Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
NRI Indonesia Correspondence: t3-okada@nri.co.jp Abstract In adapting LLMs to specific domains, achieving both domain expertise and reasoning ability remains an urgent challenge. This study proposes a general method for constructing high-quality synthetic instruction data for any domain, starting from domain-specific vocabulary. As a demonstration, we applied this method to the financial domain and constructed a large-scale instruction dataset totaling approximately 9.5 billion tokens with Chain-of-Thought reasoning traces. Evaluation results confirmed performance improvements over baseline models on financial benchmarks, demonstrating the effectiveness of our approach. We also report findings on the impact of reasoning trace length on performance and its limitations. Lastly, we open-source our models and datasets on HuggingFace. Constructing Synthetic Instruction Datasets for Improving Reasoning in Domain-Specific LLMs: A Case Study in the Japanese Financial Domain Yuma Okochi1, Fabio Milentiansen Sim2, Tomoyasu Okada1
1
0
3
Unlock domain-specific LLM reasoning by generating synthetic instruction data from vocabulary alone, achieving gains on financial benchmarks.