Report Industry Investment Rating The document does not provide the industry investment rating. Core Viewpoints of the Report The report, as a continuation of AI - empowered bond market investment research, focuses on the current application of AI technology in the bond market and vertical large - models in the frontier fixed - income field. It details AI applications in bond investment research, such as curve construction, investment research process optimization, and structured product pricing. Future reports will cover the practical application of quantitative means in the bond market [1]. Summary by Relevant Catalogs 1. Introduction In 2025, with the popularity of DeepSeek, AI represented by large language models has evolved rapidly, changing the research and practice paradigms in the financial market. In the fixed - income and asset allocation fields, AI introduction has more challenges and value due to the large market capacity, diverse tools, and complex trading chains. Traditional fixed - income investment methods have limitations, and large - model technology can help market participants break information barriers and improve research depth and decision - making efficiency [11]. 2. Current Development Trends of Large Models In 2025, large - model development trends are "flagship - oriented, ecological, and embedded". Flagship models like GPT - 5, Claude 4, Gemini 2.0, and Llama 4 have become mature products. The ecological trend shows parallel open - source and closed - source paths. The embedded trend is reflected in models like BondGPT, which have penetrated the whole process of investment research, trading, and risk control. For the bond market, fixed - income vertical models like BondGPT Intelligence can directly embed generative AI into bond trading, promoting the shift from "human - machine separation" to "human - machine collaboration" [13][18]. 3. Application of AI Large Models in Fixed - Income Investment BlackRock Aladdin, a global leading asset management platform, has entered the "production - level implementation" stage. In investment research, it can process non - structured text information, extract key information, and generate summaries. In investment portfolio construction and rebalancing, it can generate scenario analyses and optimization tools. In trading execution, it scores and ranks bond market liquidity, improving trading efficiency. In risk control, it can detect potential risks and generate reports. The development path of BlackRock Aladdin provides a paradigm for other financial institutions, and the future Aladdin may become an AI - driven investment operating system [19][30]. 4. Vertical Large Models in Fixed - Income and Asset Allocation Fields - **BondGPT**: Driven by GPT - 4 and bond & liquidity data from LTX, it is used for pre - trading analysis of corporate bonds, including credit spread analysis and natural language queries for illiquid securities. It can assist in key pricing decisions, etc., with advantages such as instant information access, an intuitive user interface, and fast result return, and it can increase transaction file processing speed by 40% [32]. - **BondGPT+**: As an enterprise - level version of BondGPT, it allows customers to integrate local and third - party data, provides various deployment methods and API suites, and can be embedded in enterprise applications. It has functions like real - time liquidity pool analysis and automatic RFQ response, significantly improving the matching efficiency between dealers and customers [35]. 5. Implemented AI Applications in Fixed - Income and Asset Allocation Fields - **Curve Building**: It transforms discrete market quotes into continuous and interpolatable discount/forward curves. Generative AI has brought significant changes to traditional interest - rate modeling, with AI - based models showing better accuracy and adaptability than traditional methods. For example, a new deep - learning framework has 12% higher accuracy than the Nelson - Siegel model, and the error of the improved Libor model for 1 - 10 - year term interest rates is less than 0.5% [40]. - **Reshaping the Bond Investment Research Ecosystem**: Large language models and generative AI are reshaping the fixed - income investment research ecosystem. In trading, they provide natural - language interfaces and generation capabilities for bond analysis. They can summarize market data, policies, and research. For example, they can conduct sentiment analysis, generate summaries, and complete bond analysis tasks. BondGPT+ can improve trading counter - party matching efficiency by 25% [41]. - **ABS, MBS, Structured Products**: In structured product markets, AI - driven valuation frameworks can achieve automated cash - flow analysis, improve prepayment speed prediction accuracy by 10 - 20%, and reduce pricing errors of complex CMO tranches. Generative AI can simulate over 10,000 housing market scenarios, predict default rates with 89% accuracy, and help investors optimize portfolios and strategies [44][45].