FinRpt: Dataset, Evaluation System and LLM-based Multi-agent Framework for Equity Research Report Generation

The recent paper 'FinRpt: Dataset, Evaluation System and LLM-based Multi-agent Framework for Equity Research Report Generation' marks a significant advancement in the field of financial technology by introducing the first comprehensive framework for automating Equity Research Report (ERR) generation. This initiative addresses the existing data scarcity and lack of evaluation metrics by presenting the FinRpt benchmark, which integrates seven types of financial data to create a high-quality ERR dataset. The authors also propose a multi-agent framework, FinRpt-Gen, which utilizes advanced training methods such as supervised fine-tuning and reinforcement learning. Experimental results demonstrate the effectiveness of the evaluation metrics and the strong performance of the FinRpt-Gen framework, indicating its potential to drive innovation in equity research. The availability of all code and datasets publicly further enhances the accessibility and applicability of this research, paving the …