QianfanHuijin Technical Report: A Novel Multi-Stage Training Paradigm for Finance Industrial LLMs

TL;DR

This work targets the finance domain by delivering QianfanHuijin, a specialized LLM that combines deep domain knowledge, robust numerical reasoning, and agentic tool usage. It introduces a four-stage Progressive Post-training pipeline (SFT → Reasoning RL → Agentic RL → General RL) complemented by a Continual Pre-training phase and a Controllable Instruction Synthesis Framework (CIS-F) with a Dual-Verifier Reward Model to ensure high-quality data and reinforcement signals. Empirical results across finance knowledge, numerical reasoning, and tool-enabled tasks show strong gains over comparable models, with ablations confirming the critical role of the Reasoning RL and Agentic RL stages. The proposed framework aims to enable reliable, efficient, and compliant deployment of industrial finance LLMs capable of real-world decision support and regulatory alignment.

Abstract

Domain-specific enhancement of Large Language Models (LLMs) within the financial context has long been a focal point of industrial application. While previous models such as BloombergGPT and Baichuan-Finance primarily focused on knowledge enhancement, the deepening complexity of financial services has driven a growing demand for models that possess not only domain knowledge but also robust financial reasoning and agentic capabilities. In this paper, we present QianfanHuijin, a financial domain LLM, and propose a generalizable multi-stage training paradigm for industrial model enhancement. Our approach begins with Continual Pre-training (CPT) on financial corpora to consolidate the knowledge base. This is followed by a fine-grained Post-training pipeline designed with increasing specificity: starting with Financial SFT, progressing to Finance Reasoning RL and Finance Agentic RL, and culminating in General RL aligned with real-world business scenarios. Empirical results demonstrate that QianfanHuijin achieves superior performance across various authoritative financial benchmarks. Furthermore, ablation studies confirm that the targeted Reasoning RL and Agentic RL stages yield significant gains in their respective capabilities. These findings validate our motivation and suggest that this fine-grained, progressive post-training methodology is poised to become a mainstream paradigm for various industrial-enhanced LLMs.

Figures (6)

• Figure 1: Illustration of the multi-stage training framework of QianfanHuijin
• Figure 2: The Data Cleaning and Filtering Pipeline
• Figure 3: The Data Synthesis Workflow
• Figure 4: FinanceIQ Submetrics
• Figure 5: Illustration of 4 Stages