Fin-R1
| Property | Value |
|---|---|
| Parameter Count | 7B |
| Base Model | Qwen2.5-7B-Instruct |
| Developer | SUFE-AIFLM-Lab |
| Model URL | https://huggingface.co/SUFE-AIFLM-Lab/Fin-R1 |
What is Fin-R1?
Fin-R1 is a specialized financial reasoning language model developed by Shanghai University of Finance and Economics' AIFLM Lab. Built on Qwen2.5-7B-Instruct, it's specifically designed for complex financial reasoning tasks, achieving state-of-the-art performance across multiple financial benchmarks despite its relatively compact 7B parameter size.
Implementation Details
The model employs a two-stage training approach combining supervised fine-tuning (SFT) and reinforcement learning (RL). It's trained on a carefully curated dataset of 60,091 high-quality financial samples covering various aspects including financial code, calculations, and ESG analysis.
- Uses GRPO (Group Relative Policy Optimization) algorithm for reinforcement learning
- Implements dual reward mechanism for format and accuracy optimization
- Incorporates model-based verifier using Qwen2.5-Max for answer evaluation
Core Capabilities
- Financial code interpretation and generation
- Complex financial calculations in both English and Chinese
- Regulatory compliance checking
- Intelligent risk control
- ESG analysis and reporting
- Multi-turn financial reasoning
Frequently Asked Questions
Q: What makes this model unique?
Fin-R1 achieves remarkable performance with just 7B parameters, scoring 75.2 on average across financial benchmarks, making it highly efficient for deployment while maintaining high accuracy in financial reasoning tasks.
Q: What are the recommended use cases?
The model is ideal for banking, securities, insurance, and trust operations, particularly excelling in financial QA tasks, regulatory compliance checking, and complex financial calculations. It achieves top scores in FinQA (76.0) and ConvFinQA (85.0) benchmarks.
