Baichuan4-Finance Technical Report

2024-12-17

AI Summary

Large language models (LLMs) exhibit immature financial domain knowledge understanding and application. Method: This paper introduces the Baichuan4-Finance series—comprising the foundational model Baichuan4-Finance-Base and the aligned dialogue model Baichuan4-Finance—designed specifically for the financial vertical. We propose an in-domain self-constrained continual pretraining strategy to strengthen financial expertise while preserving general capabilities, and a dual-path reinforcement learning framework integrating both human feedback (RLHF) and AI feedback (RLAIF) to jointly optimize performance on financial certification benchmarks and real-world financial tasks. Data quality governance, supervised fine-tuning (SFT), and multi-stage alignment training are systematically employed. Contribution/Results: Baichuan4-Finance-Base achieves state-of-the-art results on major financial benchmarks without sacrificing general-domain performance; Baichuan4-Finance demonstrates superior effectiveness across practical financial applications, advancing the ecosystem of financial LMs.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: (Large) Language Models
• Computer Vision: Large Vision Models

Application Category

• Economics, Online Markets and Human Computation: Humans versus LLMs for data annotation and labeling
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Search Tool Learning with LLM: Teaching LLMs to invoke search and make use of retrieved information

Abstract

Large language models (LLMs) have demonstrated strong capabilities in language understanding, generation, and reasoning, yet their potential in finance remains underexplored due to the complexity and specialization of financial knowledge. In this work, we report the development of the Baichuan4-Finance series, including a comprehensive suite of foundational Baichuan4-Finance-Base and an aligned language model Baichuan4-Finance, which are built upon Baichuan4-Turbo base model and tailored for finance domain. Firstly, we have dedicated significant effort to building a detailed pipeline for improving data quality. Moreover, in the continual pre-training phase, we propose a novel domain self-constraint training strategy, which enables Baichuan4-Finance-Base to acquire financial knowledge without losing general capabilities. After Supervised Fine-tuning and Reinforcement Learning from Human Feedback and AI Feedback, the chat model Baichuan4-Finance is able to tackle various financial certification questions and real-world scenario applications. We evaluate Baichuan4-Finance on many widely used general datasets and two holistic financial benchmarks. The evaluation results show that Baichuan4-Finance-Base surpasses almost all competitive baselines on financial tasks by significant margins without sacrificing performance on general LLM benchmarks. At the same time, Baichuan4-Finance demonstrates even more impressive performance on financial application scenarios, showcasing its potential to foster community innovation in the financial LLM field.

Problem

• Large Language Models
• Financial Domain
• Complex Financial Knowledge

Innovation

• Baichuan4-Finance Model
• Financial Language Specialization
• Enhanced Performance in Financial Applications