feat: add complete English version of the wiki (270 pages)

docs/en/FAQ.md

@@ -0,0 +1,40 @@
+# F.A.Q
+
+This page addresses frequently asked questions about quantitative investing and the Quant-Wiki open-source project.
+
+---
+
+## Project Questions
+
+**Q: Why did you create the Quant-Wiki project?**
+**A**: Quant-Wiki was created to help learners who lack access to quality quantitative finance resources. Our goal is to contribute to the popularization and development of quantitative finance.
+
+---
+
+**Q: How can I get involved?**
+**A**: Quant-Wiki is hosted on GitHub. Visit our [repository](https://github.com/LLMQuant/Quant-Wiki) and participate by submitting Issues or Pull Requests, sharing ideas in discussion groups, or helping promote the project.
+
+---
+
+**Q: I feel my skills are limited. What can I do?**
+**A**: Every contribution matters — reviewing content, fixing minor errors, or organizing documentation. No contribution is too small.
+
+---
+
+**Q: Who maintains this project?**
+**A**: A group of core maintainers from quantitative finance and technology oversees the project, but we always welcome new contributors.
+
+---
+
+**Q: How do you ensure content sustainability?**
+**A**: All content is hosted on GitHub with regular backups to protect contributions long-term.
+
+---
+
+**Q: There are many content gaps!**
+**A**: True — we actively seek contributors to help fill them. See our [How to Contribute](contribute.md) page.
+
+---
+
+**Q: Why not just update Wikipedia?**
+**A**: We aim to provide content specifically tailored to quantitative finance learners — more focused and practical than general-purpose platforms.

docs/en/about.md

@@ -0,0 +1,5 @@
+![](https://fastly.jsdelivr.net/gh/bucketio/img11@main/2024/10/21/1729466068183-23134fce-3131-4262-b18c-f378d71af4f6.gif)
+
+LLMQuant is a cutting-edge community of researchers and practitioners from the world's leading universities and quantitative finance firms, dedicated to exploring the intersection of artificial intelligence (AI) and quantitative finance. Our members hail from Cambridge, Oxford, Harvard, ETH Zurich, Peking University, USTC, and other top institutions, with advisors from Microsoft, HSBC, Citadel, Man Group, Citi, Jump Trading, and leading proprietary trading firms.
+
+![alt text](asset/image.png)

docs/en/advanced/index.md

@@ -0,0 +1,85 @@
+# Frontiers in Quantitative Finance
+
+![](https://fastly.jsdelivr.net/gh/bucketio/img11@main/2024/10/21/1729466068183-23134fce-3131-4262-b18c-f378d71af4f6.gif)
+
+## Introduction
+
+This section features cutting-edge research in quantitative investing, including the latest technology developments and research reports from major financial institutions.
+
+## Latest Research
+
+### AI Applications in Quantitative Investing
+
+- [Balyasny Builds a Dedicated AI Team](./最新技术/chatgpt-balyasny.md) - How a hedge fund giant is deploying AI for quantitative finance
+- [Using LLMs to Uncover Hidden Bad News in Annual Reports](./最新技术/llm-report.md) - Applications of large language models in financial filing analysis
+- [RD-Agent: An AI Automation Tool Revolutionizing Quantitative Trading](./最新技术/rd-agent.md) - AI agents in quantitative trading
+- [Advanced Portfolio Management](./最新技术/advanced-portfolio-management.md) - A quant's guide to portfolio management
+
+### Selected Research Reports
+
+- [Multi-Factor Series](./研报精选/index.md#多因子系列) - Multi-factor model research from leading brokerages
+  - Factor mining and construction
+  - Factor efficacy studies
+  - Portfolio optimization methods
+  - Risk model construction
+
+- [AI Series](./研报精选/index.md#人工智能系列) - AI applications in quantitative investing
+  - Machine learning models
+  - Deep learning applications
+  - NLP and alternative data
+  - AI-driven stock selection strategies
+
+- [High-Frequency Trading Series](./研报精选/index.md#高频交易系列) - High-frequency trading strategy research
+  - Market microstructure
+  - High-frequency factor mining
+  - Trading strategy design
+  - Live trading experience summaries
+
+## Research Topics
+
+### 1. Machine Learning Strategies
+
+- Deep learning prediction
+- Reinforcement learning for trading
+- NLP sentiment analysis
+- Factor mining
+
+### 2. High-Frequency Trading Strategies
+
+- Market making strategies
+- Statistical arbitrage
+- Latency arbitrage
+- Order flow prediction
+
+### 3. Alternative Data Strategies
+
+- Satellite imagery analysis
+- Social media mining
+- Alternative data applications
+- Web scraping
+
+### 4. Multi-Factor Strategies
+
+- Factor construction
+- Factor combination
+- Risk models
+- Portfolio optimization
+
+## Recommended Approach
+
+1. Start by browsing the latest research to understand current industry trends
+2. Choose specific topics based on your interests for deeper study
+3. Combine research report content with practical case studies for deeper understanding
+
+## Update Schedule
+
+We continuously track:
+
+1. Technological innovations at top hedge funds
+2. The latest academic research findings
+3. In-depth research reports from major financial institutions
+4. Industry best practices and lessons learned
+
+## About LLMQuant
+
+LLMQuant is a cutting-edge community of professionals from the world's leading universities and quantitative finance practitioners, dedicated to exploring the intersection of artificial intelligence (AI) and quantitative finance (Quant). Our team members come from Cambridge, Oxford, Harvard, ETH Zurich, Peking University, USTC, and other prestigious institutions, with external advisors from Microsoft, HSBC, Citadel, Man Group, Citi, Jump Trading, and leading proprietary trading firms.

docs/en/advanced/最新技术/advanced-portfolio-management.md

@@ -0,0 +1,124 @@
+
+## Advanced Portfolio Management: A Quant's Guide for Fundamental Investors
+
+### About the Author
+
+The author is a leading quantitative investment expert with extensive experience in risk management and portfolio construction. His career spans multiple major hedge funds, and his research and practice have provided invaluable portfolio construction guidance for countless fund managers.
+
+![ ](https://m.media-amazon.com/images/I/71Ki2qadzPL._AC_UF1000,1000_QL80_.jpg)
+
+---
+
+### Chapter 1: Who Is This Book For? Why Was It Written? How Should You Read It?
+
+This book is primarily aimed at portfolio managers and quantitative researchers who want to develop a deeper understanding of quantitative investing. Whether you are a seasoned investor or a quantitative researcher, the book provides a complete path from fundamental theory to practical application. Giuseppe emphasizes that while the book involves considerable mathematical concepts, understanding these tools is essential for portfolio management. He encourages readers to study the sections marked with a star symbol carefully for deeper technical insight.
+
+---
+
+### Chapter 2: From Ideas to Profits
+
+The central challenge of quantitative investing is converting theoretical concepts into actual profits. This chapter discusses core investment questions such as how to invest based on your edge and reduce uncertainty through hedging. Giuseppe also explores how to effectively leverage data analysis to optimize investment decisions.
+
+Key formula:
+$$ R(t) = \alpha + \beta_m \cdot M(t) + \epsilon(t) $$
+This formula expresses asset returns as a function of the market factor return $$ M(t) $$, the asset's market risk exposure $$ \beta_m $$, and idiosyncratic risk $$ \epsilon(t) $$.
+
+---
+
+### Chapter 3: A Tour of Risk and Performance
+
+Factor models are among the most widely used tools in quantitative investing for decomposing risk and attributing performance. This chapter introduces how to use a simple single-factor model to explain asset returns using the market factor.
+
+Core formula:
+$$ r_i = \alpha_i + \beta_i \cdot F + \epsilon_i $$
+
+Here, $$ \beta_i $$ is the asset's exposure to the factor, $$ F $$ is the factor return, and $$ \epsilon_i $$ is the asset's idiosyncratic return. The chapter further demonstrates how to estimate $$ \alpha $$ and $$ \beta $$ through regression analysis, thereby separating systematic risk from idiosyncratic risk.
+
+---
+
+### Chapter 4: Introduction to Multi-Factor Models
+
+Single-factor models have limitations because they cannot account for multiple market drivers. This chapter introduces multi-factor models, extending the risk decomposition framework.
+
+Formula:
+
+$$ r_i = \alpha + \beta_1 f_1 + \beta_2 f_2 + ... + \beta_k f_k + \epsilon $$
+
+With multiple factors, investors can more precisely measure the risk and return characteristics of different assets. The chapter also discusses in detail how to compute asset exposures across multiple factors through factor regression.
+
+---
+
+### Chapter 5: Understanding Factors
+
+Factor models help investors understand the primary drivers in markets, including macroeconomic factors, sector factors, and valuation factors. This chapter provides an in-depth analysis of how these factors affect asset returns and demonstrates how to manage risk by adjusting factor exposures within a portfolio.
+
+Formula:
+$$ F_i = \frac{E[R_i] - R_f}{\beta_i} $$
+
+Using this formula, investors can estimate the contribution of each factor to returns, enabling portfolio optimization.
+
+---
+
+### Chapter 6: Effective Alpha Sizing Strategies
+
+This chapter focuses on how to size alpha positions based on risk. The Sharpe ratio is the standard metric for measuring risk-adjusted portfolio returns:
+
+$$ Sharpe = \frac{E[R_p - R_f]}{\sigma_p} $$
+
+Through risk adjustment, investors can maximize returns for a given level of risk. The chapter also discusses position sizing strategies based on historical data, helping readers find the optimal balance during market volatility.
+
+---
+
+### Chapter 7: Factor Risk Management
+
+Factor risk management is a core component of quantitative investing. This chapter discusses how to control factor risk by setting market exposure limits and single-stock position limits. Through these risk management tools, investors can reduce non-systematic risk in their portfolios.
+
+Additionally, the chapter introduces how to reduce the impact of market volatility through strategic and tactical factor adjustments.
+
+Formula:
+$$ V = w' \Sigma w $$
+
+This formula calculates portfolio risk exposure $$ V $$, where $$ \Sigma $$ is the covariance matrix and $$ w $$ represents portfolio weights.
+
+![ ](https://tradingalpha.io/wp-content/uploads/2022/01/Trading-Alpha-Logo-1.png)
+
+---
+
+### Chapter 8: Understanding Your Performance
+
+Performance attribution is an essential tool for understanding the sources of investment returns. This chapter explains how factor analysis helps investors understand where their returns come from. Through regression analysis and attribution models, investors can identify which factors contribute most to portfolio returns.
+
+Formula:
+$$ Return = \Sigma (\beta_i \cdot f_i) + \alpha $$
+
+This formula shows the factor decomposition of portfolio returns, helping investors optimize their strategies.
+
+---
+
+### Chapter 9: Managing Losses
+
+Stop-loss strategies are critical in risk management. This chapter provides a detailed guide to setting effective stop-loss levels, along with an analysis of the costs and benefits of these strategies.
+
+Formula:
+$$ Loss = max(R - Threshold, 0) $$
+
+This formula describes the loss calculation, where a loss is triggered when the asset return falls below a specified threshold. The chapter emphasizes the importance of stop-loss strategies in risk management, particularly during extreme market volatility.
+
+---
+
+### Chapter 10: Setting Sustainable Leverage Ratios
+
+Leverage is a powerful tool for amplifying both returns and risk. This chapter explores how to set appropriate leverage ratios to ensure long-term business sustainability. Giuseppe provides a leverage decision framework based on volatility and returns, helping investors balance reward and risk.
+
+Formula:
+$$ Leverage = \frac{\sigma_{target}}{\sigma_{portfolio}} $$
+
+This formula helps investors determine portfolio leverage multiples for optimizing risk-adjusted returns.
+
+---
+
+### Appendix: Key Risk Model Formulas
+
+The appendix provides multiple mathematical formulas including factor models, variance minimization, and mean-variance optimization. These tools help investors better understand and apply sophisticated quantitative investment strategies.
+
+---

docs/en/advanced/最新技术/chatgpt-balyasny.md

@@ -0,0 +1,49 @@
+![](https://fastly.jsdelivr.net/gh/bucketio/img11@main/2024/10/21/1729466068183-23134fce-3131-4262-b18c-f378d71af4f6.gif)
+
+# Hedge Fund AI Outperforms ChatGPT? Balyasny Builds a Dedicated AI Team
+
+![](https://fastly.jsdelivr.net/gh/bucketio/img9@main/2024/10/20/1729465031968-b3c8959e-1d37-4b8a-91b1-b0b0dfe25143.png)
+
+In the current AI landscape, Large Language Models (LLMs) have become a transformative force in information processing and decision-making workflows. However, a model's training data alone cannot always meet the high-quality information demands of complex applications. This is where Retrieval Augmented Generation (RAG) comes in. In simple terms, RAG provides LLMs with an "external brain" -- when answering complex questions, the system can retrieve and incorporate external data in real time to deliver more targeted responses. For the financial industry, this technology is both highly relevant and challenging, since high-quality financial intelligence is difficult to obtain and requires specialized interpretation.
+
+## RAG in Finance: Challenges and Opportunities
+
+In financial investing, timeliness, precision, and domain expertise are paramount. A hedge fund researcher or trader must rapidly assess the direction of stocks, bonds, or derivatives and estimate the potential impact of major events -- policy shifts, geopolitical conflicts, industry transformations. These decisions require not only the model's built-in general knowledge and historical data, but also the latest financial news, research reports, earnings filings, and regulatory information.
+
+![](https://fastly.jsdelivr.net/gh/bucketio/img10@main/2024/12/19/1734652285742-c535017a-a9bd-40a7-9cdf-673e08067855.png)
+
+Conventional LLMs are trained on general-purpose text corpora. While they broadly understand human language and possess general knowledge, they may lack sensitivity to industry-specific jargon or specialized data structures. Hedge fund managers and data scientists cannot fully rely on a general-purpose model that, despite having knowledge gaps, is expected to deliver on extremely high standards of timeliness and expertise. RAG thus becomes an ideal solution: by embedding external information retrieval results into the LLM's context, the model can draw on the latest, more specialized external data sources when answering questions.
+
+## Balyasny's AI Team: Recruiting from Google and DeepMind to Build Proprietary Tools
+
+Among the institutions applying RAG to finance, hedge fund Balyasny Asset Management stands out as a pioneer. The fund has recruited senior professionals from Google and DeepMind to form an applied AI research team dedicated to developing internal AI tools. Their flagship product is **BAMChatGPT** -- a chatbot custom-built for the firm's traders and analysts.
+
+According to Michal Mucha, the engineer leading the tool's development, these AI tools are now used by approximately 80% of Balyasny's employees. Whether someone needs an instant update on a portfolio holding or wants to assess how a geopolitical event might impact the fund's positions, BAMChatGPT can provide valuable preliminary guidance. While these tools cannot replace human expert judgment, they significantly enhance research and decision-making efficiency.
+
+## BAM Embeddings: Custom-Built for Financial Jargon
+
+Earlier this month, Balyasny published an academic paper introducing **BAM Embeddings**. To understand the significance of this technology, it helps to know the role that embeddings play in the RAG pipeline. For LLMs, embeddings convert text into computable vector representations, enabling the model to understand text semantics mathematically. When an LLM needs to draw on external materials, those materials must first be vectorized through a specialized embedding model so that the most relevant information fragments can be efficiently retrieved and matched to the user's query.
+
+However, general-purpose embedding models often struggle with industry-specific terminology. Finance is filled with abbreviations, complex jargon, and specialized expressions that appear infrequently in general corpora, leaving models with limited comprehension. BAM Embeddings address this gap by training specifically on financial industry "jargon," enabling the embedding model to better understand and match relevant information.
+
+Preliminary results are impressive. When researchers asked an LLM (in this case, the Mistral 7B Instruct model) to search a set of financial documents for the most relevant passages, BAM Embeddings returned the optimal fragment over 60% of the time, compared to less than 40% with OpenAI's general-purpose embeddings. On the public FinanceBench benchmark, BAM Embeddings achieved 55% query accuracy versus 47% for OpenAI's ada-002 embeddings. These results demonstrate that specialized embeddings can significantly improve information retrieval accuracy in financial contexts.
+
+## Imperfect Realities: Hallucinations and Residual Risks
+
+While encouraging, these results do not mean Balyasny's approach is flawless. In FinanceBench testing, responses using BAM Embeddings still had an approximately 30% error rate. This indicates that even with RAG, the LLM "hallucination" problem persists -- the model can still produce fabricated or illogical answers.
+
+Research shows that hallucination remains prevalent across current LLM applications. A July study found that GPT-3.5, widely used in algorithmic trading, still exhibited a 27.8% hallucination rate when using RAG. Balyasny's BAM Embeddings were trained on 14.3 million synthetic queries. While this massive synthetic dataset can strengthen domain-specific understanding, some experts worry it may introduce biases or context-specific misinformation, making it difficult to further reduce hallucination rates.
+
+## Industry Trends: The Rise of RAG in Enterprise Applications
+
+Despite these limitations, Balyasny's initiative remains forward-looking and competitive. In enterprise applications, RAG has already become a mainstream trend. According to a report by venture capital firm Menlo Ventures, RAG will serve as the primary AI architecture for 51% of enterprises in 2024, up from 31% the previous year. For a hedge fund that needs to respond rapidly to market changes and information demands, early adoption of this cutting-edge technology provides a significant competitive advantage.
+
+## Looking Ahead: Specialization and Customization
+
+As more enterprises incorporate RAG into their core strategies, general-purpose LLMs and embedding solutions may prove insufficient for highly specialized domains. Balyasny's case illustrates a likely future trend: the rise of industry-customized AI tools. This goes beyond algorithmic optimization -- it represents a deep integration of data, models, and application contexts.
+
+In this evolution, we may see more proprietary embedding models and LLMs trained by financial institutions, delivering higher precision, greater timeliness, and improved interpretability in understanding financial terminology, capturing market signals, and analyzing macroeconomic variables. As Balyasny's ongoing experiment demonstrates, continuous optimization and iteration have the potential to bring new insights and value to the broader financial AI ecosystem.
+
+## About LLMQuant
+
+LLMQuant is a cutting-edge community of professionals from the world's leading universities and quantitative finance practitioners, dedicated to exploring the intersection of artificial intelligence (AI) and quantitative finance (Quant). Our team members come from Cambridge, Oxford, Harvard, ETH Zurich, Peking University, USTC, and other prestigious institutions, with external advisors from Microsoft, HSBC, Citadel, Man Group, Citi, Jump Trading, and leading proprietary trading firms.

docs/en/advanced/最新技术/index.md

@@ -0,0 +1 @@
+