EDINET-BENCH: EVALUATING LLMS ON COMPLEX FINANCIAL TASKS USING JAPANESE FINANCIAL STATEMENTS

2026

Open publication workspace · Sign in to read the full PDF.

AI-generated summary

1) This paper introduces EDINET-Bench, a novel benchmark for evaluating Large Language Models on complex Japanese financial tasks, highlighting current LLM limitations and the need for more realistic financial evaluation frameworks.
2) * A comprehensive benchmark, EDINET-Bench, built from ten years of Japanese financial reports.
* Three challenging tasks: accounting fraud detection, earnings forecasting, and industry classification.
* Experimental results showing state-of-the-art LLMs struggle, performing only marginally better than simpler models.
3) LLMs, Financial Benchmarks, Japanese Financial Statements, Fraud Detection, Earnings Forecasting

Check the original publication for accuracy and context.