Xuanyuan: The first 100-billion-level Chinese financial chat model
XuanYuan: A Large Chinese Financial Chat Model with Hundreds of Billions Parameters
## 1. Introduction to the model
Xuanyuan is the first open-source 100-billion-level Chinese dialogue model in China, and it is also the first 100-billion-level open source dialogue model optimized for the Chinese financial field. On the basis of BLOOM-176B, Xuanyuan has carried out targeted pre-training and fine-tuning for Chinese general fields and financial fields. It can not only deal with problems in general fields, but also answer various questions related to finance, providing users with accurate, Comprehensive financial information and advice.
## 2. Model download
The hundreds of billions of Xuanyuan models can be applied for download in Huggingface .
## 3. Model Evaluation
### 3.1. Financial Field Effect Evaluation Professionalism, we have built a detailed evaluation system for the financial field. The evaluation system involves four major areas: financial term understanding, financial market commentary, financial data analysis, and financial news understanding, and comprehensively examines the financial processing capabilities of the model from multiple perspectives. The evaluation results show that Xuanyuan has won 63.33% of the 150 answers in the comparison of the four mainstream open source large models, highlighting its significant advantages in the financial field.
[ 
——————————————————— \|
\| Please explain what is meant by "share repurchase" and why companies do share repurchases? \| Share repurchases ( Share buyback) means that a company repurchases its issued stock for cancellation or for future use. Share repurchases can reduce a company's share capital, thereby boosting dividends per share and earnings per share. There are many reasons why companies conduct share repurchases, here are some common ones: 1. To increase dividends per share and earnings per share: By repurchasing shares, companies can reduce their share capital, thereby increasing dividends per share and earnings per share. 2. Boosting the stock price: Share repurchases can reduce a company's share capital, which increases dividends and earnings per share, which in turn boosts the stock price. 3. Reduce debt: By repurchasing shares, a company can reduce debt and thus reduce interest expenses. 4. Increase shareholder returns: By repurchasing shares, companies can reduce their share capital, thereby increasing dividends per share and earnings per share, thereby increasing shareholder returns. 5. Improve the company's image: By repurchasing shares, the company can reduce its share capital, thereby increasing dividends per share and earnings per share, thereby improving the company's image. It should be noted that stock repurchase may have an impact on the company's financial status, so the company needs to carefully consider whether to carry out stock repurchase. \|
\| Can you explain what is a "credit line"? \| The credit line refers to the maximum loan amount provided by a bank or financial institution to a customer, that is, the loan amount that the customer can freely control within this line. Usually, the bank will determine the size of the credit line based on factors such as the customer's personal credit history, income, and liabilities. Credit line is one of the important indicators for banks and financial institutions to provide credit services to customers. The specific calculation method of the credit line will vary with different banks and financial institutions, but generally includes the following aspects: 1. Personal credit history: Banks and financial institutions usually consider the customer's credit history, including loan repayment records, credit card Usage conditions, etc. These factors will directly affect the probability of customers obtaining credit lines. 2. Income level: Banks and financial institutions usually require customers to provide a stable source of income, such as salary income, return on investment, etc. Banks and financial institutions determine a customer's credit limit by reviewing the customer's proof of income and financial reports. 3. Indebtedness: Banks and financial institutions usually keep an eye on the indebtedness of their customers, including other outstanding loans, credit card arrears, etc. Banks and financial institutions may limit the amount of credit extended to customers if they already have a high debt load. 4. Guarantee ability: If the customer does not have a good credit record or the income status is poor, banks and financial institutions may require the customer to provide guarantees, such as real estate, vehicles and other properties as guarantees, to increase the possibility of credit lines. In short, the line of credit is an important loan tool that can help customers respond to various occasions more flexibly when they need funds. \|
### 3.2. Evaluation of effects in the general field excellent. Using a general evaluation set consisting of 200 diverse questions, we evaluated Xuanyuan's performance in 13 main dimensions including mathematical calculation, scenario writing, logical reasoning, and text summarization, and compared it with ChatGPT. The results show that Xuanyuan performed as well as ChatGPT in 71% of the questions on this evaluation set, which verified its comprehensive capabilities.
## 4. Related papers
If Xuanyuan related methods and models are useful, please cite the following papers:
XuanYuan 2.0: A Large Chinese Financial Chat Model with Hundreds of Billions Parameters
Self-QA: Unsupervised Knowledge Guided Language Model Alignment
CGCE: A Chinese Generative Chatval E uation Benchmark for General and Financial Domains
## 5. Instructions for use
This model is recommended to run on 8-card A100 GPU or equivalent configuration for best performance.
You can call this model through the following code:
from
from transformers import AutoTokenizer, AutoModel
tokenizer = AutoTokenizer.from_pretrained("MODEL_NAME", trust_remote_code=True)
model = AutoModel.from_pretrained("MODEL_NAME", trust_remote_code=True)
【Popular Question] How to invoke the Xuanyuan model?
Since this model is large and does not support online API testing, please download the model and use the AutoTokenizer and AutoModel of the transformers library to call it. Input example of Xuanyuan dialogue model:
BOS_TOKEN + "Human: " + query + "\n\nAssistant: "
Xuanyuan dialogue model generation example:
output = model.generate(**input, do_sample=True, temperature=0.8, top_k=50, top_p=0.9, early_stopping
Visit Official Website
