Update README.md

README.md

@@ -125,24 +125,26 @@ tags:
 ---
 
 # EDINET-Bench
-EDINET-Bench is a comprehensive financial benchmark built from the securities reports of publicly listed companies in Japan, sourced from [EDINET](https://disclosure2.edinet-fsa.go.jp/).
-It includes challenging tasks such as accounting fraud detection, earnings forecasting, and industry prediction—problems that require sophisticated financial reasoning.
+EDINET-Bench is a Japanese financial benchmark designed to evaluate the performance of LLMs on challenging financial tasks including accounting fraud detection, earnings forecasting, and industry prediction.
+This dataset is built leveraging [EDINET](https://disclosure2.edinet-fsa.go.jp), a platform managed by the Financial Services Agency (FSA) of Japan that provides access to disclosure documents such as securities reports.
 
+## Resources
+- Paper: https://arxiv.org/abs/xxx
+- Counstruction code of EDINET-Bench: https://github.com/SakanaAI/edinet2dataset
+- Evaluation code using EDINET-Bench: https://github.com/SakanaAI/edinet-bench
 
-## Dataset Construction
+## Dataset Construction Pipeline
 
 <img src="EDINET-Bench.png" alt="Overview of EDINET-Bench" width="50%"/>
 
-EDINET-Bench is built by downloading the past 10 years of annual reports from Japanese listed companies on EDINET and automatically annotating labels for each task. For detailed implementation.
-The complete dataset construction pipeline is available at https://github.com/SakanaAI/edinet2dataset.
+EDINET-Bench is built by downloading the past 10 years of annual reports from Japanese listed companies on EDINET and automatically annotating labels for each task. 
+For detailed information, please read our paper and code.
 
-## How to Use
 
-Each task contains (Annual report, Label) pairs.
-Each report includes pre-extracted information such as the EDINET Code and DOC ID associated with the report, along with data sections (META, BS, CF, PL, Summary, and Text).
+## How to Use
 
 - Acounting fraud detection
-
+This task is a binary classification problem aimed at predicting whether a given annual report is fraudulent.
 The label is either fraud (1) or non-fraud (0).
 
 ```python
@@ -150,29 +152,27 @@ ds = load_dataset("SakanaAI/EDINET-Bench", "fraud_detection")
 ```
 
 - Earnings forecast
-
+This task is a binary classification problem that predicts whether a company's earnings will increase or decrease in the next fiscal year based on its current annual report.
 The label is either increase (1) or not (0).
 ```python
 ds = load_dataset("SakanaAI/EDINET-Bench", "earnings_forecast")
 ```
 
 - Industry prediction
+This task is a multi-class classification problem that predicts a company's industry type (e.g., IT) based on its current annual report.
+Each label represents one of 16 possible industry types.
 
-The label is one of 17 industry names.
 ```python
 ds = load_dataset("SakanaAI/EDINET-Bench", "industry_prediction")
 ```
 
-## Limitations
-
-Mislabeling: For the Accounting Fraud Detection task, the fraud labels are determined by feeding the text information from the correction reports into Claude Sonnet 3.7, which classifies whether the correction is related to fraud.
-As a result, there is a possibility of false positives. Additionally, for non-fraud cases, there may be instances that actually involve fraud but have not yet been identified.
-
 ## LICENSE
 
 EDINET-Bench is released under [CC BY 4.0](https://creativecommons.org/licenses/by/4.0/deed.en).
 
-## References
-- EDINET, https://disclosure2.edinet-fsa.go.jp/
-- https://github.com/SakanaAI/edinet2dataset
-- https://github.com/SakanaAI/edinet-bench
+## Citation
+
+```
+TODO:
+```
+