metadata
license: apache-2.0
task_categories:
- question-answering
language:
- zh
- en
pretty_name: e
configs:
- config_name: cn
data_files:
- split: anomaly_information_tracing
path: cn/anomaly_information_tracing_cn.jsonl
- split: conterfactual
path: cn/conterfactual_cn.jsonl
- split: event_logic_reasoning
path: cn/event_logic_reasoning_cn.jsonl
- split: financial_data_description
path: cn/financial_data_description_cn.jsonl
- split: financial_multi_turn_perception
path: cn/financial_multi-turn_perception_cn.jsonl
- split: financial_quantitative_computation
path: cn/financial_quantitative_computation_cn.jsonl
- split: financial_report_analysis
path: cn/financial_report_analysis.jsonl
- split: stock_price_predict
path: cn/stock_price_predict_cn.jsonl
- split: user_sentiment_analysis
path: cn/user_sentiment_analysis_cn.jsonl
- config_name: en
data_files:
- split: anomaly_information_tracing
path: en/anomaly_information_tracing_en.jsonl
- split: conterfactual
path: en/conterfactual_en.jsonl
- split: event_logic_reasoning
path: en/event_logic_reasoning_en.jsonl
- split: financial_data_description
path: en/financial_data_description_en.jsonl
- split: financial_multi_turn_perception
path: en/financial_multi-turn_perception_en.jsonl
- split: financial_quantitative_computation
path: en/financial_quantitative_computation_en.jsonl
- split: stock_price_predict
path: en/stock_price_predict_en.jsonl
- split: user_sentiment_analysis
path: en/user_sentiment_analysis_en.jsonl
BizFinBench.v2: A Unified Dual-Mode Bilingual Benchmark for Expert-Level Financial Capability Alignment
BizFinBench.v2 is the secend release of 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.
🌟 Key Features
- Authentic & Real-Time: 100% derived from real financial platform queries, integrating online assessment capabilities.
- Expert-Level Difficulty: A challenging dataset of 29,578 Q&A pairs requiring professional financial reasoning.
- Comprehensive Coverage: Spans 4 core business scenarios, 8 fundamental tasks, and 2 online tasks.
📊 Key Findings
- High Difficulty: Even ChatGPT-5 achieves only 61.5% accuracy on main tasks, highlighting a significant gap vs. human experts.
- 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%.
📢 News
- 🚀 [06/01/2026] TBD
📕 Data Distrubution
BizFinBench.v2 contains multiple subtasks, each focusing on a different financial understanding and reasoning ability, as follows:
Distribution Visualization
Detailed Statistics
| Scenarios | Tasks | Avg. Input Tokens | # Questions |
|---|---|---|---|
| Business Information Provenance | Anomaly Information Tracing | 8,679 | 4,000 |
| Financial Multi-turn Perception | 10,361 | 3,741 | |
| Financial Data Description | 3,577 | 3,837 | |
| Financial Logic Reasoning | Financial Quantitative Computation | 1,984 | 2,000 |
| Event Logic Reasoning | 437 | 4,000 | |
| Counterfactual Inference | 2,267 | 2,000 | |
| Stakeholder Feature Perception | User Sentiment Analysis | 3,326 | 4,000 |
| Financial Report Analysis | 19,681 | 2,000 | |
| Real-time Market Discernment | Stock Price Prediction | 5,510 | 4,000 |
| Portfolio Asset Allocation | — | — | |
| Total | — | — | 29,578 |
✒️Citation
Coming Soon
📄 License
Usage and License Notices: The data and code are intended and licensed for research use only.
License: Attribution-NonCommercial 4.0 International It should abide by the policy of OpenAI: https://openai.com/policies/terms-of-use
💖 Acknowledgement
- 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.