Update README.md
Browse files
README.md
CHANGED
|
@@ -6,4 +6,86 @@ language:
|
|
| 6 |
- zh
|
| 7 |
- en
|
| 8 |
pretty_name: e
|
| 9 |
-
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 6 |
- zh
|
| 7 |
- en
|
| 8 |
pretty_name: e
|
| 9 |
+
---
|
| 10 |
+
|
| 11 |
+
<p align="center">
|
| 12 |
+
<h1 align="center">
|
| 13 |
+
<img src="static/logo.png" alt="BizFinBench logo" height="40" style="position:relative; top:6px;">
|
| 14 |
+
BizFinBench.v2: A Unified Dual-Mode Bilingual Benchmark for Expert-Level Financial Capability Alignment</h1>
|
| 15 |
+
<p align="center">
|
| 16 |
+
<span class="author-block">
|
| 17 |
+
Xin Guo<sup>1,2,*</sup> </a>,</span>
|
| 18 |
+
<span class="author-block">
|
| 19 |
+
Rongjunchen Zhang<sup>1,*,♠</sup></a>, Guilong Lu<sup>1</sup>, Xuntao Guo<sup>1</sup>, Jia Shuai<sup>1</sup>, Zhi Yang<sup>2</sup>, Liwen Zhang<sup>2,♠</sup>
|
| 20 |
+
</span>
|
| 21 |
+
</div>
|
| 22 |
+
<div class="is-size-5 publication-authors" style="margin-top: 10px;">
|
| 23 |
+
<span class="author-block">
|
| 24 |
+
<sup>1</sup>HiThink Research, <sup>2</sup>Shanghai University of Finance and Economics
|
| 25 |
+
</span>
|
| 26 |
+
<br>
|
| 27 |
+
<span class="author-block">
|
| 28 |
+
<sup>*</sup>Co-first authors, <sup>♠</sup>Corresponding author, zhangrongjunchen@myhexin.com,zhang.liwen@shufe.edu.cn
|
| 29 |
+
</span>
|
| 30 |
+
</div>
|
| 31 |
+
</p>
|
| 32 |
+
<p>
|
| 33 |
+
📖<a href="">Paper</a> |🏠<a href="">Homepage</a></h3>|🤗<a href="">Huggingface</a></h3>
|
| 34 |
+
</p>
|
| 35 |
+
<div align="center"></div>
|
| 36 |
+
<p align="center">
|
| 37 |
+
|
| 38 |
+
**BizFinBench.v2** is the secend release of [BizFinBench](https://github.com/HiThink-Research/BizFinBench). It is built entirely on real-world user queries from Chinese and U.S. equity markets. It bridges the gap between academic evaluation and actual financial operations.
|
| 39 |
+
|
| 40 |
+
<img src="static/score_sequence.png" alt="Evaluation Result">
|
| 41 |
+
|
| 42 |
+
### 🌟 Key Features
|
| 43 |
+
|
| 44 |
+
* **Authentic & Real-Time:** 100% derived from real financial platform queries, integrating online assessment capabilities.
|
| 45 |
+
* **Expert-Level Difficulty:** A challenging dataset of **29,578 Q&A pairs** requiring professional financial reasoning.
|
| 46 |
+
* **Comprehensive Coverage:** Spans **4 core business scenarios**, 8 fundamental tasks, and 2 online tasks.
|
| 47 |
+
|
| 48 |
+
### 📊 Key Findings
|
| 49 |
+
* **High Difficulty:** Even **ChatGPT-5** achieves only 61.5% accuracy on main tasks, highlighting a significant gap vs. human experts.
|
| 50 |
+
* **Online Prowess:** **DeepSeek-R1** outperforms all other commercial LLMs in dynamic online tasks, achieving a total return of 13.46% with a maximum drawdown of -8%.
|
| 51 |
+
|
| 52 |
+
## 📢 News
|
| 53 |
+
- 🚀 [06/01/2026] TBD
|
| 54 |
+
|
| 55 |
+
## 📕 Data Distrubution
|
| 56 |
+
BizFinBench.v2 contains multiple subtasks, each focusing on a different financial understanding and reasoning ability, as follows:
|
| 57 |
+
|
| 58 |
+
### Distribution Visualization
|
| 59 |
+
<div align="center">
|
| 60 |
+
<img src="static/distribution.png" alt="Data Distribution" width="600">
|
| 61 |
+
</div>
|
| 62 |
+
|
| 63 |
+
### Detailed Statistics
|
| 64 |
+
| Scenarios | Tasks | Avg. Input Tokens | # Questions |
|
| 65 |
+
|:---|:---|---:|---:|
|
| 66 |
+
| **Business Information Provenance** | Anomaly Information Tracing | 8,679 | 4,000 |
|
| 67 |
+
| | Financial Multi-turn Perception | 10,361 | 3,741 |
|
| 68 |
+
| | Financial Data Description | 3,577 | 3,837 |
|
| 69 |
+
| **Financial Logic Reasoning** | Financial Quantitative Computation | 1,984 | 2,000 |
|
| 70 |
+
| | Event Logic Reasoning | 437 | 4,000 |
|
| 71 |
+
| | Counterfactual Inference | 2,267 | 2,000 |
|
| 72 |
+
| **Stakeholder Feature Perception** | User Sentiment Analysis | 3,326 | 4,000 |
|
| 73 |
+
| | Financial Report Analysis | 19,681 | 2,000 |
|
| 74 |
+
| **Real-time Market Discernment** | Stock Price Prediction | 5,510 | 4,000 |
|
| 75 |
+
| | Portfolio Asset Allocation | — | — |
|
| 76 |
+
| **Total** | **—** | **—** | **29,578** |
|
| 77 |
+
|
| 78 |
+
|
| 79 |
+
## ✒️Citation
|
| 80 |
+
|
| 81 |
+
```
|
| 82 |
+
Coming Soon
|
| 83 |
+
```
|
| 84 |
+
|
| 85 |
+
## 📄 License
|
| 86 |
+
  **Usage and License Notices**: The data and code are intended and licensed for research use only.
|
| 87 |
+
License: Attribution-NonCommercial 4.0 International It should abide by the policy of OpenAI: https://openai.com/policies/terms-of-use
|
| 88 |
+
|
| 89 |
+
## 💖 Acknowledgement
|
| 90 |
+
* Special thanks to Ning Zhang, Siqi Wei, Kai Xiong, Kun Chen and colleagues at HiThink Research's data team for their support in building BizFinBench.v2.
|
| 91 |
+
|