Summary of Thai Financial Domain Adaptation Of Thalle — Technical Report, by Kbtg Labs et al.
Thai Financial Domain Adaptation of THaLLE – Technical Report
by KBTG Labs, Atthakorn Petchsod, Pornchanan Balee, Danupat Khamnuansin, Anuruth Lertpiya, Chanatip Saetia, Tawunrat Chalothorn, Thadpong Pongthawornkamol, Monchai Lertsutthiwong
First submitted to arxiv on: 27 Nov 2024
Categories
- Main: Computation and Language (cs.CL)
- Secondary: Artificial Intelligence (cs.AI)
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 proposed Large Language Model (LLM) excels in general tasks but struggles with domain-specific challenges. The existing financial LLMs, FinGPT and BloombergGPT, lack support for the Thai financial domain. To address this gap, a new Thai Financial LLM is developed using the Investment Consultant (IC) exam dataset from the Stock Exchange of Thailand. Data augmentation, ReLoRA, Continued Pretraining (CPT), and Rank-Stabilized LoRA (rsLoRA) are applied to improve training efficiency. Supervised Fine-Tuning (SFT) simulates exam scenarios, while Direct Preference Optimization (DPO) refines the model using feedback. The model achieves scores of 72%, 72%, and 84% on IC exam levels P1, P2, and P3, respectively, demonstrating its effectiveness in Thai financial advisory tasks. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary The research develops a special language model for the Thai financial domain. This model is better at understanding and generating text about finance in Thailand than other models. It uses data from the Investment Consultant (IC) exam to learn what words and phrases are important in this area of finance. The researchers tried different ways to make the model more accurate, such as using extra training data or fine-tuning the model with feedback. They tested the model on real-world financial advisory tasks and found it was very good at answering questions and providing advice. |
Keywords
» Artificial intelligence » Data augmentation » Fine tuning » Language model » Large language model » Lora » Optimization » Pretraining » Supervised
