abocide/Qwen2.5-7B-Instruct-R1-forfinance

Overview

This model, abocide/Qwen2.5-7B-Instruct-R1-forfinance, is a 7.6 billion parameter large language model derived from Qwen2.5-7B-Instruct. It has undergone full parameter fine-tuning specifically for the financial sector. The training incorporated a blend of open-source financial question-and-answer datasets and high-quality chain-of-thought reasoning data. Notably, the chain-of-thought data was generated by DeepSeek-R1 and then rigorously quality-scored by GPT-5 to ensure high fidelity.

Key Capabilities

• Financial Knowledge Q&A: Answers questions on basic financial concepts and complex topics.
• Financial Calculations and Analysis: Capable of solving financial calculation problems.
• Financial Concept Explanations: Provides detailed explanations of financial terms and theories.
• Chain-of-Thought Reasoning: Enhanced reasoning abilities for complex financial scenarios.

Training Details

The model was fine-tuned using Supervised Fine-Tuning (SFT) on 8 NVIDIA A100 GPUs. The training process involved 2 epochs, achieving a final training loss of 0.7332 over 312 steps. Future plans include reinforcement learning training using GRPO to further enhance performance and safety in the financial domain.

Use Cases

• Financial knowledge inquiry
• Financial calculation and analysis support
• Investment advice consultation (for research, not direct advice)
• Risk assessment assistance

Important Note: This model is intended for educational and research purposes only and does not provide investment advice. Users should exercise caution and fact-check outputs.