bguzzo2k/tkrs_timeless_report

--- license: apache-2.0 --- # Timeless Synthetic Financial Asset Reports ## Dataset Summary A fully synthetic repository of financial instrument descriptions designed for Transformer-based trading systems. It provides a multi-dimensional conceptual framework for 700+ assets while **strictly eliminating look-ahead bias** by abstracting structural and behavioral characteristics away from specific dates and named historical events. **Dataset available on Hugging Face:** [bguzzo2k/tkrs_timeless_report](https://huggingface.co/datasets/bguzzo2k/tkrs_timeless_report) ## Key Features - **Zero Look-Ahead Bias:** Scrubbed of temporal anchors; historical behaviors are translated into abstract dynamics (e.g., "systemic credit crises"). - **Multi-Asset Coverage:** Over 700+ tickers spanning Global Indices, ETFs, Fixed Income, Commodities, FX, S&P 500, and Cryptocurrencies. - **15 Expert Perspectives:** Every asset is analyzed through 15 specialized "Personas" to ensure high-granularity and diverse embedding subspaces. ## Dataset Structure The dataset is provided in `Parquet` format with the following columns: | Column | Description | | :--- | :--- | | `ticker` | Financial instrument identifier. | | `gemini_master` | Foundational structural report (Grounding via `gemini-3.1-pro-preview`). | | `macroeconomist` | Inflation, interest rates, and monetary policy. | | `geopolitical_risk_analyst`| Conflicts, trade wars, and sanctions. | | `supply_chain_logistics` | Raw materials, bottlenecks, and labor markets. | | `quantitative_factor_analyst`| Beta, volatility regimes, and correlation patterns. | | `fundamental_value_investor`| Cash flow, capital allocation, and moats. | | `growth_innovation_strategist`| Secular tech trends and R&D efficiency. | | `commodity_market_strategist`| Energy, metals, and agricultural price shocks. | | `regulatory_compliance_officer`| Antitrust, taxation, and subsidies. | | `fx_sovereign_risk_strategist`| Currency translation and EM sovereign debt. | | `corporate_credit_risk_analyst`| Capital structure and solvency risks. | | `consumer_behavioral_economist`| Discretionary vs. staple spending and demographics. | | `institutional_microstructure` | Liquidity, ownership, and short-squeeze risks. | | `esg_transition_risk_assessor` | Climate regulations and corporate governance. | | `dividend_yield_strategist` | Payout sustainability and bond-proxy behavior. | | `intangible_asset_network_analyst`| Network effects, brand equity, and IP defense. | ## Methodology 1. **Ticker Sourcing:** Based on global liquidity (`config/tickers_list.toml`). 2. **Master Report:** Generated with search grounding and abstractive transformation to remove temporal anchors. 3. **Multidimensional Expansion:** Expanded by 15 specialized personas using `gemini-2.5-flash-lite` to capture idiosyncratic risks and behaviors. ## Usage Ideal for: - **Transformer Trading Systems:** Contextual embeddings for asset behavior. - **LLM Financial Reasoning:** Fine-tuning for macroeconomic shock responses. - **Factor Modeling:** Augmenting quantitative models with qualitative data. ## Technical Details - **Models:** `gemini-3.1-pro-preview` (Master), `gemini-2.5-flash-lite` (Personas). - **Format:** Parquet (Snappy compression). - **Dimensions:** Ticker + Master + 15 Personas. ## Disclaimer This dataset is **fully synthetic** and generated by LLMs. It is for research and development purposes only and does not constitute financial advice.--- 许可证：Apache-2.0 --- # 永恒合成金融资产报告 ## 数据集摘要 本数据集为专为基于Transformer的交易系统设计的全合成金融工具描述库，构建了覆盖700余种资产的多维度概念框架，并**严格消除前瞻偏差**：通过将资产的结构与行为特征从特定日期和命名历史事件中抽象剥离，实现无时间锚点的分析。 **数据集可在Hugging Face获取：** [bguzzo2k/tkrs_timeless_report](https://huggingface.co/datasets/bguzzo2k/tkrs_timeless_report) ## 核心特性 - **零前瞻偏差**：已清理所有时间锚点，将历史行为转化为抽象动态（例如“系统性信贷危机”）。 - **多资产覆盖**：涵盖全球指数、交易型开放式指数基金（ETF）、固定收益产品、大宗商品、外汇、标普500指数以及加密货币在内的700余个交易代码。 - **15种专家视角**：每种资产均通过15个专业化“分析视角”进行拆解分析，以确保生成高粒度且多样化的嵌入子空间。 ## 数据集结构 本数据集以Parquet格式存储，包含以下列： | 列名 | 描述 | | :--- | :--- | | `ticker` | 金融工具标识符。 | | `gemini_master` | 基础结构报告（通过`gemini-3.1-pro-preview`生成锚定内容）。 | | `macroeconomist` | 通胀、利率与货币政策视角。 | | `geopolitical_risk_analyst` | 冲突、贸易战与制裁视角。 | | `supply_chain_logistics` | 原材料、供应链瓶颈与劳动力市场视角。 | | `quantitative_factor_analyst` | 贝塔值、波动区间与相关性模式视角。 | | `fundamental_value_investor` | 现金流、资本配置与竞争壁垒视角。 | | `growth_innovation_strategist` | 长期科技趋势与研发效率视角。 | | `commodity_market_strategist` | 能源、金属与农产品价格冲击视角。 | | `regulatory_compliance_officer` | 反垄断、税收与补贴政策视角。 | | `fx_sovereign_risk_strategist` | 货币兑换与新兴市场主权债务视角。 | | `corporate_credit_risk_analyst` | 资本结构与偿付风险视角。 | | `consumer_behavioral_economist` | 可选消费与必选消费支出、人口结构视角。 | | `institutional_microstructure` | 流动性、股权结构与卖空挤压风险视角。 | | `esg_transition_risk_assessor` | 气候监管与公司治理视角。 | | `dividend_yield_strategist` | 派息可持续性与债券替代属性视角。 | | `intangible_asset_network_analyst` | 网络效应、品牌资产与知识产权保护视角。 | ## 生成方法 1. **交易代码遴选**：基于全球流动性水平（配置文件路径：`config/tickers_list.toml`）。 2. **主报告生成**：通过搜索锚定与抽象化转换，移除所有时间锚点。 3. **多维度扩展**：使用`gemini-2.5-flash-lite`通过15个专业化分析视角进行扩展，以捕捉资产的异质性风险与行为特征。 ## 适用场景 本数据集适用于： - **基于Transformer的交易系统**：用于生成资产行为的上下文嵌入。 - **大语言模型（Large Language Model，LLM）金融推理**：用于微调以适配宏观经济冲击响应任务。 - **因子建模**：为量化模型补充定性分析数据。 ## 技术细节 - **所用模型**：`gemini-3.1-pro-preview`（主报告生成）、`gemini-2.5-flash-lite`（分析视角生成）。 - **存储格式**：Parquet（Snappy压缩）。 - **数据维度**：交易代码 + 主报告 + 15个分析视角。 ## 免责声明 本数据集为**全合成数据**，由大语言模型生成，仅用于研发与学术研究，不构成任何投资建议。