Reproduce “FinSphere, a Real-Time Stock Analysis Agent Powered by Instruction-Tuned LLMs and Domain Tools” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Shijie Han, Jingshu Zhang, Yiqing Shen, Kaiyuan Yan, Hongguang Li
FinSphere is a real-time stock analysis agent developed by JF SmartInvest Holdings Ltd. that utilizes an instruction-tuned Qwen2-72B language model integrated with proprietary quantitative tools and a real-time financial database. The system produces high-quality, professional stock reports, achieving an AnalyScore of 70.88 out of 100, outperforming leading general-purpose and domain-specific LLMs as well as other agent-based systems.
Key insights
- Instruction tuning an LLM with expert-curated data, such as the Stocksis dataset, is crucial for achieving professional-grade analytical depth and expression in complex financial tasks.
- Seamless integration of LLMs with real-time financial databases and specialized quantitative tools is essential for overcoming knowledge cut-off limitations and providing dynamic, current market insights.
- A systematic, expert-developed evaluation framework like AnalyScore is fundamental for objectively benchmarking performance, ensuring reliability, and driving further advancements in financial AI.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.