engineering Agent!

MLE-bench is a comprehensive benchmark evaluating the performance of AI agents on machine learning engineering tasks. Utilizing datasets from 75 Kaggle competitions, MLE-bench provides robust assessments of AI systems' capabilities in real-world ML engineering scenarios.

R&D-Agent currently leads as the top-performing machine learning engineering agent on MLE-bench:

AgentLow == Lite (%)Medium (%)High (%)All (%)
R&D-Agent o3(R)+GPT-4.1(D)51.52 ± 6.919.3 ± 5.526.67 ± 030.22 ± 1.5
R&D-Agent o1-preview48.18 ± 2.498.95 ± 2.3618.67 ± 2.9822.4 ± 1.1
AIDE o1-preview34.3 ± 2.48.8 ± 1.110.0 ± 1.916.9 ± 1.1

Notes:

  • O3(R)+GPT-4.1(D): This version is designed to both reduce average time per loop and leverage a cost-effective combination of backend LLMs by seamlessly integrating Research Agent (o3) with Development Agent (GPT-4.1).
  • AIDE o1-preview: Represents the previously best public result on MLE-bench as reporte