Summary

QuantaAlpha is an open-source framework automating quantitative alpha factor discovery by integrating LLM intelligence with evolutionary strategies. Users define research directions, and the system automatically mines, evolves, and validates factors via self-evolving trajectories, aiming to achieve superior quantitative alpha for researchers and engineers.

How It Works

The core approach is an LLM-driven, self-evolving methodology. It utilizes diversified planning initialization, trajectory-level evolution, and structured hypothesis-code constraints to explore complex factor spaces efficiently, leading to novel and robust alpha signals.

Quick Start & Requirements

Installation requires cloning the repository, creating a Conda environment (Python 3.10 for Linux, 3.11 for Windows), and installing dependencies via pip. Configuration involves setting up LLM API keys (OpenAI compatible) and specifying paths for Qlib market data and pre-computed HDF5 files, downloadable from HuggingFace. Execution is via ./run.sh (Linux) or python launcher.py (Windows). A Node.js-based web UI is also available.

Highlighted Details

• Achieved performance metrics include IC: 0.1501, Rank IC: 0.1465, ARR: 27.75%, MDD: 7.98%, CR: 3.4774.
• Provides an end-to-end workflow from research input to factor mining, validation, and backtesting.
• Includes an optional web dashboard for visual interaction.

Maintenance & Community

Founded by academics from leading universities, the project focuses on AI in finance research, with future work on CodeAgent and Agentic Reasoning. Community engagement is via a WeChat Group. Contributions are encouraged.

Licensing & Compatibility

The repository's license is not explicitly stated, posing a significant adoption risk. The project is primarily Linux-native, with Windows support requiring specific configurations and potential workarounds.

Limitations & Caveats

Windows deployment can be complex and error-prone. Reliance on external LLM APIs may incur costs. Data preparation involves large downloads or slow generation. The unspecified license is a critical barrier for many use cases.