Summary of Kodexv0.1: a Family Of State-of-the-art Financial Large Language Models, by Neel Rajani et al.
KodeXv0.1: A Family of State-of-the-Art Financial Large Language Models
by Neel Rajani, Lilli Kiessling, Aleksandr Ogaltsov, Claus Lang
First submitted to arxiv on: 13 Sep 2024
Categories
- Main: Computation and Language (cs.CL)
- Secondary: Artificial Intelligence (cs.AI); Computational Finance (q-fin.CP)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary The paper introduces KodeXv0.1, a family of large language models that surpasses GPT-4 in financial question answering. The authors adapt Llama 3.1 base variants to the financial domain through a custom training regime using publicly available financial documents and generate a synthetic dataset for instruction tuning. They perform extensive model evaluations on FinanceBench, FinQABench, and their own dataset, showing that KodeX-8Bv0.1 is more reliable in financial contexts than cutting-edge instruct models, and even outperforms state-of-the-art proprietary models like GPT-4 by up to 7.07%. The authors also introduce KodeX-70Bv0.1, which further improves upon this performance. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary The paper creates special language models for financial questions that are better than the current best ones. They take two big language models and teach them to understand financial documents like earnings calls and business reports. Then they test these new models on financial tasks and find that they do a lot better than other models, even some very good ones made by companies. |
Keywords
» Artificial intelligence » Gpt » Instruction tuning » Llama » Question answering
