Spaces:
Running
Running
| title: OpenFinMind | |
| emoji: 🧠 | |
| colorFrom: blue | |
| colorTo: purple | |
| sdk: static | |
| pinned: false | |
| short_description: Open research community advancing AI for finance | |
| ## OpenFinMind | |
| **Open Financial Intelligence** · **An Open Research Community for AI in Finance** | |
| OpenFinMind is an open research community at **The University of Hong Kong (HKU)**, advancing artificial intelligence in finance through open research, collaborative development, and shared resources. | |
| **Research Areas** | |
| - **AI Agents & Financial Research** — LLM-based agents, multi-agent systems, and deep research for financial analysis and decision-making. | |
| - **Quantitative Finance & Alpha Discovery** — Code-based alpha mining, factor research, quantitative strategies, and systematic investment. | |
| - **Financial Forecasting & Evaluation** — Financial prediction, temporal evaluation, and rigorous benchmarks for AI systems. | |
| - **Financial Data & Research Infrastructure** — Curated datasets, reproducible workflows, and open-source tools for financial AI. | |
| **Selected Research** | |
| - **CogAlpha** — An LLM-driven framework combining financial reasoning and code-based evolutionary search for automated alpha discovery. [Paper](https://aclanthology.org/2026.acl-long.538/) · [PDF](https://aclanthology.org/2026.acl-long.538.pdf) · [Code](https://github.com/OpenFinMind/CogAlpha_Prompt) · ACL 2026. | |
| - **X-Alpha** — An AI Quant Researcher that integrates report-grounded financial knowledge and persistent discovery feedback into a closed-loop hypothesis-to-code research process. [Paper](https://arxiv.org/abs/2607.08332) · arXiv:2607.08332. | |
| - **FinForesight** — A diagnostic benchmark evaluating the financial forecasting capabilities of deep research systems across CN, US, and HK markets. [Paper](https://arxiv.org/abs/2611.01621) · [Code](https://github.com/OpenFinMind/FinForesight) · [Dataset](https://huggingface.co/datasets/OpenFinMind/FinForesight) · arXiv:2611.01621. | |
| - **MC-JEPA** — A market-conditioned joint-embedding predictive architecture that leverages realized market shocks during self-supervised training to learn asset-specific representations, improving financial return prediction, market exposure estimation, risk, and liquidity modeling. [Paper](https://arxiv.org/abs/2611.13572) · [Code](https://github.com/OpenFinMind/MC-JEPA) · arXiv:2611.13572. | |
| **Open Research & Resources** | |
| We develop and share research projects, datasets, benchmarks, models, and tools to support reproducible and collaborative financial AI research. [GitHub](https://github.com/OpenFinMind) · [Hugging Face](https://huggingface.co/OpenFinMind). | |
| **Open Collaboration** | |
| We welcome researchers, developers, and practitioners interested in AI for finance to explore our projects, share resources, and contribute to open research. | |
| **Build. Share. Discover. Advance.** |