Construction of Domain-specified Japanese Large Language Model for Finance through Continual Pre-training

By Masanori Hirano, Kentaro Imajo

Rating

1230
Battle Count: 76

Relevance

7/10
The model shows potential for improving financial text analysis and information extraction, which could be valuable for quantitative trading strategies.

Implementation Complexity

6/10
The paper provides detailed methodology, but implementing and tuning large language models requires significant computational resources and expertise.

Reproducibility

4/5
The paper provides detailed information on the methodology, datasets, and evaluation process. The tuned model is publicly available on Hugging Face.

About this paper

Methodology: Continual Pre-training. Problem types: Natural Language Processing, Domain Adaptation.

The interactive Everscope explorer (charts, battles, favorites) loads below.