<aside> 🚀
Show us how you’ll help to realise our mission to make the 3-day working week a reality.
</aside>
We kindly ask that you complete this task within one week. If this is not possible, please let us know a suitable date.
We’d like you to demonstrate a LLM driven prototype that can answer conversational dialogue questions on any given financial document. For this task we ask you to use a cleaned version of the ConvFinQA Dataset.
<aside> 👉
This dataset consists of multi-hop numerical reasoning questions over semi-structured financial documents, containing text & datatables.
</aside>
We would like to see what you can build with the data provided. Feel free to present the results in any format you prefer and explore any additional ideas you have with the dataset. You may use any model or architecture of your choice. The goal is to demonstrate your knowledge and experience. We are particularly interested in the logic and reasoning behind your choice of accuracy metrics and your ability to communicate your solutions and ideas effectively.
Here is an example question from the paper :
dataset.md which contains information on the dataset fields and key highlights from the original paper .