The University of British Columbia and Invertible AI jointly launched FinTral, a new multi-modal large-scale language model for the financial field. The model is based on Mistral-7b and is capable of processing text, numeric and visual data, and its performance even surpasses GPT-4. FinTral has demonstrated excellent performance in financial analysis tasks through domain pre-training and AI feedback optimization, especially the FinTral-DPO model, whose performance is close to the level of GPT-4. This research result has brought new breakthroughs to the application of artificial intelligence in the financial field.
Research results show that FinTral performs outstandingly in multiple financial tasks, especially surpassing other models in visual data processing and tool retrieval. This shows that FinTral has significant advantages in processing complex financial information and assisting financial analysis.
The emergence of FinTral marks a major progress in the application of multi-modal large-scale language models in the financial field, providing a powerful tool for financial analysis and decision-making, and pointing out a new direction for future financial technology development. Its advantages in visual data processing deserve more attention, indicating that financial technology may achieve more intuitive and intelligent development in the future.