Finance LLM
โ Large Language Models, Finance LLM, Reasoning โ 1 min read
Overview
๐กI was recently working on an LLM assignment, the problem statement was to train one of the LLM models on Financial data (mostly tables) to seek answers of the mathematical questions. ๐ง
I came across a research paper and discovered that LLMs are good in language pattern matching. But when it comes to imitation of complex reasoning abilities like human beings, their performance suffers.
๐With this intent, I used FinQA dataset1 (Question Answering on Financial Report) to conduct experiment around that. I compared few of the results of BERT model trained on FinQA with ChatGPT. Results can be seen in the video ๐ค
๐Given a tricky scenario where I ask the question and given partial table to ChatGPT to answer, the finding was interesting.๐
For code and experiment you can visit the Github repository
https://github.com/dravinash/FinanceLLM
The live implementation can be seen in the below video.
A detailed document can be seen here. document
References
Footnotes
-
Chen, Z., Chen, W., Smiley, C., Shah, S., Borova, I., Langdon, D., ... & Wang, W. Y. (2021). Finqa: A dataset of numerical reasoning over financial data. arXiv preprint arXiv:2109.00122. โฉ