File size: 15,750 Bytes
a086cd4 677e898 a086cd4 9800fde 677e898 a617f35 677e898 a086cd4 677e898 a086cd4 8d1471a ad5a21b e220fe2 a086cd4 5549287 a086cd4 5549287 2aba161 677e898 2a972f2 677e898 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 |
---
language:
- zh
license: cc-by-4.0
size_categories:
- 10K<n<100K
task_categories:
- question-answering
- text-classification
pretty_name: BizFinBench
tags:
- finance
configs:
- config_name: Anomalous_Event_Attribution
data_files:
- split: test
path: Anomalous_Event_Attribution/*.jsonl
- config_name: Emotion_Recognition
data_files:
- split: test
path: Emotion_Recognition/*.jsonl
- config_name: Financial_Data_Description
data_files:
- split: test
path: Financial_Data_Description/*.jsonl
- config_name: Financial_Knowledge_QA
data_files:
- split: test
path: Financial_Knowledge_QA/*.jsonl
- config_name: Financial_Named_Entity_Recognition
data_files:
- split: test
path: Financial_Named_Entity_Recognition/*.jsonl
- config_name: Financial_Numerical_Computation
data_files:
- split: test
path: Financial_Numerical_Computation/*.jsonl
- config_name: Financial_Time_Reasoning
data_files:
- split: test
path: Financial_Time_Reasoning/*.jsonl
- config_name: Financial_Tool_Usage
data_files:
- split: test
path: Financial_Tool_Usage/*.jsonl
- config_name: Stock_Price_Prediction
data_files:
- split: test
path: Stock_Price_Prediction/*.jsonl
---
# BizFinBench: A Business-Driven Real-World Financial Benchmark for Evaluating LLMs
📖<a href="https://arxiv.org/abs/2505.19457">Paper</a> |🐙<a href="https://github.com/HiThink-Research/BizFinBench/">Github</a></h3>|🤗<a href="https://huggingface.co/datasets/HiThink-Research/BizFinBench">Huggingface</a></h3>
Large language models excel in general tasks, yet assessing their reliability in logic‑heavy, precision‑critical domains like finance, law, and healthcare remains challenging. To address this, we introduce BizFinBench, the first benchmark specifically designed to evaluate LLMs in real-world financial applications. BizFinBench comprises over 100,000+ bilingual (English & Chinese) financial questions, each rooted in real-world business scenarios. The first public release, BizFinBench.v1, delivers 6,781 well annotated Chinese queries, spanning five dimensions: numerical calculation, reasoning, information extraction, prediction recognition, and knowledge-based question answering, grouped into nine fine-grained categories. The benchmark includes both objective and subjective metrics. We also introduce Iterajudge, a novel LLM evaluation method that reduces bias when LLMs serve as evaluators in objective metrics. We benchmark 25 models, including both proprietary and open-source systems. Extensive experiments show that no model dominates across all tasks. Our evaluation reveals distinct capability patterns: (1) In Numerical Calculation, Claude-3.5-Sonnet (63.18) and DeepSeek-R1 (64.04) lead, while smaller models like Qwen2.5-VL-3B (15.92) lag significantly; (2) In Reasoning, proprietary models dominate (ChatGPT-o3: 83.58, Gemini-2.0-Flash: 81.15), with open-source models trailing by up to 19.49 points; (3) In Information Extraction, the performance spread is the largest, with DeepSeek-R1 scoring 71.46, while Qwen3-1.7B scores 11.23; (4) In Prediction Recognition, performance variance is minimal, with top models scoring between 39.16 and 50.00. We find that while current LLMs handle routine finance queries competently, they struggle with complex scenarios requiring cross-concept reasoning. BizFinBench offers a rigorous, business-aligned benchmark for future research. The code and dataset are available at https://github.com/HiThink-Research/BizFinBench.
## 📢 News
- 🚀 [16/05/2025] We released <strong>BizFinBench</strong> benchmark (V1), the first benchmark grounded in real-world financial applications.
## 💡 Highlights
- 🔥 **Benchmark:** We propose **BizFinBench**, the first evaluation benchmark in the financial domain that integrates business-oriented tasks, covering 5 dimensions and 9 categories. It is designed to assess the capacity of LLMs in real-world financial scenarios.
- 🔥 **Judge model:** We design a novel evaluation method, i.e., **Iterajudge**, which enhances the capability of LLMs as a judge by refining their decision boundaries in specific financial evaluation tasks.
- 🔥 **key insights:** We conduct a comprehensive evaluation with **25 LLMs** based on BizFinBench, uncovering key insights into their strengths and limitations in financial applications.
## 📕 Data Distrubution
This dataset contains multiple subtasks, each focusing on a different financial understanding and reasoning ability, as follows:
| Dataset | Description | Evaluation Dimensions | Volume |
| -------------------------------------- | ------------------------------------------------------------ | ------------------------------------------------------------ | ------ |
| **Anomalous Event Attribution** | A financial anomaly attribution evaluation dataset assessing models' ability to trace stock fluctuations based on given information (e.g., timestamps, news articles, financial reports, and stock movements). | Causal consistency, information relevance, noise resistance | 1,064 |
| **Financial Numerical Computation** | A financial numerical computation dataset evaluating models' ability to perform accurate numerical calculations in financial scenarios, including interest rate calculations, gain/loss computations, etc. | Calculation accuracy, unit consistency | 581 |
| **Financial Time Reasoning** | A financial temporal reasoning evaluation dataset assessing models' ability to comprehend and reason about time-based financial events, such as "the previous trading day" or "the first trading day of the year." | Temporal reasoning correctness | 514 |
| **Financial Data Description** | A financial data description evaluation dataset measuring models' ability to analyze and describe structured/unstructured financial data, e.g., "the stock price first rose to XX before falling to XX." | Trend accuracy, data consistency | 1,461 |
| **Stock Price Prediction** | A stock price movement prediction dataset evaluating models' ability to forecast future stock price trends based on historical data, financial indicators, and market news. | Trend judgment, causal rationality | 497 |
| **Financial Named Entity Recognition** | A financial named entity recognition dataset assessing models' ability to identify entities (Person, Organization, Market, Location, Financial Products, Date/Time) in short/long financial news. | Recognition accuracy, entity category correctness | 433 |
| **Emotion_Recognition** | A financial sentiment recognition dataset evaluating models' ability to discern nuanced user emotions in complex financial market environments. Inputs include multi-dimensional data such as market conditions, news, research reports, user holdings, and queries, covering six emotion categories: optimism, anxiety, pessimism, excitement, calmness, and regret. | Emotion classification accuracy, implicit information extraction and reasoning correctness | 600 |
| **Financial Tool Usage** | A financial tool usage dataset evaluating models' ability to understand user queries and appropriately utilize various financial tools (investment analysis, market research, information retrieval, etc.) to solve real-world problems. Tools include calculators, financial encyclopedia queries, search engines, data queries, news queries, economic calendars, and company lookups. Models must accurately interpret user intent, select appropriate tools, input correct parameters, and coordinate multiple tools when necessary. | Tool selection rationality, parameter input accuracy, multi-tool coordination capability | 641 |
| **Financial Knowledge QA** | A financial encyclopedia QA dataset assessing models' understanding and response accuracy regarding core financial knowledge, covering key domains: financial fundamentals, markets, investment theories, macroeconomics, etc. | Query comprehension accuracy, knowledge coverage breadth, answer accuracy and professionalism | 990 |
## 🛠️ Usage
Please clone the github [link](https://github.com/HiThink-Research/BizFinBench/) to start evaluation
### Quick Start – Evaluate a Local Model
```sh
export MODEL_PATH=model/Qwen2.5-0.5B # Path to the model to be evaluated
export REMOTE_MODEL_PORT=16668
export REMOTE_MODEL_URL=http://127.0.0.1:${REMOTE_MODEL_PORT}/model
export MODEL_NAME=Qwen2.5-0.5B
export PROMPT_TYPE=chat_template # Hithink llama3 llama2 none qwen chat_template; chat_template is recommended
# First start the model as a service
python inference/predict_multi_gpu.py \
--model ${MODEL_PATH} \
--server_port ${REMOTE_MODEL_PORT} \
--prompt ${PROMPT_TYPE} \
--preprocess preprocess \
--run_forever \
--max_new_tokens 4096 \
--tensor_parallel ${TENSOR_PARALLEL} &
# Pass in the config file path to start evaluation
python run.py --config config/offical/eval_fin_eval_diamond.yaml --model_name ${MODEL_NAME}
```
### Quick Start – Evaluate a Local Model and Score with a Judge Model
```sh
export MODEL_PATH=model/Qwen2.5-0.5B # Path to the model to be evaluated
export REMOTE_MODEL_PORT=16668
export REMOTE_MODEL_URL=http://127.0.0.1:${REMOTE_MODEL_PORT}/model
export MODEL_NAME=Qwen2.5-0.5B
export PROMPT_TYPE=chat_template # llama3 llama2 none qwen chat_template; chat_template is recommended
# First start the model as a service
python inference/predict_multi_gpu.py \
--model ${MODEL_PATH} \
--server_port ${REMOTE_MODEL_PORT} \
--prompt ${PROMPT_TYPE} \
--preprocess preprocess \
--run_forever \
--max_new_tokens 4096 \
--tensor_parallel ${TENSOR_PARALLEL} \
--low_vram &
# Start the judge model
export JUDGE_MODEL_PATH=/mnt/data/llm/models/base/Qwen2.5-7B
export JUDGE_TENSOR_PARALLEL=1
export JUDGE_MODEL_PORT=16667
python inference/predict_multi_gpu.py \
--model ${JUDGE_MODEL_PATH} \
--server_port ${JUDGE_MODEL_PORT} \
--prompt chat_template \
--preprocess preprocess \
--run_forever \
--manual_start \
--max_new_tokens 4096 \
--tensor_parallel ${JUDGE_TENSOR_PARALLEL} \
--low_vram &
# Pass in the config file path to start evaluation
python run.py --config "config/offical/eval_fin_eval.yaml" --model_name ${MODEL_NAME}
```
> **Note**: Add the `--manual_start` argument when launching the judge model, because the judge must wait until the main model finishes inference before starting (this is handled automatically by the `maybe_start_judge_model` function in `run.py`).
## ✒️Results
The models are evaluated across multiple tasks, with results color-coded to represent the top three performers for each task:
- 🥇 indicates the top-performing model.
- 🥈 represents the second-best result.
- 🥉 denotes the third-best performance.
| Model | AEA | FNC | FTR | FTU | FQA | FDD | ER | SP | FNER | Average |
| ---------------------------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- | ------- |
| **Proprietary LLMs** | | | | | | | | | | |
| ChatGPT-o3 | 🥈 86.23 | 61.30 | 🥈 75.36 | 🥇 89.15 | 🥈 91.25 | 🥉 98.55 | 🥉 44.48 | 53.27 | 65.13 | 🥇 73.86 |
| ChatGPT-o4-mini | 🥉 85.62 | 60.10 | 71.23 | 74.40 | 90.27 | 95.73 | 🥇 47.67 | 52.32 | 64.24 | 71.29 |
| GPT-4o | 79.42 | 56.51 | 🥇 76.20 | 82.37 | 87.79 | 🥇 98.84 | 🥈 45.33 | 54.33 | 65.37 | 🥉 71.80 |
| Gemini-2.0-Flash | 🥇 86.94 | 🥉 62.67 | 73.97 | 82.55 | 90.29 | 🥈 98.62 | 22.17 | 🥉 56.14 | 54.43 | 69.75 |
| Claude-3.5-Sonnet | 84.68 | 🥈 63.18 | 42.81 | 🥈 88.05 | 87.35 | 96.85 | 16.67 | 47.60 | 63.09 | 65.59 |
| **Open Source LLMs** | | | | | | | | | | |
| Qwen2.5-7B-Instruct | 73.87 | 32.88 | 39.38 | 79.03 | 83.34 | 78.93 | 37.50 | 51.91 | 30.31 | 56.35 |
| Qwen2.5-72B-Instruct | 69.27 | 54.28 | 70.72 | 85.29 | 87.79 | 97.43 | 35.33 | 55.13 | 54.02 | 67.70 |
| Qwen2.5-VL-3B | 53.85 | 15.92 | 17.29 | 8.95 | 81.60 | 59.44 | 39.50 | 52.49 | 21.57 | 38.96 |
| Qwen2.5-VL-7B | 73.87 | 32.71 | 40.24 | 77.85 | 83.94 | 77.41 | 38.83 | 51.91 | 33.40 | 56.68 |
| Qwen2.5-VL-14B | 37.12 | 41.44 | 53.08 | 82.07 | 84.23 | 7.97 | 37.33 | 54.93 | 47.47 | 49.52 |
| Qwen2.5-VL-32B | 76.79 | 50.00 | 62.16 | 83.57 | 85.30 | 95.95 | 40.50 | 54.93 | 🥉 68.36 | 68.62 |
| Qwen2.5-VL-72B | 69.55 | 54.11 | 69.86 | 85.18 | 87.37 | 97.34 | 35.00 | 54.94 | 54.41 | 67.53 |
| Qwen3-1.7B | 77.40 | 35.80 | 33.40 | 75.82 | 73.81 | 78.62 | 22.40 | 48.53 | 11.23 | 50.78 |
| Qwen3-4B | 83.60 | 47.40 | 50.00 | 78.19 | 82.24 | 80.16 | 42.20 | 50.51 | 25.19 | 59.94 |
| Qwen3-14B | 84.20 | 58.20 | 65.80 | 82.19 | 84.12 | 92.91 | 33.00 | 52.31 | 50.70 | 67.05 |
| Qwen3-32B | 83.80 | 59.60 | 64.60 | 85.12 | 85.43 | 95.37 | 39.00 | 52.26 | 49.19 | 68.26 |
| Xuanyuan3-70B | 12.14 | 19.69 | 15.41 | 80.89 | 86.51 | 83.90 | 29.83 | 52.62 | 37.33 | 46.48 |
| Llama-3.1-8B-Instruct | 73.12 | 22.09 | 2.91 | 77.42 | 76.18 | 69.09 | 29.00 | 54.21 | 36.56 | 48.95 |
| Llama-3.1-70B-Instruct | 16.26 | 34.25 | 56.34 | 80.64 | 79.97 | 86.90 | 33.33 | 🥇 62.16 | 45.95 | 55.09 |
| Llama 4 Scout | 73.60 | 45.80 | 44.20 | 85.02 | 85.21 | 92.32 | 25.60 | 55.76 | 43.00 | 61.17 |
| DeepSeek-V3 (671B) | 74.34 | 61.82 | 72.60 | 🥈 86.54 | 🥉 91.07 | 98.11 | 32.67 | 55.73 | 🥈 71.24 | 71.57 |
| DeepSeek-R1 (671B) | 80.36 | 🥇 64.04 | 🥉 75.00 | 81.96 | 🥇 91.44 | 98.41 | 39.67 | 55.13 | 🥇 71.46 | 🥈 73.05 |
| QwQ-32B | 84.02 | 52.91 | 64.90 | 84.81 | 89.60 | 94.20 | 34.50 | 🥈 56.68 | 30.27 | 65.77 |
| DeepSeek-R1-Distill-Qwen-14B | 71.33 | 44.35 | 16.95 | 81.96 | 85.52 | 92.81 | 39.50 | 50.20 | 52.76 | 59.49 |
| DeepSeek-R1-Distill-Qwen-32B | 73.68 | 51.20 | 50.86 | 83.27 | 87.54 | 97.81 | 41.50 | 53.92 | 56.80 | 66.29 |
## ✒️Citation
```
@article{lu2025bizfinbench,
title={BizFinBench: A Business-Driven Real-World Financial Benchmark for Evaluating LLMs},
author={Lu, Guilong and Guo, Xuntao and Zhang, Rongjunchen and Zhu, Wenqiao and Liu, Ji},
journal={arXiv preprint arXiv:2505.19457},
year={2025}
}
```
## 📄 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 |