[
  {
    "id": "1",
    "question": "You are the head of strategy and risk management at a global reinsurance group. After several years of strong profitability, the global reinsurance market is showing signs of transitioning from a hard to a more competitive phase amid abundant capital, the mainstreaming of alternative capital, and persistent natural catastrophe and social inflation risks.\n\nPlease write a deep research report titled \"Strategic Discipline and Earnings Quality in the Next Reinsurance Cycle\" and address the following:\n\n1. Market cycle and supply–demand dynamics: Analyze the key drivers behind the market’s shift from hard to softer conditions, including capital growth, selective rate pressure, and changing cedent behavior, and assess the implications for underwriting discipline and long-term profitability;\n2. Evolving risk structure and capital sources: Examine how natural catastrophe risk, social inflation, liability uncertainty, and the increasing role of alternative capital (e.g., insurance-linked securities and catastrophe bonds) are reshaping risk transfer, capital allocation, and portfolio management for reinsurers;\n3. Strategic response and long-term competitiveness: In a softening market environment, discuss how reinsurers can balance strategic discipline with earnings growth, and summarize key levers—such as cycle management, technology investment, and talent strategy—for building sustainable competitive advantage.",
    "classification": "Insurance-Reinsurance",
    "classification_code": "INS-RIN",
    "report_type": "Reinsurance Market Cycle and Strategic Response Report",
    "report_type_zh": "再保险行业周期与战略应对研究报告",
    "language": "en",
    "expert_evaluation_criteria": "## Evaluation Criteria for Strategic Resilience and Earnings-Quality Management Under the New Reinsurance Cycle\n\nThis evaluation framework is designed to assess the quality of research reports themed on strategic resilience and earnings-quality management in the context of a new underwriting-cycle evolution in the reinsurance industry. The evaluation emphasizes: the forward-looking quality of reinsurance cycle-phase judgment; the depth of analysis on the evolution of risk and capital structure; the professional fit of data and methodology; and the practical guidance value for improving reinsurers' long-term earnings stability and capital efficiency. The framework includes 4 primary dimensions and 12 secondary dimensions, with a total score of 100. It supports consistent application for both human expert review and large language model (LLM) automated scoring.\n\n### 1. Core Viewpoints and Cycle-Assessment Quality (25 points)\n**(1) Clarity and forward-looking nature of cycle-phase judgment (10 points)**\n- 9-10: Clearly states a phased conclusion of \"hard -> soft\" (tight -> loose) and specifies a time window and/or trigger conditions (e.g., 6-18 months, renewal seasons, capital-return threshold changes), benchmarked against market consensus; the conclusion can be validated or falsified by subsequent data.\n- 6-8: Directional judgment is clear, but time guidance, trigger conditions, or comparison with consensus is insufficient; forward-looking quality is average.\n- 0-5: Conclusion is vague or merely repeats commonplaces such as \"rate declines\" or \"abundant capital,\" lacking testable judgments.\n\n**(2) Decomposition of supply-demand drivers and transmission chain (8 points)**\n- 7-8: Links capital supply (traditional reinsurance capital, retained earnings, buyback/dividend constraints, rating/regulatory factors), demand side (cedants' retention changes, premium growth, terms and conditions rollback), and pricing/terms (rate-on-line, attachment, exclusions, occurrence/aggregate) into a closed loop; explains the transmission mechanism from \"underwriting discipline -> loss ratio/expense ratio -> earnings quality.\"\n- 4-6: Covers major factors, but the chain has logical jumps; arguments on key links (terms, layers, cedant behavior) are overly conceptual.\n- 0-3: Primarily lists factors; lacks causal relationships and mechanism explanation.\n\n**(3) Definition and measurement framework for \"earnings quality\" (7 points)**\n- 6-7: Proposes an earnings-quality indicator system suitable for reinsurance (e.g., risk-adjusted ROE/RAROC, split of underwriting profit vs. investment contribution, volatility and tail metrics such as TVaR/PML, reserve adequacy and development years, model/parameter uncertainty, capital consumption and reinsurance cost), and applies it throughout to evaluate \"growth vs. resilience.\"\n- 3-5: Mentions earnings quality but metrics are not systematic, or relies only on coarse measures such as ROE/combined ratio.\n- 0-2: Barely defines earnings quality; strategic recommendations are disconnected from measurement definitions.\n\n---\n\n### 2. Depth of Analysis on Risk Structure and Evolution of Capital Sources (30 points)\n**(1) Characterization of natural catastrophe risk and climate/secondary perils (10 points)**\n- 9-10: Distinguishes peak vs. non-peak zones and frequency vs. severity events; discusses secondary perils, inflation amplification of repair costs, model drift and \"non-stationarity\"; translates into concrete portfolio implications (layers, regions, perils, reinsurance structures, retrocession/retro).\n- 6-8: Covers key NatCat risk points, but lacks translation into portfolio impacts and terms-level actions.\n- 0-5: Generic discussion of increasing catastrophes/climate change; lacks reinsurance business context and actionable meaning.\n\n**(2) Social inflation and long-tail liability tail uncertainty (10 points)**\n- 9-10: Clearly explains drivers of social inflation (litigation environment, jury verdicts, legal fees, liability expansion, re-inflation transmission), reserving and claims-development risks in long-tail lines; provides specific impacts on underwriting/terms (limits, deductibles, claims-made, aggregate, exclusions) and pricing.\n- 6-8: Describes social inflation and long-tail risks, but lacks \"how to manage\" at the reserving/capital/terms level.\n- 0-5: Stops at the conclusion that \"liability uncertainty is high\"; no mechanisms or management levers.\n\n**(3) Mechanisms and impacts of the mainstreaming of alternative capital (ILS/cat bonds, etc.) (10 points)**\n- 9-10: Accurately distinguishes major ILS forms (cat bonds, collateralized reinsurance, sidecars, etc.), investor preferences and constraints (tenor, triggers, basis risk, liquidity, secondary markets), and differences vs. traditional reinsurance in pricing/terms/cycle elasticity; assesses marginal contribution to market softening and implications for reinsurers' capital strategy.\n- 6-8: Recognizes the importance of alternative capital and provides qualitative analysis, but insufficient explanation of instrument mechanics, terms, and market transmission.\n- 0-5: Treats alternative capital as a simple \"capital inflow/diversion\"; lacks professional mechanisms and verifiable inference.\n\n---\n\n### 3. Data Support, Methodology, and Verifiability (25 points)\n**(1) Reliability and granularity of data sources (10 points)**\n- 9-10: Data sources are clearly stated and traceable (e.g., disclosures of major reinsurers, rating agency/regulatory definitions, industry loss databases, ILS issuance and pricing data, renewal-terms change information); data can be segmented by line/region/layer or provides proxy indicators; specifies definition differences and comparability adjustments.\n- 6-8: Data is generally reliable, but granularity is coarse or definition explanations are insufficient.\n- 0-5: Sources are vague, years are outdated, or many key conclusions lack data support.\n\n**(2) Fit of pricing/capital/portfolio analysis methods (8 points)**\n- 7-8: Uses tools suitable for reinsurance analysis (e.g., decomposition of rate change vs. loss trend, adjustment for exposure/attachment changes, PML/TVaR/capital usage, scenario and stress testing, retrocession cost and net risk-exposure changes, decomposition of earnings volatility) and aligns with the earnings-quality framework.\n- 4-6: Methods are broadly reasonable, but missing key adjustments (terms/layers/inflation/model uncertainty), reducing robustness.\n- 0-3: Weak methodology; mostly narrative judgments that cannot support key conclusions.\n\n**(3) Disclosure of key assumptions and reproducibility (7 points)**\n- 6-7: Explicitly lists key assumptions and rationale (e.g., loss trends, inflation path, cost of capital, investment returns, ILS supply elasticity, cedant retention changes), provides sensitivity/scenario comparisons; readers can recompute the direction of major conclusions.\n- 3-5: Discloses some assumptions, but lacks parameter ranges/sensitivities; hard to reproduce.\n- 0-2: Assumptions are implicit or contradictory; conclusions are not verifiable.\n\n---\n\n### 4. Implementability of Strategic Responses and Long-Term Competitiveness (20 points)\n**(1) Concrete measures for cycle management and underwriting discipline (8 points)**\n- 7-8: Provides an executable action list for \"maintain discipline + grow opportunistically\" (e.g., tiered pricing floors, red lines for terms rollback, peak-risk limits and net-exposure management, dynamic retro/retrocession strategy, exit/shrink triggers, cedant relationship management and structured solutions), and explains how these improve long-term earnings quality rather than short-term scale.\n- 4-6: Direction is right, but measures are slogan-like; lacks triggers, metrics, and execution mechanisms.\n- 0-3: Only states principles such as \"operate prudently\" or \"underwrite cautiously\"; no actionable plan.\n\n**(2) Technology investment and capability building (6 points)**\n- 5-6: Presents a roadmap around models and data (NatCat model governance, data assets, pricing and portfolio optimization, automated underwriting, risk monitoring), including governance for model risk/bias; ties to decision scenarios (renewals, structure design, capital allocation).\n- 3-4: Notes the importance of technology/models, but lacks implementation path or governance arrangements.\n- 0-2: Vague technology statements, or disconnected from core reinsurance processes.\n\n**(3) Talent, organizational governance, and risk/compliance (6 points)**\n- 5-6: Clearly defines roles and checks-and-balances across strategy and risk/actuarial/underwriting/investment/retro teams (risk appetite, limits, model approval, reserve governance, performance and incentive alignment), and identifies key risks (model risk, concentration, liquidity/counterparty, reputational and regulatory).\n- 3-4: Mentions governance and talent, but coordination mechanisms and performance constraints are unclear.\n- 0-2: Neglects governance and compliance constraints, or reduces accountability to \"strengthen management.\""
  },
  {
    "id": "2",
    "question": "Machine learning models demonstrate advantages in capturing complex nonlinear interactions among factors, but they also introduce significant risks of overfitting and lack of interpretability (black-box issues). Please prepare a research report on machine learning-based stock selection models, covering at least the following aspects:\n\n1. Overview of machine learning stock selection models:\nCompare with traditional linear cross-sectional regression models, and provide an in-depth discussion of how commonly used models such as GBDT or other machine learning approaches capture multi-factor interaction effects and nonlinear return patterns.\n\n2. Model design and construction framework:\nDesign a machine learning model training framework, detailing key aspects of model construction, including how to design rolling windows to avoid look-ahead bias in time series. Discuss methods to control overfitting, such as early stopping mechanisms and sample weighting schemes (e.g., based on market capitalization or volatility decay).\n\n3. Model interpretability analysis:\nFor the selected machine learning stock selection model, introduce SHAP or similar feature attribution methods, and explain how SHAP values can be used in cross-sectional settings to identify key factors driving excess returns under current market conditions.\n\n4. Risk analysis of machine learning models:\nBased on relevant literature and research reports (with cited references), analyze potential risks of machine learning-based stock selection models, such as model breakdown under structural changes in market conditions.",
    "classification": "Capital Markets-Financial Engineering & Quantitative Research",
    "classification_code": "CAP-QUA",
    "report_type": "Research Report on Quantitative Stock Selection Models Based on Machine Learning Methods",
    "report_type_zh": "基于机器学习方法的量化选股模型研究报告",
    "language": "en",
    "expert_evaluation_criteria": "## Evaluation Criteria for *Research Report on Machine-Learning Stock Selection Models*\n\nThis framework evaluates research reports on machine-learning stock selection models. The focus is whether the report clearly compares ML models with traditional linear models, designs a robust training and anti-overfitting framework, provides meaningful interpretability analysis, and adequately identifies model risks and failure scenarios. Total: 100 points.\n\n---\n\n### 1. Model Principles and Method Comparison (25 points)\n\n#### (1) Comparison between traditional and ML models (10 points)\n- **9–10**: Clearly compares linear cross-sectional regression with GBDT or other ML models in nonlinear fitting, interaction capture, and use cases.\n- **6–8**: Basic comparison is provided, but depth is limited.\n- **0–5**: Only gives concepts without meaningful comparison.\n\n#### (2) Nonlinearity and interaction effects (8 points)\n- **7–8**: Clearly explains why ML models are better suited to capture complex factor relations in stock selection.\n- **4–6**: Provides some explanation, but lacks practical linkage.\n- **0–3**: Discussion is vague.\n\n#### (3) Model choice rationale (7 points)\n- **6–7**: Clearly justifies the choice of GBDT or another model, with strengths and limitations.\n- **3–5**: Mentions the model, but without clear rationale.\n- **0–2**: Model choice appears arbitrary.\n\n---\n\n### 2. Model Design and Training Framework (30 points)\n\n#### (1) Training pipeline and rolling-window design (10 points)\n- **9–10**: Presents a complete framework with clear train/validation/test splits and proper rolling-window design that avoids look-ahead bias.\n- **6–8**: Framework is broadly complete, but time handling is not rigorous enough.\n- **0–5**: Lacks a sound training design.\n\n#### (2) Overfitting control (10 points)\n- **9–10**: Systematically explains early stopping, regularization, sample weighting, and related controls, with relevance to stock selection.\n- **6–8**: Lists control methods, but mechanism analysis is limited.\n- **0–5**: Mentions overfitting risk without concrete solutions.\n\n#### (3) Metrics and validation design (10 points)\n- **9–10**: Clearly defines evaluation metrics and backtesting design, balancing predictive power and investment usability.\n- **6–8**: Includes metrics, but not comprehensively.\n- **0–5**: Lacks a proper evaluation framework.\n\n---\n\n### 3. Interpretability Analysis (25 points)\n\n#### (1) SHAP or similar method explanation (8 points)\n- **7–8**: Accurately explains the core idea and value of SHAP or similar attribution methods.\n- **4–6**: Basically correct, but not deep enough.\n- **0–3**: Explanation is vague or incorrect.\n\n#### (2) Cross-sectional factor attribution (10 points)\n- **9–10**: Uses SHAP values to identify key drivers of excess returns under the current market regime, with a clear analytical structure.\n- **6–8**: Includes attribution analysis, but not systematically.\n- **0–5**: Stops at method introduction only.\n\n#### (3) Investment implications of interpretability results (7 points)\n- **6–7**: Connects attribution results to factor logic, market style, or portfolio construction.\n- **3–5**: Some linkage is made, but depth is limited.\n- **0–2**: Lacks investment interpretation.\n\n---\n\n### 4. Risk Analysis and Research Quality (20 points)\n\n#### (1) Model failure risk identification (10 points)\n- **9–10**: Systematically analyzes risks such as regime shifts, sample drift, style rotation, and liquidity shocks.\n- **6–8**: Covers major risks, but mechanism analysis is limited.\n- **0–5**: Risk analysis is insufficient.\n\n#### (2) Literature and research support (5 points)\n- **5**: Uses sufficient and credible references, properly cited and well integrated into the argument.\n- **3–4**: Includes references, but with limited support value.\n- **0–2**: Lacks reliable references.\n\n#### (3) Report structure and professionalism (5 points)\n- **5**: Clear structure, proper terminology, and professional quantitative research style.\n- **3–4**: Basically complete, but average in expression.\n- **0–2**: Weak structure and professionalism."
  },
  {
    "id": "3",
    "question": "Please prepare an in-depth fundamental research report centered on the theme of \"the economics and fundamental impact of AWS AI infrastructure investment,\" and systematically address the following questions:\n\n1. Based on key takeaways from AWS re:Invent, summarize AWS’s latest strategic developments in AI infrastructure, including in-house chips (e.g., the Trainium series), computing platforms, and the AI technology stack. Explain how these developments differ from the traditional cloud computing model.\n2. From the perspectives of capital expenditure and unit economics, analyze the impact of large-scale AI investments on cost structure, depreciation dynamics, EBITDA/EBIT margins, and free cash flow. Explain the mechanism of \"short-term profitability pressure versus medium- to long-term profit release,\" and identify key forward-looking operating indicators from the perspective of operational data.\n3. Considering the growth of AI workloads, changes in customer mix, and the competitive landscape, evaluate the impact of AI-related businesses on medium-term revenue growth and return on invested capital (ROIC). Discuss the potential re-rating logic for AWS and Amazon’s overall valuation.\n4. Identify key risks and uncertainties (e.g., compute supply constraints, power and infrastructure limitations, AI pricing pressure, and intensifying competition), and assess the sensitivity of Amazon’s medium- to long-term fundamentals to these factors.\n\nRequirements: The analysis should focus on underlying mechanisms, maintain logical consistency with clearly stated assumptions, and demonstrate an integrated understanding of technological evolution, financial structure, and capital market dynamics.",
    "classification": "Capital Markets-Company & Equity Research",
    "classification_code": "CAP-COM",
    "report_type": "AWS Cloud and AI Investment Economics",
    "report_type_zh": "AWS 云计算与 AI 投资经济性分析",
    "language": "en",
    "expert_evaluation_criteria": "## Evaluation Criteria for *AWS AI Infrastructure Investment Economics and Fundamental Impact*\n\nThis framework evaluates deep research reports on AWS AI infrastructure investment. It emphasizes **tech–financial–valuation linkage**, **verifiable unit economics**, **short-term pressure vs. medium-term release**, and **risk scenarios with monitoring indicators**. The framework includes **5 primary dimensions and 14 secondary dimensions, totaling 100 points**.\n\n---\n\n### 1. Technical Strategy and Information Extraction (25 points)\n\n#### (1) Key takeaways from re:Invent 2025 (8 points)\n- **7–8**: Accurately extracts relevant disclosures (chips, platforms, supply, pricing, ecosystem), distinguishes facts vs. management views, provides sources, and clarifies uncertainty.\n- **4–6**: Covers main points but weak linkage to business implications.\n- **0–3**: Generic or inaccurate summary, insufficient for analysis.\n\n#### (2) Depth of AI infrastructure architecture understanding (10 points)\n- **9–10**: Explains chip strategy (e.g., Trainium), workload differences, bottlenecks (network, memory, power), and linkage to upper-layer services, with quantifiable variables.\n- **6–8**: Covers components but lacks constraint/mechanism analysis.\n- **0–5**: Superficial or concept-heavy without substance.\n\n#### (3) Identification of paradigm shift vs. traditional cloud (7 points)\n- **6–7**: Clearly identifies shifts (capex front-loading, utilization-driven economics, pricing changes, ecosystem lock-in) and links to financial impact.\n- **3–5**: Partial qualitative insight without financial linkage.\n- **0–2**: Fails to recognize key differences.\n\n---\n\n### 2. Capex and Unit Economics Analysis (30 points)\n\n#### (1) Capex breakdown and accounting rigor (10 points)\n- **9–10**: Breaks down AI investments (DC, servers, networking, power, facilities) with clear capitalization scope and linkage to delivery capacity.\n- **6–8**: Reasonable breakdown but lacks granularity or accounting clarity.\n- **0–5**: Generic capex discussion.\n\n#### (2) Depreciation and P&L transmission (10 points)\n- **9–10**: Clearly explains D&A impact (EBIT vs EBITDA), timing of asset ramp and utilization, and short-term margin pressure vs. medium-term release.\n- **6–8**: Understands direction but lacks dynamic modeling.\n- **0–5**: Misuses EBITDA/EBIT or lacks mechanism.\n\n#### (3) Unit economics model and testability (10 points)\n- **9–10**: Builds reproducible model (TCO, utilization, pricing, power, depreciation) with sensitivity analysis.\n- **6–8**: Includes key variables but incomplete framework.\n- **0–5**: Lacks verifiable economic model.\n\n---\n\n### 3. Growth, Competition, ROIC, and Valuation (25 points)\n\n#### (1) AI workload growth and revenue modeling (10 points)\n- **9–10**: Quantifies growth drivers (training vs inference, customer mix, pricing, backlog) with measurable indicators.\n- **5–8**: Describes drivers but lacks quantification.\n- **0–4**: Generic growth narrative.\n\n#### (2) Competitive dynamics and pricing pressure (7 points)\n- **6–7**: Analyzes competition across supply, chips, ecosystem, pricing, and lock-in, linking to growth, margins, and efficiency.\n- **4–5**: Basic comparison without financial linkage.\n- **0–3**: Weak or incorrect analysis.\n\n#### (3) ROIC and valuation re-rating framework (8 points)\n- **7–8**: Links ROIC to AI investment and builds valuation logic (growth, margins, capital efficiency), identifying triggers.\n- **4–6**: Mentions ROIC/valuation but lacks decomposition.\n- **0–3**: Weak or inconsistent framework.\n\n---\n\n### 4. Methodology, Assumptions, and Data Quality (10 points)\n\n#### (1) Assumptions clarity and consistency (4 points)\n- **4**: Key assumptions clearly stated, consistent, and linked to sensitivity.\n- **2–3**: Partial or weakly justified assumptions.\n- **0–1**: Missing or inconsistent assumptions.\n\n#### (2) Data quality and reproducibility (4 points)\n- **4**: Reliable sources, clear definitions, and reproducible calculations.\n- **2–3**: Reasonable data but incomplete transparency.\n- **0–1**: Unverifiable or unsupported data.\n\n#### (3) Structure and model transparency (2 points)\n- **2**: Clear structure, charts/models support logic.\n- **1**: Acceptable but scattered.\n- **0**: Poor structure.\n\n---\n\n### 5. Risk and Scenario Analysis (10 points)\n\n#### (1) Risk identification and transmission (6 points)\n- **5–6**: Covers key risks (supply, power, capex overruns, pricing, competition, regulation) with clear impact paths.\n- **3–4**: Lists risks without mechanism.\n- **0–2**: Weak or irrelevant risks.\n\n#### (2) Scenario analysis and monitoring indicators (4 points)\n- **4**: Provides at least two scenarios with quantified impact on margins/FCF/ROIC and clear tracking indicators.\n- **2–3**: Scenarios present but not quantified.\n- **0–1**: No scenario analysis."
  },
  {
    "id": "4",
    "question": "Based on the latest disclosed monthly shipment data of Q Technology (1478.HK), and following the analytical framework commonly adopted by sell-side research institutions for consumer electronics component companies, prepare an in-depth research report titled \"Shipment Structure Evolution and Growth Logic Assessment of Q Technology,\" and complete at least the following tasks:\n\n1. Shipment Structure and Business Differentiation: Compare the shipment trends of camera modules (CCM) and fingerprint recognition modules (FPM), explain the reasons behind the YoY pressure on CCM and the growth in FPM, and analyze these dynamics in conjunction with product cycles and customer structure.\n\n2. Specification Upgrades and Product Mix: Review the changes over the past five years in the proportion of high-resolution modules (32MP and above) within the company's CCM shipments, and discuss the impact of specification upgrades on ASP, revenue composition, and profitability.\n\n3. Non-Smartphone Applications and New Growth Drivers: Based on shipment performance in non-smartphone applications such as IoT and automotive electronics, evaluate the contribution of new scenarios to mid- to long-term growth, and analyze how these emerging growth drivers affect the company’s valuation.\n\n4. Strategic Initiatives and Investment Implications: Considering the company’s positioning in optical actuators and advanced optical technologies, as well as potential M&A activities, analyze the logic of vertical integration and valuation re-rating, and summarize key stock price catalysts and risks.",
    "classification": "Capital Markets-Company & Equity Research",
    "classification_code": "CAP-COM",
    "report_type": "Consumer Electronics Component Supplier Shipment Mix and Growth Assessment Report",
    "report_type_zh": "消费电子零部件公司出货结构与成长性评估报告",
    "language": "en",
    "expert_evaluation_criteria": "## Evaluation Criteria for “Q Technology Shipment Mix Changes and Growth Logic Assessment Research Report”\n\nThis evaluation system is designed to assess the quality of research reports centered on Q Technology’s shipment mix changes, business segmentation (CCM vs. FPM), and mid-to-long-term growth logic. It is compatible with both human review and consistent LLM-based automated scoring. Scoring emphasizes: rigorous handling of monthly shipment data and statistical definitions; a verifiable causal chain behind CCM pressure vs. FPM growth; mapping of spec upgrades to the ASP→revenue→gross margin transmission; and actionable judgments on sustainable growth and valuation re-rating conditions in non-smartphone scenarios.\n\nTotal score: 100 points; 4 primary dimensions; 13 secondary dimensions.\n\n### 1. Insights on Shipment Mix and Business Segmentation (30 points)\n#### (1) Monthly shipment data processing and definition consistency (8 points)\n- 7–8: Clearly discloses data sources and time window; makes necessary adjustments for “monthly volatility/seasonality/working days/Lunar New Year timing mismatch”; identifies and addresses definition changes (e.g., statistical scope, product categories, customer definition) and their impact on YoY/MoM comparisons; provides reproducible charts (at minimum: monthly CCM/FPM series, YoY/MoM, rolling 3M/12M).\n- 4–6: Completes basic YoY/MoM calculations and charting, but insufficiently flags seasonality/definition risks, or handles them only descriptively.\n- 0–3: Simply lists numbers without time-series processing or definition explanations, making conclusions questionable or not reproducible.\n\n#### (2) Quality of causal explanation for CCM pressure vs. FPM growth (12 points)\n- 11–12: Builds a clear, verifiable explanation chain across “demand/supply/competition/customers/product cycle,” and distinguishes the relative weight of drivers (e.g., Android supply-chain demand changes, customer share migration, per-device camera configuration changes, pricing pressure, yield/delivery constraints, fingerprint module penetration and solution switching); conclusions are cross-validated with shipment mix, base effects, and industry cycle indicators.\n- 7–10: Covers major factors, but lacks weighting judgment or has an incomplete evidence chain (e.g., discusses industry demand but not customers/competition, or discusses customers but not product cycle).\n- 0–6: Stays at generalized statements like “the market is weak/demand is soft,” without data or industry-logic support.\n\n#### (3) Sell-side level of customer mix and product cycle analysis (10 points)\n- 9–10: Clearly breaks down key customer types/platforms (using compliant phrasing such as “top customers/overseas customers/automotive customers”), and infers trends for the next 1–2 quarters based on new product launch cadence, build cycle, and inventory cycle; provides judgments and evidence on customer concentration, bargaining power, and order visibility.\n- 6–8: Mentions customers and cycles, but projection is rough, lacks time specificity, or lacks analysis of how order visibility/concentration affects outcomes.\n- 0–5: Hardly addresses customers and cycles, or only repeats common knowledge.\n\n---\n\n### 2. Spec Upgrades and Profitability Mapping (25 points)\n#### (1) Quantification and decomposition of 32MP+ mix changes (10 points)\n- 9–10: Provides monthly/quarterly trend of 32MP+ shipment share, YoY changes, and turning-point explanations; further decomposes contributions into “volume growth/structural upgrade/base effect/customer pull”; charts and definitions are clear.\n- 6–8: Includes share data and trend judgment, but lacks contribution decomposition or has weak turning-point reasoning.\n- 0–5: Only states qualitatively that “upgrading is significant,” without quantified share or with non-rigorous trend analysis.\n\n#### (2) Completeness of the ASP→revenue→gross margin transmission chain (10 points)\n- 9–10: Builds a reproducible bridge model: shipment volume × ASP (or ASP by spec) → revenue; derives gross margin/profit elasticity by incorporating yield, material cost, depreciation/expense amortization, and product mix changes; explains conditions under which “higher pixel does not necessarily mean higher profit” (e.g., price competition, yield ramp, customer rebates, foundry vs. in-house manufacturing mix).\n- 6–8: Can infer direction of ASP/revenue from spec upgrades, but insufficiently decomposes gross margin drivers, or fails to disclose key assumptions.\n- 0–5: Sticks to a linear narrative “higher pixels → higher ASP → better profit,” without cost/yield/competition constraints.\n\n#### (3) Key assumptions and sensitivity/scenario analysis (5 points)\n- 5: Lists at least 3 key assumptions (e.g., 32MP+ mix, ASP changes, yield/cost, customer share, industry demand growth) and provides quantified scenario/sensitivity (bull/base/bear) showing impact on revenue and profit.\n- 3–4: Has assumptions, but insufficient quantification or missing sensitivity analysis.\n- 0–2: No explicit assumptions; calculations are not verifiable.\n\n---\n\n### 3. Non-Smartphone Applications and Mid-to-Long-Term Growth Assessment (20 points)\n#### (1) Shipment performance and sustainability judgment for IoT/automotive electronics, etc. (8 points)\n- 7–8: Clearly presents scale, growth rate, volatility patterns, and drivers of non-smartphone application shipments; discusses correlation with the smartphone business (counter-cyclical vs. pro-cyclical); provides reasonable judgments on order visibility, certification cycles (automotive), and customer onboarding cadence.\n- 4–6: Describes growth but lacks analysis of “sustainability/visibility/certification cycle.”\n- 0–3: Slogan-like claims that “non-smartphone = new growth driver,” without data or industry mechanisms.\n\n#### (2) Competitive moats and commercialization path (7 points)\n- 6–7: Explains moat sources in new scenarios (technology, process, reliability, certification, delivery, cost, customer relationships, etc.) and competitor landscape; provides actionable commercialization path with stage breakdown (introduction/ramp/scale).\n- 3–5: Mentions moats but stays conceptual; lacks path and staging.\n- 0–2: Does not analyze moats or competition; conclusions are not credible.\n\n#### (3) Quantitative framework for mid-to-long-term contribution (5 points)\n- 5: Uses a clear framework to quantify contribution (e.g., 2–3 year revenue share/increment, profit contribution, or cash flow impact) and explains key constraints (capacity, certification, ASP, yield, customer concentration).\n- 3–4: Has quantitative targets but lacks constraints or assumption disclosure.\n- 0–2: No quantification or only vague judgments.\n\n---\n\n### 4. Strategic Initiatives, Valuation Re-Rating, and Investment Implications (25 points)\n#### (1) Strategic alignment and feasibility of optical actuators/frontier optics initiatives (8 points)\n- 7–8: Explains synergy with core business (performance, cost, yield, lead time, customer solution lock-in, etc.); discusses R&D/capex, mass-production ramp, and customer onboarding thresholds; identifies milestones (samples/design-in/MP) and their timeline.\n- 4–6: Clarifies strategic direction but lacks feasibility discussion on ramp/onboarding and milestones.\n- 0–3: Only lists concepts without an execution path.\n\n#### (2) M&A potential and value assessment of vertical integration (7 points)\n- 6–7: Evaluates M&A/integration benefits from synergies, cost structure, supply-chain security, bargaining power, and technology gap filling; also covers integration risks (culture/process/customer overlap, goodwill, cash flow pressure) and trigger conditions.\n- 3–5: Emphasizes synergies only; does not discuss integration risks or triggers.\n- 0–2: Treats M&A as a slogan-like positive, lacking an assessment framework.\n\n#### (3) Valuation methodology, re-rating logic, and peer benchmarking (6 points)\n- 5–6: Uses a reasonable valuation framework (one or a combination of P/E, P/B, EV/EBITDA, SOTP) and benchmarks against comparable companies/businesses; clearly states “conditions for re-rating to hold” (earnings inflection, business mix change, new business ramp, cash flow improvement, etc.) and cases where it does not hold.\n- 3–4: Has a valuation conclusion but insufficient benchmarking or unclear conditions.\n- 0–2: No valuation framework, or only provides target price/multiple without basis.\n\n#### (4) Specificity of stock catalysts and risk list (4 points)\n- 4: Both catalysts and risks include “event/indicator + realization timing/trigger condition + impact pathway,” and they balance (cross-check) the main-text logic.\n- 2–3: Lists points but lacks timing/impact pathway, or repeats the main text as a stack of items.\n- 0–1: Template-like risks/catalysts that do not match the company’s characteristics."
  },
  {
    "id": "5",
    "question": "Based on the current computational principles of large language model inference, the direction of model evolution, and the demand characteristics of downstream customers, please study how the large-scale deployment of model inference may drive the evolution of AI chip architectures (including but not limited to Processing-Near-Memory and 3D memory-logic stacking), changes in chip/server interconnect methods, and adjustments in model architectures. Further analyze which specific industry sectors may be affected by these changes.\n\nPlease follow the requirements below:\n\n1. From a technical perspective, conduct a comprehensive study of the full operational pipeline of current large-model inference services, including but not limited to model computation mechanisms, how hardware infrastructure performs the corresponding data processing, and storage/communication mechanisms. Under current trends in inference demand, identify the technical bottlenecks that may arise at each stage and discuss potential future improvement directions. All technical arguments should be supported by rigorously verified data sources to ensure the analysis is accurate, comprehensive, and up to date.\n\n2. For each major technological improvement direction, assess the industrial beneficiaries and adversely affected parties (including both directly impacted stakeholders and important indirectly affected stakeholders). Analyze the channels of impact, transmission speed, and potential competitive dynamics across industries, and based on this, evaluate the likely industrial landscape over the next two to three years.",
    "classification": "Capital Markets-Sector & Thematic Research",
    "classification_code": "CAP-STR",
    "report_type": "Analysis Report on Industry Transformation and Investment Opportunities Driven by Large Language Model Inference Demand",
    "report_type_zh": "大语言模型推理需求驱动的产业变革与投资机会分析报告",
    "language": "en",
    "expert_evaluation_criteria": "## Evaluation Criteria for *AI Chip Architecture Evolution and Industry Impact under Large-Scale LLM Inference Deployment*\n\nThis framework evaluates deep research reports on large-scale LLM inference deployment. It focuses on whether the report accurately deconstructs the inference pipeline, analyzes the evolution of chips, interconnects, and model architectures, and maps technological changes to industry segments and competitive dynamics over the next 2–3 years. Total: 100 points.\n\n---\n\n### 1. Inference Pipeline and Bottleneck Analysis (30 points)\n\n#### (1) Completeness of inference pipeline decomposition (10 points)\n- **9–10 points**: Systematically breaks down the inference workflow, covering compute, memory, data movement, interconnect, and scheduling.\n- **6–8 points**: Covers major steps, but some components lack depth.\n- **0–5 points**: Superficial description without a complete pipeline.\n\n#### (2) Accuracy of bottleneck identification (10 points)\n- **9–10 points**: Accurately identifies key constraints such as compute, bandwidth, latency, power, memory wall, and interconnect efficiency, and explains root causes.\n- **6–8 points**: Identifies main bottlenecks, but lacks depth or clarity on underlying causes.\n- **0–5 points**: Bottleneck analysis is vague or incorrect.\n\n#### (3) Rigor of data and technical evidence (10 points)\n- **9–10 points**: Arguments are supported by reliable, verifiable data; technical descriptions are accurate and up to date.\n- **6–8 points**: Some data support, but rigor or timeliness is limited.\n- **0–5 points**: Lacks evidence or contains technical inaccuracies.\n\n---\n\n### 2. Evolution of Chips, Interconnects, and Model Architectures (30 points)\n\n#### (1) Chip architecture evolution (10 points)\n- **9–10 points**: Clearly analyzes directions such as PNM and 3D memory-logic stacking, including drivers, use cases, and constraints.\n- **6–8 points**: Covers main directions but lacks mechanism analysis.\n- **0–5 points**: Lists concepts without substantive insight.\n\n#### (2) Network and system architecture changes (10 points)\n- **9–10 points**: Explains changes in chip, server, and cluster interconnects, and their impact on inference efficiency and system cost.\n- **6–8 points**: Provides directional analysis, but lacks clear linkage.\n- **0–5 points**: Lacks system-level perspective.\n\n#### (3) Model architecture adaptation and co-optimization (10 points)\n- **9–10 points**: Analyzes model structure, inference strategies, and hardware-software co-optimization, and links them to chip/system evolution.\n- **6–8 points**: Some analysis, but weak integration logic.\n- **0–5 points**: Ignores model-side adjustments or remains superficial.\n\n---\n\n### 3. Industry Impact and Competitive Dynamics (25 points)\n\n#### (1) Identification of beneficiaries and impacted players (10 points)\n- **9–10 points**: Accurately identifies direct and indirect beneficiaries and impacted segments, with clear logic.\n- **6–8 points**: Identifies key players, but coverage is incomplete.\n- **0–5 points**: Weak or generic industry mapping.\n\n#### (2) Transmission mechanisms and impact pathways (8 points)\n- **7–8 points**: Clearly explains how technological changes propagate along the value chain, including timing, pace, constraints, and key quantitative drivers.\n- **4–6 points**: Provides directional insight, but lacks systematic linkage.\n- **0–3 points**: Lacks clear transmission analysis.\n\n#### (3) 2–3 year outlook on industry structure (7 points)\n- **6–7 points**: Provides well-supported views on future competition based on supply, demand, technology maturity, and industry dynamics.\n- **3–5 points**: Offers directional views but lacks support.\n- **0–2 points**: Conclusions are vague.\n\n---\n\n### 4. Report Quality and Strategic Implications (15 points)\n\n#### (1) Structure and professionalism (5 points)\n- **5 points**: Clear structure with a coherent tech–industry–competition narrative.\n- **3–4 points**: Generally structured but average logic.\n- **0–2 points**: Poor structure.\n\n#### (2) Actionability of conclusions (6 points)\n- **5–6 points**: Provides clear industry insights, key directions, and tracking indicators.\n- **3–4 points**: Clear conclusions but limited actionability.\n- **0–2 points**: Vague conclusions.\n\n#### (3) Risk identification (4 points)\n- **4 points**: Covers key risks such as slower tech adoption, weak demand realization, cost constraints, supply chain bottlenecks, and technology shifts.\n- **2–3 points**: Lists major risks but lacks depth.\n- **0–1 point**: Missing risk disclosure."
  },
  {
    "id": "6",
    "question": "As the global AI computing architecture evolves from general-purpose GPUs toward ASICs (Application-Specific Integrated Circuits), the market has begun to reassess supply constraints in key upstream segments. Assume you are a securities analyst specializing in technology hardware and semiconductor supply chains. Please prepare an in-depth research report titled \"Supply Chain Bottleneck Analysis of China's PCB Industry under the AI ASIC Wave.\" \n\nPlease complete the following analysis tasks:\n\n1. Evolution of Computing Architecture and Changes in PCB Demand: Outline the core logic behind the transition of AI computing from GPUs to ASICs, and analyze the new requirements imposed on PCBs in terms of layer count, materials (high-speed/low-loss), manufacturing processes, and reliability. Evaluate how these requirements reshape the PCB market structure;\n2. Mechanism of PCB Bottleneck Formation: From the perspectives of capacity expansion cycles, yield ramp-up, and supporting equipment and materials, explain why high-end PCBs (e.g., high-layer HDI boards and high-speed server boards) may become a key constraint for AI ASIC expansion by 2026;\n3. Assessment of China's PCB Supply Chain: Evaluate Chinese PCB manufacturers in terms of technological capabilities, customer structure, and global competitive positioning. Analyze their strengths, weaknesses, and domestic substitution potential in meeting AI ASIC demand;\n4. Industry Impact and Investment Implications: Analyze how PCB bottlenecks affect the scaling pace of AI chips, computing costs, and capital expenditures of cloud providers. Based on this, summarize medium-term investment implications and key risks for China's PCB industry.\n\nPlease incorporate relevant data in the report to substantiate the analysis and support key arguments.",
    "classification": "Capital Markets-Sector & Thematic Research",
    "classification_code": "CAP-STR",
    "report_type": "China PCB Industry Value Chain Deep-Dive Report",
    "report_type_zh": "中国PCB行业产业链深度研究报告",
    "language": "en",
    "expert_evaluation_criteria": "## Evaluation Criteria for: Supply Chain Bottleneck Analysis Report on China's PCB Industry Under the AI ASIC Wave\n\nThis evaluation framework is used to conduct an executable, quantitative assessment of deep-research reports under the theme of 'technology hardware / semiconductor supply chain', focusing on high-end PCBs driven by AI ASICs. Scoring emphasizes: accuracy in understanding technology and system architecture; verifiability and quantifiability of bottleneck mechanisms; fact-based support for China's supply-chain positioning; and actionable guidance for industry timing and investment decisions. The framework contains 4 primary dimensions and 12 secondary dimensions, totaling 100 points, and is designed for consistent use across human review and LLM-based automated scoring.\n\n---\n\n### 1. Compute Architecture Evolution and Insights into PCB Technical Requirements (25 points)\n\n#### (1) GPU -> ASIC evolution logic and demand-side derivation (8 points)\n- 7-8 points: Clearly explains the drivers for moving from general-purpose GPUs to ASICs / specialized accelerators (cost/energy efficiency/TCO, model-specific operators, system integration and supply chain considerations, multi-vendor strategies, etc.), and connects them to how changes in system form factor impact PCBs (add-in card vs full system vs rack, interconnect topology, power density, I/O bandwidth, and networking-side changes), with an explicit causal pathway.\n- 4-6 points: Explains why ASICs are rising, but the transmission chain to PCB demand is coarse and misses key intermediate variables (I/O rate, board-level interconnect density, form-factor differences).\n- 0-3 points: Stays at the level of 'ASICs are more specialized and cheaper', or gets GPU/ASIC system differences wrong or confused.\n\n#### (2) PCB spec changes: layer count/materials/process and reliability requirements (10 points)\n- 9-10 points: Clearly identifies new requirements imposed by AI-ASIC-related carriers (accelerator cards, server motherboards, switches, etc.) for high layer count, high-speed low-loss materials, HDI/high-density interconnect, impedance control, back-drilling/blind-buried vias, low warpage, and thermal reliability; and provides checkable ways to express technical metrics (e.g., signal-rate generations, loss/dielectric parameter dimensions, line width/spacing and via-structure difficulties, reliability test focus areas).\n- 6-8 points: Covers the upgrade direction for layer count and materials, but provides insufficient rationale for why these indicators become binding constraints, or lacks expansion on process/reliability.\n- 0-5 points: Only states that 'high-end PCB demand is growing' with generalized requirements; or shows major concept confusion (e.g., consistently conflating PCB with packaging substrates/ABF, etc., without clarifying boundaries).\n\n#### (3) Technology roadmap and standards/generation cadence (7 points)\n- 6-7 points: Incorporates a 2-3 year generation roadmap (e.g., SerDes rate upgrades, networking generation shifts, changes in system power density) to form staged judgments on PCB demand; and identifies potential substitution/diversion factors (e.g., optical interconnect penetration, advanced packaging shifting board-level pressure, architectural integration changing PCB value content).\n- 3-5 points: Provides trend judgments but lacks a time scale or generation-based evidence.\n- 0-2 points: No roadmap cadence, or timing judgments clearly conflict with industry common sense.\n\n---\n\n### 2. Bottleneck Formation Mechanisms and Quantification of Supply Constraints (30 points)\n\n#### (1) Capacity build cycle and capex constraints (10 points)\n- 9-10 points: Clearly decomposes the critical path from project approval to mass production (plant and environmental compliance, equipment installation, process validation, customer qualification, ramp-up) and explains why constraints may form in 2026 (e.g., overlapping build and qualification cycles, equipment lead times, mismatch between expansion and demand peak). Higher scores if it provides a capacity definition and expansion timeline based on public information/research (e.g., by area, layer-count mix, utilization).\n- 6-8 points: Explains that the build cycle is long, but lacks milestone decomposition or a clear linkage to the 2026 constraint.\n- 0-5 points: Provides conclusions without mechanisms; or describes expansion as simply 'buy equipment'.\n\n#### (2) Yield ramp and process window: why high-end boards are hard to scale (10 points)\n- 9-10 points: Explains core difficulties of yield and consistency from lamination/stack-up, drilling/laser, copper plating, imaging/pattern transfer, impedance consistency, warpage control, reliability, etc.; and clarifies the feedback loop of 'yield-cost-delivery-qualification' (e.g., unstable yield causes delivery and cost to go out of control, prolonging customer validation).\n- 6-8 points: Mentions yield ramp-up but lacks breakdown of key process steps/failure modes.\n- 0-5 points: Summarizes as 'high technical barriers/low yield' without verifiable content.\n\n#### (3) Equipment/material matching, qualification, and supply-chain rigidity (10 points)\n- 9-10 points: Clearly identifies key equipment/material bottlenecks (e.g., high-precision imaging/exposure, drilling and inspection, vacuum lamination, testing capability, high-speed low-loss copper-clad laminates and related material systems), discusses lead times and supply elasticity, and analyzes how customer qualification (cloud vendors/ODMs/OEMs/chip companies) constrains supplier switching and requires time.\n- 6-8 points: Covers either equipment or materials, but insufficiently develops supply-chain rigidity and qualification barriers.\n- 0-5 points: Ignores material/equipment constraints, or contains no discussion of qualification/onboarding cycles.\n\n---\n\n### 3. Competitiveness of China's PCB Supply Chain and Localization/Substitution Potential (25 points)\n\n#### (1) Technical capability and product mapping of Chinese manufacturers (10 points)\n- 9-10 points: Provides a 'capability-product-application' mapping for Chinese PCB makers (e.g., high-layer HDI, high-speed server boards across product lines), explains key gaps (process stability, material systems, yield, qualification track record, global delivery, etc.), and gives plausible catch-up paths.\n- 6-8 points: Lists representative companies and rough capabilities, but lacks mapping to AI-ASIC carriers/specs.\n- 0-5 points: Stays at company lists and market-share enumeration without describing technical capabilities.\n\n#### (2) Customer structure, onboarding path, and pricing power (8 points)\n- 7-8 points: Identifies the key customer chain (cloud vendors/server ODMs/OEMs/switch vendors/chip companies) and their requirements on quality, delivery, qualification, and cost; analyzes customer concentration, price pass-through, and capacity-locking mechanisms (e.g., long-term orders, qualification, joint development).\n- 4-6 points: Mentions customer structure but does not explain onboarding and pricing/power mechanisms.\n- 0-3 points: Only states 'downstream demand is strong' without customer-side constraints.\n\n#### (3) Global competitive landscape and boundaries of localization (7 points)\n- 6-7 points: Benchmarks global players to position China (technology generation, yield stability, scale and cost, overseas capacity footprint, geopolitics) and states boundary conditions for localization/substitution potential (which specs are more substitutable vs which are still constrained by materials/equipment/qualification).\n- 3-5 points: Provides comparisons but is mostly qualitative and lacks boundary conditions.\n- 0-2 points: Only slogans 'localization' without conditions/constraints.\n\n---\n\n### 4. Industry Impact, Investment Implications, and Key Risks (20 points)\n\n#### (1) Impact mechanism on AI chip scaling, compute cost, and cloud capex (10 points)\n- 9-10 points: Builds a transmission model from 'PCB supply -> delivery lead time/cost -> accelerator/server shipments -> compute supply and capex', provides at least one reproducible quantitative framework (e.g., based on board value content/area, shipment constraints from extended lead time, impact of cost increases on system BOM and depreciation payback), and states timing and conditions for the impact.\n- 6-8 points: Explains direction of impact, but quantification is insufficient or key assumptions are not disclosed.\n- 0-5 points: Only qualitative statements like 'it will affect scaling and cost' without mechanisms/variables.\n\n#### (2) Actionability of medium-term investment conclusions (6 points)\n- 5-6 points: Provides clear investment conclusions and applicable scenarios (base/bull/bear), identifies key tracking indicators and catalysts (expansion commissioning pace, qualification progress, pricing/utilization, material supply, etc.), and distinguishes drivers (volume growth vs price uplift vs share gains).\n- 3-4 points: Has conclusions but trigger conditions/tracking indicators are unclear.\n- 0-2 points: Slogan-like conclusions without an executable tracking framework.\n\n#### (3) Identification of key risks and constraint conditions (4 points)\n- 4 points: Risks are comprehensive and counter-balance the main thesis, including but not limited to demand volatility and cloud capex cuts, technology-route changes (e.g., interconnect form-factor shifts), overly aggressive expansion causing a downcycle, yield/quality incidents, trade and export controls, material/equipment supply disruption, customer concentration and receivables risk, with explicit impact pathways.\n- 2-3 points: Lists major risks but lacks mechanisms or is insufficiently targeted.\n- 0-1 point: Risk section is missing or purely formalistic.\n"
  },
  {
    "id": "7",
    "question": "Recently, the correlation structure among multiple asset classes in global financial markets has undergone notable changes, with some assets significantly deviating from their historical behavior patterns and exhibiting pronounced \"outlier\" characteristics. This has sparked market discussions about a potential regime shift.\n\nIn this context, please prepare an in-depth research report titled \"Correlation Reconfiguration and Asset Anomalies: Is a Regime Shift Underway in Global Markets?\" and address the following analytical tasks:\n\n1. Changes in multi-asset correlation structure:\nBased on major asset classes such as equities, bonds, foreign exchange, and commodities, compare recent market behavior with long-term benchmarks and previous structural regimes. Identify key manifestations of correlation reconfiguration and asset anomalies (e.g., weakening correlation between China’s equity and bond markets, diminished safe-haven properties of gold, and sustained high-level volatility in U.S. equities).\n\n2. Drivers of asset anomalies:\nAnalyze the core factors driving deviations from historical correlations and pricing logic, incorporating perspectives from macroeconomic cycles, monetary policy divergence, geopolitical developments, and evolving industry themes. Distinguish between structural and cyclical influences.\n\n3. Market regime assessment and portfolio implications:\nEvaluate whether current market conditions reflect a short-term structural reshuffling or the emergence of a new medium-term asset pricing regime. Discuss the implications of unstable correlations for multi-asset allocation and risk management, and provide actionable strategy recommendations.",
    "classification": "Capital Markets-Macro & Strategy Research",
    "classification_code": "CAP-MAC",
    "report_type": "Global Multi-Asset Correlation Restructuring and Market Regime Shift Research Report",
    "report_type_zh": "全球多资产相关性重构与市场范式切换研究报告",
    "language": "en",
    "expert_evaluation_criteria": "## Evaluation Criteria for the Research Report \"Correlation Regime Rebuild and Asset Outliers: Is a Paradigm Shift Happening in Global Markets?\"\n\nThis evaluation framework is designed to provide a systematic, actionable, and quantitative assessment of a deep research report covering: multi-asset correlation regime rebuild + outlier asset identification + causal attribution + paradigm-shift judgment + portfolio implications.\n\nThe framework contains 4 primary dimensions and 13 secondary dimensions, with a total score of 100 points. It supports both human review and consistent LLM-based automated scoring.\n\n### 1. Quality of Key Findings and Insights (25 points)\n\n#### (1) Clarity and Testability of Core Conclusions (10 points)\n- 9-10: Explicitly answers whether a paradigm shift is occurring / which regime is more likely. Provides testable criteria (e.g., break tests, state-switching probability thresholds, key indicator trigger conditions) and time horizon (short-term / medium-term). Makes an explicit comparison against the market consensus narrative (\"the market thinks... but based on ... we think...\").\n- 6-8: Conclusions are fairly clear, but criteria are mostly qualitative or missing triggers/time orientation; differences vs consensus and falsifiable points are not explicit enough.\n- 0-5: Conclusions are vague or \"argues both sides\"; lacks testable standards and time dimension.\n\n#### (2) \"Incremental Insight\" on Correlation Regime Rebuild (8 points)\n- 7-8: Goes beyond describing correlation changes to identify structurally new features (e.g., \"shifting from rate-driven to geopolitics/supply-shock-driven,\" \"USD/real yields become a common cross-asset factor,\" \"conditions under which the classic stock-bond hedge fails change\"). Can pinpoint specific asset pairs/regions/tenors.\n- 4-6: Some novelty, but largely extends common narratives (e.g., \"inflation makes stocks and bonds move together\"); lacks granular pinpointing and an evidence chain.\n- 0-3: Mostly repeats common knowledge or provides news-style summaries; no meaningful incremental insight.\n\n#### (3) Research Framework and Communication Structure (7 points)\n- 6-7: Presents a clear framework (e.g., \"measurement -> outlier identification -> attribution -> regime classification -> portfolio mapping\"); chapter organization supports a closed-loop reasoning chain.\n- 3-5: Structure is broadly complete, but key links (outlier definition / decision criteria / portfolio mapping) are broken or weak.\n- 0-2: Loose structure; research path is hard to reproduce.\n\n### 2. Rigor of Data, Measurement, and Methods (30 points)\n\n#### (1) Data Coverage, Comparability, and Documentation (10 points)\n- 9-10: Covers equities/bonds/FX/commodities (at minimum with representative indices or yield tenors and clearly defined conventions). Specifies data frequency, sample period, currency conventions, trading time zones, whether FX-hedged, and missing-value handling. Ensures comparability across \"recent\" vs \"long-run benchmark\" vs \"prior regime\".\n- 6-8: Broad asset coverage, but insufficient convention details (e.g., mixed tenors, unclear currency basis), reducing credibility of some comparisons.\n- 0-5: Data source/conventions unclear or asset representation insufficient, making conclusions non-verifiable.\n\n#### (2) Appropriateness of Correlation Measures (8 points)\n- 7-8: Uses at least two types of measures and explains when each is appropriate, e.g., rolling correlation + conditional correlation (DCC/GARCH) / quantile correlation / tail dependence (copula, co-ES). Discusses correlation bias under rising volatility and the limitation that \"correlation != causality\".\n- 4-6: Uses basic methods such as rolling correlation, but lacks treatment/discussion of window choice, heteroskedasticity effects, or tail co-movement.\n- 0-3: Only static correlations or a few screenshot-style results; methods cannot support claims of \"structural change\".\n\n#### (3) Definition and Detection of \"Outliers\" (6 points)\n- 6: Provides a clear, reproducible definition (covers at least one of: correlation outlier / pricing anomaly / volatility anomaly / tail anomaly, with explanation). Uses threshold rules or statistical tests (z-score, quantile thresholds, Mahalanobis distance, anomaly detection, extreme value theory, etc.) and explains false-positive control.\n- 3-5: Describes outliers and uses simple thresholds, but lacks robustness (across windows/frequencies/baselines) or does not justify why the definition matches the research question.\n- 0-2: Outliers selected subjectively based on narrative; not reproducible.\n\n#### (4) Tests and Robustness for Structural Change / Paradigm Shift (6 points)\n- 6: Uses at least one structural-change identification method (breakpoints/change-points: Bai-Perron, CUSUM; regime switching: Markov switching; correlation-network rebuild metrics, etc.) and performs robustness checks (different sample windows, frequencies, excluding crisis periods, alternative proxies).\n- 3-5: Shows awareness of testing but relies on a single method or weak robustness; medium credibility.\n- 0-2: Entirely qualitative \"paradigm shift\" claims with no verifiable support.\n\n### 3. Causal Analysis and Structural vs Cyclical/Phase Distinction (25 points)\n\n#### (1) Mechanism Chain: Macro Cycle and Policy Divergence (10 points)\n- 9-10: Maps macro variables (inflation, growth, real rates, term premium, liquidity, fiscal expectations) to cross-asset pricing mechanisms (e.g., discount rates, risk premia, carry vs safe-haven channels). Validates with data or events (policy path divergence, rate differentials, inflation-expectation decompositions).\n- 6-8: Mechanisms are broadly correct, but evidence is weak (missing key variables, lacking quantitative validation, or weak causal identification).\n- 0-5: Slogan-like attribution (\"rate hikes explain everything\") without mechanism and evidence.\n\n#### (2) Explanatory Power: Geopolitics, Supply Shocks, and Thematic Forces (7 points)\n- 6-7: Links geopolitical risk, energy/shipping/supply-chain constraints, and themes (AI, energy transition, defense, etc.) to specific assets (commodity curves, related currencies, sector equities, credit spreads, etc.) and validates via event windows or risk-premium indicators.\n- 3-5: Mentions factors, but mapping to asset outliers is loose; lacks verifiable evidence.\n- 0-2: Generic discussion; cannot explain why those assets become outliers.\n\n#### (3) Identification and Weighting of Structural vs Cyclical/Phase Effects (8 points)\n- 7-8: Provides explicit criteria (persistence, changes in institutional/policy frameworks, permanent shifts in supply-demand curves, regime residence probabilities, etc.). Layers drivers into \"structural\" vs \"phase\" components and assesses relative weights (qualitative + quantitative where possible).\n- 4-6: Attempts to distinguish but criteria are unclear, or weights are subjective.\n- 0-3: No distinction, or distinction contradicts the evidence.\n\n### 4. Market Regime Judgment, Portfolio Mapping, and Risk-Management Actionability (20 points)\n\n#### (1) Regime-Decision Framework and Monitoring Indicators (8 points)\n- 7-8: Proposes an executable \"regime dashboard\" (e.g., inflation/growth percentiles, real rates, USD liquidity, volatility and correlation metrics, risk appetite, cross-asset risk premia) with clear thresholds/triggers and update frequency.\n- 4-6: Has a framework but lacks thresholds and operational details; hard to use for ongoing monitoring.\n- 0-3: Purely subjective calls without a monitoring system.\n\n#### (2) Implementability of Multi-Asset Allocation Recommendations (7 points)\n- 6-7: Under unstable correlations, provides clear allocation principles and instruments (diversify factors rather than assets, dynamic hedging, risk budgeting/volatility targeting, cross-asset relative value, tail hedges). States applicability conditions, transaction costs/margin constraints, and potential side effects.\n- 3-5: Directional advice exists, but lacks implementation path (which instruments, when to adjust, how to constrain drawdowns).\n- 0-2: Generic advice such as \"be cautious/diversify/time the market\".\n\n#### (3) Risk Management: Stress Tests, Tail Risk, and Model Risk (5 points)\n- 4-5: Provides stress scenarios (policy re-acceleration, growth cliff, geopolitical escalation, liquidity tightening, etc.) and maps them to portfolio P&L. Discusses rising tail dependence and hedge failure; flags model risks (window sensitivity, non-stationarity, structural breaks) and mitigation measures.\n- 2-3: Mentions stress tests or tail risk, but scenarios are vague or lacks portfolio-level implementation.\n- 0-1: Barely addresses risk management or only lists risks formally.\n"
  },
  {
    "id": "8",
    "question": "The global trade system is undergoing a new round of adjustments driven by the combined effects of tariff frictions and geopolitical competition. The United States has strengthened its tariff-centered trade pressure strategy, leading to increasing divergence in policy stances and industrial positioning among major economies.\n\nAssume you are a macro strategy portfolio manager at a leading asset management institution. Focusing on the theme of \"Tariff Policy from a Game-Theoretic Perspective: Strategic Intent of U.S. Tariff Policies and Differentiated Responses of Major Economies,\" please prepare a macro research report addressing the following tasks:\n\n1. Strategic intent of U.S. tariff policies:\nMove beyond surface-level indicators such as trade deficits and tariff rates, and analyze the underlying objectives of the U.S. in frequently deploying tariffs from a game theory and negotiation strategy perspective. Based on differentiated tariff policies toward major economies (with a focus on the EU, Japan and Korea, China, Southeast Asia, Canada, and Australia), examine the varying roles of tariffs—for example, whether they are primarily aimed at long-term supply chain restructuring and \"de-risking,\" used as bargaining tools to secure favorable negotiation outcomes, or serve other strategic purposes.\n\n2. Comparative response paths under multi-party strategic interactions:\nConstruct a multilateral game framework to identify the key industries affected in different economies and the degree of impact. Compare potential strategic responses under tariff shocks, such as reciprocal retaliation, strategic compromise, or absorbing industrial relocation. Analyze the constraints imposed by regional trade agreements and rules of origin.\n\n3. Equilibrium outcomes and market implications:\nBased on the interaction of strategies among multiple parties, assess the potential equilibrium outcomes of the current tariff conflict. Evaluate whether U.S. tariff policies may be adjusted again in response to domestic economic conditions under the Trump administration. Analyze the implications for global supply chain restructuring, U.S. inflation dynamics, and major asset classes (e.g., the U.S. dollar, gold, sovereign bonds), and propose corresponding asset allocation and risk hedging strategies.",
    "classification": "Capital Markets-Macro & Strategy Research",
    "classification_code": "CAP-MAC",
    "report_type": "Macroeconomic Thematic Research Report",
    "report_type_zh": "宏观经济专题研究报告",
    "language": "en",
    "expert_evaluation_criteria": "## Evaluation Criteria for the Macroeconomic Thematic Research Report: *Tariff War from a Game-Theoretic Perspective*\n\nThis evaluation framework is designed to conduct a structured and quantitative assessment of a macro thematic report centered on: “U.S. tariff strategy intentions — multi-party game responses — equilibrium and market impact/asset allocation.” It contains 4 primary dimensions and 12 secondary dimensions, with a total score of 100 points, and is compatible with both human review and consistent scoring by large language models (LLMs).\n\n### 1. Core Views and Strategic Insights (30 points)\n#### (1) Depth of Decomposing the “Underlying Objectives” of U.S. Tariff Strategy (12 points)\n- 11–12: Goes beyond surface-level deficit/tariff-rate narratives; clearly decomposes and balances at least two categories of objectives (e.g., supply-chain restructuring and de-risking, alignment with technology/industrial policy, domestic political-economic constraints, bargaining leverage over allies and adversaries, rule reshaping/bloc formation). Provides an evidence chain for “primary objective vs. secondary objective” and a time dimension (e.g., short-term negotiation leverage vs. medium-to-long-term industrial security).\n- 7–10: Proposes a deep-objective framework, but the weighting between primary/secondary goals is overly subjective, or it does not properly address consistency and conflicts between “short-term vs. long-term” objectives.\n- 0–6: Largely stays at the level of trade deficits, tariff changes, or slogan-like “decoupling/de-risking” descriptions, without forming a testable judgment of strategic intent.\n\n#### (2) Quality of Applying Game Theory and Negotiation Strategy (10 points)\n- 9–10: Explicitly uses and explains key mechanisms (e.g., credible threats/commitments, signaling and incomplete information, repeated games and reputation, bargaining/BATNA, first-mover advantage, coalition formation and free-riding), and maps them to observable tariff-policy behaviors (tariff escalation tempo, exemption lists, investigation procedures, differentiated clauses for allies, etc.).\n- 6–8: Uses some game/negotiation concepts, but the mapping to real policy actions is not tight enough, or the analysis is concept-stacking with a weak causal chain.\n- 0–5: Almost no game/negotiation analysis, or only name-checking that does not support conclusions.\n\n#### (3) Incremental Insight and Differentiated Conclusions (8 points)\n- 7–8: Conclusions explicitly contrast mainstream market narratives (“the market commonly believes… but this report argues…”), offering falsifiable/verifiable differentiated judgments (e.g., tariffs function more as a rule-restructuring tool toward allies than a single-point pressure tool on adversaries; or constraints on “de-risking” stem from inflation/election cycles), and identifies key validation indicators.\n- 4–6: Some novelty, but mostly minor reframing; validation indicators are unclear.\n- 0–3: Repeats common views; lacks new information or independent inference.\n\n---\n\n### 2. Multilateral Game Framework and Strategy Comparison (25 points)\n#### (1) Completeness of Multi-Party Game Model Construction (10 points)\n- 9–10: Clearly defines participants (at minimum the U.S. + two categories of major economies, such as allies like the EU/Japan/Korea, key rival economies, and emerging economies receiving relocation). Specifies strategy sets (reciprocal retaliation, selective compromise, industrial relocation/route diversion, non-tariff tools, alliance coordination, etc.), payoff/loss components (economic growth, inflation, employment, industrial security, political constraints), information structure, and action sequence. Can be expressed via a payoff matrix, extensive-form game tree, or scenario tree.\n- 6–8: Identifies main players and strategies, but payoff functions/constraints are largely qualitative; model readability is average.\n- 0–5: Only lists national stances or news events; lacks an auditable game structure.\n\n#### (2) Depth of Comparing Differentiated Response Paths (9 points)\n- 8–9: Provides mechanism-based explanations for “why responses differ” across economies (industrial structure, dependence on the U.S., supply-chain position, trade openness/capital account, domestic politics, alternative markets), and compares the costs and boundary conditions of strategy bundles (e.g., sustainability of retaliation, negotiation gains from compromise, capacity/institutional constraints for capturing relocation).\n- 5–7: Covers strategy differences across major economies, but remains descriptive; lacks constraints or quantitative characterization.\n- 0–4: Over-generalizes into “tough/compromise/wait-and-see,” without structured comparison.\n\n#### (3) Constraint Analysis of Regional Trade Agreements and Rules of Origin (6 points)\n- 6: Clearly explains how RTAs/FTAs, rules of origin, customs unions/cumulation rules, export controls, and sanctions compliance change the feasibility and cost of “rerouting/relocation/substitution,” and ties it to concrete impact channels (firm location decisions, entrepot/transshipment trade, certification costs, compliance risks).\n- 3–5: Mentions institutional constraints, but does not explain how they alter the strategy space or equilibrium outcomes.\n- 0–2: Largely ignores institutional and rule constraints, leading to distorted strategy inference.\n\n---\n\n### 3. Equilibrium Reasoning and Macro-to-Market Transmission (25 points)\n#### (1) Rigor of Deriving Game Equilibria / Outcome Sets (10 points)\n- 9–10: Provides a clear equilibrium concept or outcome set (Nash equilibrium/subgame perfection, trigger strategies in repeated games, coalition equilibrium, mixed strategies, or a “stable outcome interval”), states key conditions for equilibrium, and presents at least 2–3 scenarios (escalation, freeze/phased deal, targeted exemptions and bloc formation) with trigger variables.\n- 6–8: Forms scenario analysis, but equilibrium judgments are largely verbal; lacks necessary conditions or discussion of deviations.\n- 0–5: Only makes trend statements (“may escalate/ease”); lacks interactive derivation.\n\n#### (2) Impact Mechanisms on Supply-Chain Restructuring and Verifiable Indicators (7 points)\n- 6–7: Maps equilibrium outcomes to supply-chain restructuring paths (friend-shoring, regionalization, onshoring of critical segments, dual-circulation backups), and proposes trackable indicators (share of trade diversion/transshipment, FDI destination flows, concentration of import sources for key intermediates, freight rates/lead times, corporate CAPEX and site selection).\n- 3–5: Discusses supply-chain restructuring but indicators and validation pathways are incomplete.\n- 0–2: Supply-chain discussion stays at slogan level; lacks mechanisms and indicators.\n\n#### (3) Transmission to U.S. Inflation and Major Asset Classes (8 points)\n- 7–8: Clearly decomposes the “tariffs → prices → inflation → policy → assets” chain; distinguishes one-off price-level shocks from persistent inflation mechanisms (substitution costs, supply constraints, FX pass-through, profit absorption), and analyzes directional impacts and conditions for USD, gold, and sovereign bonds (e.g., real rates, safe-haven demand, fiscal/term premium, USD liquidity).\n- 4–6: Provides directional views, but parts of the transmission chain are jumpy, or asset analysis is disconnected from macro assumptions.\n- 0–3: Asset conclusions are detached from macro reasoning and rely more on heuristics.\n\n---\n\n### 4. Data, Method Verifiability, and Investment Actionability (20 points)\n#### (1) Data Source Quality, Timeliness, and Fit-for-Purpose (8 points)\n- 7–8: Data sources are clear and traceable (official statistics, international organizations, customs/trade databases, supply-chain/shipping/high-frequency price data, policy texts), and timing matches current tariff/policy developments. Explains definitional differences and limitations (nominal vs. effective tariff rates, exemptions and enforcement differences).\n- 4–6: Data is broadly reliable, but definitions are insufficiently explained or key data is missing.\n- 0–3: Data sources are vague, outdated, or weakly linked to conclusions.\n\n#### (2) Method Transparency and Reproducibility (6 points)\n- 6: Discloses key steps and parameters (scenario assumptions, weight settings, inflation pass-through/effective tariff calculations, event studies or correlation tests, etc.), enabling a third party to reproduce the main conclusions.\n- 3–5: Methods are broadly reasonable, but key parameters/steps are not clearly specified.\n- 0–2: Almost no methodological disclosure; conclusions are not verifiable.\n\n#### (3) Executability of Asset Allocation and Risk Hedging Plans (6 points)\n- 5–6: Provides scenario-linked allocation/hedging ideas (position direction, hedging instruments or substitute assets, trigger and exit conditions, main risk exposures), and specifies applicable investor types or constraints (duration, currency risk, liquidity).\n- 3–4: Offers allocation suggestions but they are generic; lacks triggers/boundaries.\n- 0–2: Only slogan-like “bullish/bearish” calls; not implementable."
  },
  {
    "id": "9",
    "question": "A systematic analysis of inflation structure and policy reaction mechanisms has become a key topic in macroeconomic research.\n\nFocusing on the theme of \"Tariff Shocks, Inflation Structure, and Federal Reserve Policy Orientation,\" and based on the latest U.S. macroeconomic data, please prepare a structured research report addressing the following issues:\n\n1. Analyze recent trends in U.S. CPI and core CPI, distinguishing among core goods, core services (especially housing-related components), as well as energy and food prices, and explain their key driving factors.\n\n2. Discuss the transmission mechanisms of U.S. tariff policy changes on inflation. Explain how tariffs affect overall inflation through goods prices, corporate costs, and profit margins, and assess the sustainability of upward inflation risks.\n\n3. In conjunction with employment growth, the unemployment rate, and the composition of newly unemployed populations, evaluate whether the U.S. economy is transitioning from an \"inflation-driven\" phase to a \"growth- and employment-risk-driven\" phase.\n\n4. Based on the above analysis, assess the likely monetary policy stance of the Federal Reserve over the coming quarters. Explain the rationale for potential rate cuts despite incomplete disinflation, and analyze key uncertainties and risks.\n\nNotes:\n1. Each section should follow a \"data–analysis–conclusion\" structure: first present specific data or factual evidence, then conduct mechanism-based analysis, and finally provide concise conclusions or judgments.\n2. The report should conclude with a dedicated \"Key Takeaways\" section summarizing the overall outlook for Federal Reserve policy, highlighting core judgments and major uncertainties.",
    "classification": "Capital Markets-Macro & Strategy Research",
    "classification_code": "CAP-MAC",
    "report_type": "U.S. Inflation Dynamics and Monetary Policy Outlook Report",
    "report_type_zh": "美国通胀形势与货币政策前瞻研究报告",
    "language": "en",
    "expert_evaluation_criteria": "## Evaluation Criteria for Macro Research Reports: Tariff Shocks, Inflation-Structure Shifts, and the Federal Reserve’s Policy Stance\n\nThis evaluation framework is designed to provide an executable, quantitative assessment of research reports centered on “tariff shock → inflation structure → Fed policy stance,” supporting consistent use in both human review and LLM-based automatic scoring. The framework contains **4 top-level dimensions and 12 sub-dimensions**, with a total score of **100 points**. Scoring emphasis: **structured decomposition capability, rigor of transmission-mechanism reasoning, verifiability of data and methods, and the operability of policy-path inference with risk awareness**.\n\n---\n\n### 1. Core Conclusions and Structured Insights (30 points)\n\n#### (1) Clarity of the Core Thesis and Time Guidance (10 points)\n- **9–10**: States a clear, testable main conclusion (e.g., “over the next 2–3 quarters the primary driver of core inflation disinflation will be… / tariffs will mainly show up as a one-off price level effect / the policy stance will tilt toward…”), including a **time window** and **directional judgment**; explicitly benchmarks against market consensus expectations or Fed communications (e.g., “the market is pricing…, but we believe…”).\n- **6–8**: Conclusion is clear but conservative / closely follows consensus; time guidance or trigger conditions are relatively vague.\n- **0–5**: No explicit conclusion or timing; mostly a compilation of materials or conceptual discussion.\n\n#### (2) Incremental Insight from Inflation-Structure Decomposition (10 points)\n- **9–10**: Goes beyond component breakdown (core goods / core services / housing / energy / food) and provides **incremental explanation**, such as: housing’s lag and leading indicators; the relationship between services inflation and wages/demand; how measures like “supercore services” affect inference; identifies structural reasons behind apparent “disinflation/reacceleration.”\n- **6–8**: Completes component breakdown and standard explanations, but mechanism discussion for key components (especially housing) is not deep enough or lacks incremental perspective.\n- **0–5**: Decomposition is rough (only headline/core) or component narratives are disconnected from data.\n\n#### (3) Key Judgment on Tariff Shocks and the Inflation Path (10 points)\n- **9–10**: Explicitly answers whether tariff-driven upside inflation risk is sustainable; distinguishes **one-off price-level increases** from **persistent inflation (second-round effects)**, and provides evidence-based justification (pass-through elasticity, lags, firms’ absorption capacity, strength of demand, anchoring of inflation expectations, etc.).\n- **6–8**: Discusses persistence, but lacks quantitative or conditional statements (e.g., “it depends on…” without testable conditions).\n- **0–5**: Only qualitatively states tariffs raise inflation, without addressing persistence and mechanisms.\n\n---\n\n### 2. Logical Rigor and Transmission-Mechanism Argumentation (25 points)\n\n#### (1) Completeness of the CPI/Core CPI Component Driver Chain (9 points)\n- **8–9**: Builds a clear chain: component trends → drivers (supply/demand, wages, rents, goods prices, policy and financial conditions) → implications for the outlook; handles key “interpretation difficulties” (e.g., housing lag, seasonal-adjustment noise, base effects).\n- **5–7**: Main storyline is largely intact, but key links (housing, services inflation) are jumpy or insufficiently evidenced.\n- **0–4**: Logical breaks; unclear mapping between component changes and driver explanations.\n\n#### (2) Decomposition of Tariff Pass-Through and Discussion of Profit/Cost Absorption (8 points)\n- **7–8**: Clearly distinguishes and substantiates at least three channels: **import prices / final-goods prices**, **firm costs and pricing strategy**, **margin absorption and redistribution**; discusses sector differences, competitive structure, substitution effects, and provides a reasonable view on lag length.\n- **4–6**: Covers main channels but misses key elements such as margin absorption, sector heterogeneity, or lags.\n- **0–3**: Treats tariffs as “directly pushing up CPI,” with no intermediate mechanism.\n\n#### (3) Evidence Closed-Loop for Regime/State-Shift Judgments (8 points)\n- **7–8**: Forms a closed-loop judgment using “job growth, unemployment rate, and the structure of increases in the unemployed” (e.g., unemployment reasons, duration, labor-force participation, job openings/quits as supporting evidence); clearly explains why the weight shifts from inflation to growth/employment risk.\n- **4–6**: Cites employment and unemployment indicators, but lacks structural dimensions, or conclusions remain at “strong/weak.”\n- **0–3**: Uses a single indicator (e.g., unemployment rate) to claim a regime shift; argumentation is insufficient.\n\n---\n\n### 3. Data Quality, Definition Consistency, and Method Verifiability (25 points)\n\n#### (1) Data Recency, Source Authority, and Definition Notes (10 points)\n- **9–10**: Uses the **latest available** official data and clearly labels sources and release dates (e.g., BLS CPI components, Employment Report/CPS structure, JOLTS, BEA/PCE, Import Price Index, etc.); explains key definition differences (CPI vs PCE, YoY vs annualized MoM, seasonal-adjustment effects, housing-component characteristics).\n- **6–8**: Data sources are reliable and fairly recent, but definition notes are insufficient or key components are missing.\n- **0–5**: Data is outdated, sources are unclear, or mixed definitions undermine robustness.\n\n#### (2) Methods for Inflation Decomposition and Contribution Analysis (8 points)\n- **7–8**: Provides at least one auditable decomposition method: component **contributions (pp contributions)**, 3/6-month annualized momentum, diffusion index / sticky-inflation metrics, etc.; explains calculation steps or key parameters so others can reproduce.\n- **4–6**: Has charts and narrative, but lacks a reproducible decomposition/estimation path.\n- **0–3**: Conclusions only; missing methods and calculation process.\n\n#### (3) Quantification and Scenario-Based Assessment of Tariff Shocks (7 points)\n- **6–7**: Provides an actionable estimation framework (e.g., affected import-category weights × pass-through elasticity × margin-absorption share × lag length), and conducts at least two scenarios (mild/strong shock) or sensitivity analysis.\n- **3–5**: Offers an estimation idea, but parameters and scenarios are coarse and hard to verify.\n- **0–2**: Entirely qualitative; no quantification or scenarios.\n\n---\n\n### 4. Operability of Policy Inference, Risks, and Uncertainty (20 points)\n\n#### (1) Fed Reaction Function and Policy-Path Inference (10 points)\n- **9–10**: Integrates inflation structure, employment/growth risks, and financial conditions into a unified framework (akin to a “reaction function / constraint set”); provides the policy stance over the next few quarters (hold / cut / re-tighten) and **trigger conditions**; explains why “rate cuts may occur even before inflation fully returns” (e.g., restrictive real rates, downside tail risks to growth, labor-market weakening leading inflation, anchored inflation expectations, etc.).\n- **5–8**: Provides directional judgment, but lacks trigger conditions or does not adequately explain constraints behind “why cuts are possible.”\n- **0–4**: Merely restates market pricing or officials’ remarks; lacks independent inference.\n\n#### (2) Key Uncertainties, Risk Items, and Their Impact Channels (6 points)\n- **5–6**: Risk discussion is specific and mechanism-based (e.g., tariff escalation scope, oil-price shock, sticky housing inflation, wage re-acceleration, abrupt shifts in financial conditions, re-rise in inflation expectations) and explains direction and magnitude (relative strength) of impacts on inflation/growth/policy path.\n- **3–4**: Lists risks but lacks impact channels or weakly links them to the conclusions.\n- **0–2**: Risks are vague or purely formal.\n\n#### (3) Tradable/Decision-Useful “Timing–Catalysts–Validation Indicators” (4 points)\n- **4**: Provides a clear watchlist: the next 1–2 key data releases and meeting milestones (e.g., CPI components, employment structure, import prices, FOMC meeting/dot plot), and for each specifies a “validate/falsify threshold” (e.g., core services momentum, speed of unemployment increase, tariff-related category price anomalies).\n- **2–3**: Includes catalysts and focus indicators, but lacks thresholds or falsifiable statements.\n- **0–1**: No clear tracking and validation framework."
  },
  {
    "id": "10",
    "question": "Demographic shifts, declining long-term interest rates, and the ongoing evolution of insurance accounting standards are fundamentally reshaping the asset-liability management (ALM) framework of life insurers. Achieving stable returns and capital resilience under long-duration liabilities has become a core strategic priority for global insurance groups.\n\nUsing Prudential Financial as a case study, please write an in-depth research report on the theme of \"Optimizing Insurers’ Asset-Liability Structures under Low Interest Rates and Demographic Change,\" addressing the following key questions:\n\n1. Ex-post Evaluation of Retirement Business Transformation: Prudential accelerated its strategic shift toward retirement solutions in the 2010s. Evaluate the financial outcomes of this transformation: Have PGIM’s alternative asset allocations delivered the expected spread returns? Has the FlexGuard variable annuity business generated returns above its cost of capital across different interest rate cycles?\n\n2. Trade-offs between GAAP and Statutory Accounting: Prudential reports under both U.S. GAAP and statutory accounting frameworks. How do differences in liability measurement and hedge accounting treatments affect the reported profitability of its long-duration annuity book? Where are the key hidden risk exposures?\n\n3. Global Practices and Implications for China: Drawing on international insurance practices, what lessons can be derived for Chinese life insurers in optimizing asset-liability structures, mitigating spread compression risks, and enhancing long-term value management capabilities?",
    "classification": "Insurance-Actuarial & Reserving",
    "classification_code": "INS-RSV",
    "report_type": "Asset–Liability Structure Optimization in Insurance",
    "report_type_zh": "保险公司资产负债结构优化研究",
    "language": "en",
    "expert_evaluation_criteria": "## Evaluation Criteria for *Insurance ALM Optimization: A Case Study of Prudential Financial*\n\nThis framework evaluates deep research reports on asset-liability management (ALM) under low interest rates, demographic shifts, and evolving accounting standards. The focus is on assessing strategy effectiveness, understanding accounting impacts on profitability and risk, and deriving actionable industry insights. Total: 100 points.\n\n---\n\n### 1. Retirement Business Transformation and Performance (30 points)\n\n#### (1) Strategy and business structure (10 points)\n- **9–10**: Clearly explains the rationale, path, and structural changes in Prudential’s retirement shift.\n- **6–8**: Captures direction but lacks depth.\n- **0–5**: Lacks systematic analysis.\n\n#### (2) PGIM alternative assets and spread income (10 points)\n- **9–10**: Evaluates contribution to yield and spread with supporting logic or data.\n- **6–8**: Provides direction but limited depth.\n- **0–5**: Weak or missing analysis.\n\n#### (3) FlexGuard annuity performance (10 points)\n- **9–10**: Assesses returns across rate cycles vs. cost of capital.\n- **6–8**: Lacks cycle or cost perspective.\n- **0–5**: Superficial analysis.\n\n---\n\n### 2. Accounting Differences and Risk Impact (30 points)\n\n#### (1) GAAP, statutory, and IFRS 17 differences (10 points)\n- **9–10**: Clearly explains liability measurement and profit recognition differences.\n- **6–8**: Generally correct but incomplete.\n- **0–5**: Contains major misunderstandings.\n\n#### (2) Impact on reported earnings (10 points)\n- **9–10**: Explains how accounting affects earnings volatility and profitability.\n- **6–8**: Directionally correct but lacks depth.\n- **0–5**: Weak analysis.\n\n#### (3) Risk exposure identification (10 points)\n- **9–10**: Identifies key risks such as interest rate risk, duration mismatch, hedging risk, and capital volatility.\n- **6–8**: Covers main risks but not systematically.\n- **0–5**: Insufficient risk analysis.\n\n---\n\n### 3. International Experience and China Implications (25 points)\n\n#### (1) Global experience (10 points)\n- **9–10**: Summarizes ALM and product strategy insights from global insurers.\n- **6–8**: Some insights but not systematic.\n- **0–5**: Weak synthesis.\n\n#### (2) Localization and applicability (10 points)\n- **9–10**: Provides actionable recommendations tailored to China’s regulatory and market context.\n- **6–8**: Insights lack practicality.\n- **0–5**: Generic suggestions.\n\n#### (3) Long-term value and spread management (5 points)\n- **5**: Clearly outlines long-term value creation and spread-risk mitigation.\n- **3–4**: Directional but not specific.\n- **0–2**: Missing analysis.\n\n---\n\n### 4. Report Quality and Conclusions (15 points)\n\n#### (1) Structure and professionalism (5 points)\n- **5**: Clear and consistent with insurance research standards.\n- **3–4**: Mostly complete.\n- **0–2**: Weak structure.\n\n#### (2) Conclusions and insights (6 points)\n- **5–6**: Clear and strategically valuable conclusions.\n- **3–4**: Moderate depth.\n- **0–2**: Vague conclusions.\n\n#### (3) Risk and limitations (4 points)\n- **4**: Covers key risks such as rates, regulation, and product design.\n- **2–3**: Limited coverage.\n- **0–1**: Missing risk analysis."
  },
  {
    "id": "11",
    "question": "Amid rising prices of innovative drugs, increasingly complex boundaries between public and commercial insurance coverage, and the acceleration of value-based care, U.S. commercial health insurers play an increasingly critical role in improving access to innovative therapies while controlling healthcare costs. However, they face a structural paradox: the need to enhance drug accessibility while simultaneously containing cost growth. How can leading institutions strike a balance between these objectives?\n\nFocusing on the theme of \"How U.S. Commercial Health Insurance Supports Access to Innovative Drugs and Value-Based Care,\" and drawing on the practices of major insurers such as UnitedHealth Group, Elevance Health, and Cigna, please prepare a structured in-depth research report addressing the following issues:\n\n1. The dilemma of PBM vertical integration:\nUnitedHealth (Optum), CVS/Aetna, and Cigna (Express Scripts) have all vertically integrated pharmacy benefit management (PBM) capabilities. Does this model genuinely improve drug affordability and access, or does it primarily function as a profit extraction mechanism? What key evidence supports each perspective?\n\n2. Value-based contracts: where is the evidence?\nInsurers have entered into numerous value-based contracts for innovative drugs. Identify specific therapeutic areas (e.g., oncology, gene therapy) and analyze whether outcome-based payment models have effectively improved access in these areas, and where meaningful progress remains limited.\n\n3. Service gaps beyond coverage:\nFor patients with rare diseases or high-cost biologics, obtaining insurance coverage is only the first hurdle. What roles should (and can) commercial insurers play in care navigation, medication adherence, and clinical support? Which current business models are most feasible and sustainable in addressing these gaps?",
    "classification": "Insurance-Health Insurance",
    "classification_code": "INS-HLT",
    "report_type": "Research Report on Commercial Health Insurance Participation in Innovative Drug Payment and Healthcare Collaboration",
    "report_type_zh": "商业健康险参与创新药支付与医疗协同机制研究报告",
    "language": "en",
    "expert_evaluation_criteria": "## Evaluation Criteria for *How U.S. Commercial Health Insurers Support Innovative Drug Access and Value-Based Care*\n\nThis framework evaluates deep research reports on commercial health insurance, innovative drug access, and value-based care. The focus is whether the report explains how insurers balance access and cost control, assesses the real effects of PBM vertical integration and value-based contracts, and identifies service gaps beyond coverage together with sustainable business models. Total: 100 points.\n\n---\n\n### 1. Core Mechanisms of Commercial Insurance Support for Drug Access (25 points)\n\n#### (1) Problem framing and core tension (10 points)\n- **9–10**: Clearly defines the structural tension between improving access and controlling costs, and accurately explains the role of commercial insurance in drug payment.\n- **6–8**: Identifies the main tension, but the framework is incomplete.\n- **0–5**: Weak definition and limited system-level perspective.\n\n#### (2) Industry structure and leading-company practices (8 points)\n- **7–8**: Effectively analyzes UnitedHealth, Elevance, Cigna, and peers, highlighting model differences and sector relevance.\n- **4–6**: Covers major firms, but with limited comparison.\n- **0–3**: Lacks company-practice analysis.\n\n#### (3) Data and evidence support (7 points)\n- **6–7**: Uses key data on pricing, claims, coverage, and utilization effectively.\n- **3–5**: Includes data, but weakly connected to conclusions.\n- **0–2**: Insufficient evidence.\n\n---\n\n### 2. PBM Vertical Integration and Value-Based Contract Assessment (35 points)\n\n#### (1) Pros and cons of PBM vertical integration (12 points)\n- **10–12**: Systematically evaluates models such as Optum, CVS/Aetna, and Express Scripts, including their impact on pricing, access, cost control, and profit allocation.\n- **7–9**: Explains key pros and cons, but lacks depth.\n- **0–6**: Surface-level description only.\n\n#### (2) Evidence and controversy analysis (10 points)\n- **9–10**: Identifies the main evidence supporting both “better affordability/access” and “profit extraction,” and gives a balanced judgment.\n- **6–8**: Discusses evidence, but controversy analysis is incomplete.\n- **0–5**: Limited evidence or one-sided judgment.\n\n#### (3) Real-world effect of value-based contracts (13 points)\n- **11–13**: Uses cases in oncology, gene therapy, and other areas to assess the real effect of outcomes-based payment on access, cost control, and execution feasibility, while distinguishing areas of progress from those still constrained.\n- **8–10**: Covers major fields, but lacks implementation analysis.\n- **0–7**: Describes the model without validating outcomes.\n\n---\n\n### 3. Service Capability Beyond Coverage and Business Models (25 points)\n\n#### (1) Identification of service gaps (8 points)\n- **7–8**: Accurately identifies gaps in care navigation, adherence support, and clinical assistance.\n- **4–6**: Identifies some gaps, but not systematically.\n- **0–3**: Lacks service-layer analysis.\n\n#### (2) Role of insurers and capability boundaries (9 points)\n- **8–9**: Clearly explains what insurers should do and what they can realistically do in patient support.\n- **5–7**: Provides role analysis, but boundary judgment is incomplete.\n- **0–4**: Role definition is unclear.\n\n#### (3) Feasibility and sustainability of business models (8 points)\n- **7–8**: Compares different support models in terms of feasibility, economics, and sustainability.\n- **4–6**: Includes comparison, but not deeply.\n- **0–3**: Lacks business-model assessment.\n\n---\n\n### 4. Report Quality and Conclusions (15 points)\n\n#### (1) Structure and professionalism (5 points)\n- **5**: Clear structure, strong logic, and solid healthcare-payment research quality.\n- **3–4**: Basically complete, but average in organization.\n- **0–2**: Weak structure.\n\n#### (2) Actionability of conclusions (6 points)\n- **5–6**: Provides clear and actionable implications for insurer strategy, payment models, or service models.\n- **3–4**: Clear conclusions, but only moderately actionable.\n- **0–2**: Vague conclusions.\n\n#### (3) Risks and boundary conditions (4 points)\n- **4**: Covers key risks such as regulation, drug pricing, execution complexity, data availability, and incentive mismatch.\n- **2–3**: Lists major risks, but without depth.\n- **0–1**: Missing risk disclosure."
  },
  {
    "id": "12",
    "question": "Under the U.S. insurance regulatory framework, investments in alternative assets such as private equity and commercial real estate must be evaluated within solvency, capital, and compliance constraints. As commercial real estate private equity funds become an important vehicle for insurers’ long-term asset allocation, their structural design and regulatory compatibility warrant careful assessment.\n\nAgainst the background of a large U.S. life insurance company planning to participate—through capital contribution—in a private equity fund primarily investing in U.S. commercial real estate, prepare a research-oriented analytical report based on legal and compliance due diligence to support investment committee review and regulatory filing.\n\nPlease address the following issues:\n\n1. From a regulatory and compliance perspective, outline key requirements for insurers investing as limited partners in real estate private equity funds, including solvency constraints, capital adequacy, investment limits, and investment purpose tests.\n2. Based on the limited partnership structure, analyze the allocation of rights and responsibilities among GPs, fund managers, and LPs, focusing on investment decision-making, related-party transactions, and disclosure mechanisms.\n3. From an asset allocation and alternative investment strategy perspective, evaluate the role of commercial real estate private equity funds within insurers’ portfolios, with emphasis on income stability, duration matching, liquidity, and concentration risk.\n4. In conjunction with U.S. GAAP and insurance regulation, assess the impact on insurers’ financial performance, asset measurement, and risk-based capital (RBC), and propose corresponding risk mitigation measures.",
    "classification": "Insurance-Life Insurance",
    "classification_code": "INS-LIF",
    "report_type": "Insurance Investment in Real Estate Funds: Compliance & Risk",
    "report_type_zh": "保险资金不动产基金合规与风险研究",
    "language": "en",
    "expert_evaluation_criteria": "## Evaluation Criteria for Research Reports on U.S. Insurance Capital Investing in Commercial Real Estate Private Equity Funds\n\nThis evaluation system is designed to assess the quality of research and analytical reports on \"large U.S. life insurers investing in private equity funds primarily targeting commercial real estate.\" It focuses on whether the report forms a complete analytical closed loop of: regulatory and capital constraints -> fund structure and governance -> asset allocation and ALM alignment -> risk control and decision executability. The report should be able to support insurance capital investment decisions and regulatory communication scenarios. The framework includes 4 primary dimensions and 14 secondary dimensions, with a total score of 100 points, and is intended to be consistently applicable to both human expert review and large language model (LLM) automated scoring.\n\n---\n\n## 1. Regulatory, Capital, and Accounting Fit (30 points)\n\n### (1) Identification of Regulatory Framework and Applicability Boundaries (8 points)\n- 7-8: Accurately explains the applicability logic of state insurance regulation and the NAIC framework, and clearly maps each step of the transaction structure to the relevant regulatory constraints and applicability boundaries.\n- 4-6: Covers key regulators, but the mapping between rules and the transaction structure is insufficient.\n- 0-3: Significant omissions or errors in regulatory understanding.\n\n### (2) Solvency and Investment Limit Constraint Analysis (8 points)\n- 7-8: Builds a closed loop of \"capital constraints -> investment size -> structural parameters,\" explaining how RBC, concentration limits, and look-through measurement affect the investment plan.\n- 4-6: Identifies major constraints but lacks parameterized analysis.\n- 0-3: Only provides high-level/principle-based descriptions of capital constraints.\n\n### (3) Accounting Measurement and Capital Impact Analysis (US GAAP + SAP/RBC) (8 points)\n- 7-8: Clearly explains GAAP vs. SAP treatment logic and the implications for earnings volatility, capital quality, and regulatory communication.\n- 4-6: Directionally correct but missing key judgments or transmission mechanisms.\n- 0-3: Accounting or capital analysis is clearly mismatched.\n\n### (4) Compliance Due Diligence and Regulatory Filing/Reporting Pathway (6 points)\n- 5-6: Forms a complete closed loop of due diligence -> remediation -> disclosure -> filing/reporting, covering compliance, conflicts of interest, and disclosure requirements.\n- 3-4: Due diligence coverage is relatively comprehensive but lacks a practical filing/reporting execution pathway.\n- 0-2: Due diligence content is templated or not actionable.\n\n---\n\n## 2. Fund Structure and Governance Mechanisms (LP Perspective) (25 points)\n\n### (1) GP/LP Rights, Responsibilities, and Incentive/Constraint Mechanisms (7 points)\n- 6-7: Clearly breaks down rights and responsibilities and fee/incentive structures, and identifies key terms affecting the safety of insurance capital.\n- 4-5: Structure description is largely complete but risk mapping is insufficient.\n- 0-3: Unclear understanding of responsibilities or mechanisms.\n\n### (2) Investment Decision Process and LP Protection Mechanisms (7 points)\n- 6-7: Clearly specifies investment decision authority, control over重大事项 (major matters), and constraints on strategy drift, and proposes actionable LP protection clauses.\n- 4-5: Process exists but lacks threshold-based design.\n- 0-3: Governance mechanisms are generic.\n\n### (3) Related-Party Transactions and Conflict-of-Interest Management (6 points)\n- 5-6: Systematically identifies potential conflicts and proposes approval, disclosure, and restriction mechanisms.\n- 3-4: Identifies some risks but control measures are insufficient.\n- 0-2: Ignores conflict risks.\n\n### (4) Information Disclosure and Look-Through Governance Capability (5 points)\n- 5: Disclosure granularity meets the needs of insurers' capital management and post-investment monitoring.\n- 3-4: Disclosure requirements are average.\n- 0-2: Does not reflect look-through management needs.\n\n---\n\n## 3. Asset Allocation Positioning and ALM Alignment (25 points)\n\n### (1) Portfolio Positioning and Liability-Side Matching (8 points)\n- 7-8: Uses an ALM framework to quantify duration, cash flows, and alignment with liability constraints.\n- 4-6: Positioning is reasonable but insufficiently quantified.\n- 0-3: Ignores liability-side constraints.\n\n### (2) Return Drivers and Sensitivity Analysis (7 points)\n- 6-7: Decomposes return drivers and conducts scenario analysis for key variables such as interest rates and cap rates.\n- 4-5: Only provides directional analysis.\n- 0-3: Lacks verifiable return analysis.\n\n### (3) Liquidity and Capital Call Management (5 points)\n- 5: Assesses capital call pacing and proposes insurer-level cash flow management plans.\n- 3-4: Identifies risks but lacks management measures.\n- 0-2: Ignores liquidity constraints.\n\n### (4) Concentration and Portfolio Correlation Control (5 points)\n- 5: Completes look-through analysis of underlying assets and proposes limits and trigger mechanisms.\n- 3-4: Only discusses concentration qualitatively.\n- 0-2: Does not assess portfolio risk.\n\n---\n\n## 4. Risk Control and Decision Executability (20 points)\n\n### (1) Risk Mitigation Measures and Post-Investment Monitoring System (12 points)\n- 11-12: Proposes a clause-based and metric-based risk control system covering entry criteria, contractual protective provisions, ongoing monitoring indicators, and disposal/response processes.\n- 7-10: Measures are relatively complete but lack thresholds or an execution closed loop.\n- 0-6: Risk control recommendations are overly generic.\n\n### (2) Report Reviewability and Decision Support Capability (8 points)\n- 7-8: Report structure fits investment committee review and regulatory communication needs; conclusions provide conditional decision basis.\n- 4-6: Structure is clear but decision guidance is insufficient.\n- 0-3: Difficult to support actual investment decision-making.\n"
  },
  {
    "id": "13",
    "question": "Following a period of elevated interest rates and the subsequent transition into a rate-cutting cycle, U.S. life insurers are reassessing the strategic role of savings-oriented insurance products within their asset-liability management (ALM) frameworks. In a declining interest rate environment, how to align product design and asset allocation in order to balance liability costs, earnings stability, and long-term operating flexibility has become a key strategic issue for life insurance companies.\n\nFocusing on the theme of “interest rate normalization and asset-liability coordination,” conduct an in-depth analysis of Participating Life Insurance in the U.S. life insurance market, addressing the following questions:\n\n1. In light of recent U.S. interest rate cycles and the characteristics of life insurance liabilities, analyze the structural advantages of participating life insurance over traditional guaranteed-rate life insurance in terms of liability cost flexibility, equity allocation capacity, and earnings stability.\n2. From a product mechanism perspective, explain the operational logic of “guaranteed benefits plus non-guaranteed dividends” in U.S. participating life insurance, with particular emphasis on dividend sources, distribution rules, and the role of surplus management in smoothing long-term returns.\n3. From the perspectives of asset allocation, accounting treatment, and regulatory constraints, analyze how participating life insurance affects insurers’ capacity to allocate to equities, high-dividend assets, and alternative investments under the current interest rate and regulatory environment, and discuss key ALM constraints and trade-offs.\n4. Based on practical experience in the U.S. insurance market, evaluate the long-term role of participating life insurance within savings-oriented and wealth management insurance product systems, and summarize the key risks and uncertainties that insurers should carefully manage when expanding participating product offerings.",
    "classification": "Insurance-Life Insurance",
    "classification_code": "INS-LIF",
    "report_type": "Participating Life Insurance and ALM Coordination Research Report",
    "report_type_zh": "参与型寿险产品与资产负债协同研究报告",
    "language": "en",
    "expert_evaluation_criteria": "## Evaluation Criteria for a Research Report on ALM Collaborative Optimization of U.S. Participating Life Insurance During an Interest-Rate Downturn\n\nThis evaluation framework is designed to assess research reports on the topic of 'Coordination between U.S. participating life insurance product mechanics and asset allocation under a declining interest-rate environment.' The framework contains 4 primary dimensions and 12 secondary dimensions, with a total score of 100 points. It supports consistent application for both human grading and LLM-based automated scoring. Scoring emphasizes: accuracy of understanding of institutional rules and product mechanisms, rigor of the ALM transmission chain, feasibility under U.S. accounting and regulatory frameworks, and the operability of conclusions with adequate risk coverage.\n\n---\n\n### 1. Insights on the Interest-Rate Cycle and Institutional Advantages on the Liability Side (25 points)\n\n#### (1) Depiction of the interest-rate cycle and scenario design (8 points)\n- 7-8: Clearly characterizes the recent U.S. tightening-high plateau-declining phases (at minimum covering key changes in policy rates, the yield curve/term spreads, reinvestment rates, and credit spreads/default cycles), and translates them into ALM-usable scenario assumptions for life insurers (e.g., 'new money yield down x bps,' curve bull-flattening/bull-steepening, spreads widening/tightening, equity drawdowns). The time horizon is explicit.\n- 4-6: Describes the direction and stages of rates, but lacks yield-curve/spread/scenario quantification elements, or the mapping to insurer operating variables is insufficient.\n- 0-3: Only vaguely states 'rates fall' without a cycle structure or scenarios usable for inference.\n\n#### (2) Liability-side characteristics and spread-pressure transmission (9 points)\n- 8-9: Accurately decomposes key attributes of U.S. life insurance liabilities (guarantee costs, adjustability of dividends/credited rates, surrender/partial withdrawal/policy loan behavior, expenses and capital usage, etc.), and explains the transmission chain under rate declines from 'liability cost -> investment yield -> capital/dividend capacity.' Distinguishes the different pressures between new business pricing and in-force liability management.\n- 4-7: Covers major liability characteristics and spread logic, but is weak on behavioral assumptions, in-force vs. new business differences, and/or capital transmission.\n- 0-3: Liability description is inconsistent with U.S. life/participating product realities, or fails to form an effective transmission chain.\n\n#### (3) Argument for the 'institutional advantage' of participating vs. guaranteed-rate products (8 points)\n- 7-8: Builds a contract/institution-level comparison around the three prompt points (liability cost flexibility, room for equity allocation, earnings stability). Examples: how the non-guaranteed nature of dividends creates a buffer where 'returns are adjustable -> costs are adjustable'; how guarantee provisions and dividend mechanics affect asset duration/cash-flow matching; why participating products have stronger advantages in return smoothing and long-term operating flexibility (while also stating prerequisites and trade-offs).\n- 4-6: Provides comparative conclusions, but the argument mostly stays at an experiential level and lacks mechanism explanations linking contract terms to ALM outcomes.\n- 0-3: Comparison is off-target or treats participating insurance as 'high-yield wealth management' without reflecting institutional differences.\n\n---\n\n### 2. Product Mechanics: How Guaranteed Benefits + Non-Guaranteed Dividends Work (25 points)\n\n#### (1) Contract structure and decomposition of cash value/guaranteed items (8 points)\n- 7-8: Clearly explains the core structure of participating life insurance (guaranteed cash value/guaranteed interest rate or guaranteed accumulation mechanism, premium and expense structure, death benefit, ways to use cash value, etc.), and identifies the rigid ALM constraints and risk points created by the 'guaranteed portion' during rate declines (e.g., guarantee floor, expense assumptions, duration characteristics).\n- 4-6: Describes the general structure, but the linkage between guaranteed items and cash-flow/ALM constraints is unclear.\n- 0-3: Misstates or confuses the basic structure (e.g., treats dividends as contractually guaranteed returns).\n\n#### (2) Dividend sources, allocation rules, and the 'contribution principle/fairness' (9 points)\n- 8-9: Accurately explains the 'three sources' of dividends (investment experience, mortality experience, expense experience, etc.) and the allocation logic under mutual/shareholder structures. Explains common allocation and measurement principles (e.g., policy contribution, intergenerational equity, dividend scale setting logic, dividend options such as cash/reduction of premium/paid-up additions) and their incentive impacts on both policyholders and the insurer.\n- 4-7: Covers dividend sources and allocation, but lacks operational detail (which experience items flow into dividends, how fairness is reflected, how dividend options affect outcomes).\n- 0-3: Dividend sources/rules are clearly inaccurate, or treats dividends like mutual fund distributions without reflecting insurance mechanics.\n\n#### (3) Surplus management and long-term smoothing: mechanisms, boundaries, and costs (8 points)\n- 7-8: Explains how insurers achieve long-term stability via dividend scale adjustments, surplus/buffers (e.g., capital buffers and mechanisms to absorb credit losses and interest-rate volatility), and intertemporal smoothing. Also specifies boundary conditions (regulatory/capital constraints, intergenerational equity disputes, hidden risks from excessive smoothing) and typical trade-offs during rate declines (e.g., 'protect dividends vs. protect capital/ratings').\n- 4-6: Mentions 'smoothing' but lacks specific tools/constraints, or does not discuss costs and negative consequences.\n- 0-3: Only slogan-level statements about 'smoothing through dividends' with no mechanism or boundaries.\n\n---\n\n### 3. Asset Allocation and ALM Constraints/Trade-offs Under U.S. Accounting and Regulatory Frameworks (30 points)\n\n#### (1) Identification of U.S. regulatory/accounting frameworks and extraction of key constraints (10 points)\n- 9-10: Grounds the discussion in real U.S. frameworks (at minimum, clearly covers Statutory/NAIC basis and RBC constraints, and highlights key mechanisms such as admissibility, capital factors, interest-rate and credit-risk capital charges, cash-flow testing/asset adequacy requirements; bonus if it connects GAAP LDTI to earnings volatility and risk management). Explains how these constraints, in turn, shape asset choice and dividend capacity for participating products.\n- 6-8: Mentions some regulatory/accounting points, but does not form a clear mapping of 'constraints -> actions -> outcomes.'\n- 0-5: Over-generalizes the framework as 'stricter/looser regulation' without key U.S.-specific elements.\n\n#### (2) Equity, high-dividend equities, and alternatives: feasible paths and limits (10 points)\n- 9-10: Discusses the allocation space for equities/high-dividend equities/alternatives (private credit, infrastructure, real estate, etc.) under the constraint framework: impacts on capital and volatility after entering the general account, liquidity and valuation issues, term matching and return sources. Distinguishes the roles of 'dividend/factor exposure' vs. 'credit spread/liquidity premia,' and provides qualitative or semi-quantitative statements on allocation caps/risk budgets.\n- 6-8: States the directional conclusion that 'more risk assets can be allocated,' but lacks path detail (under which capital/accounting basis, and how impacts show up in dividends and solvency).\n- 0-5: Treats equity/alternative allocation as a pure investment view without reflecting insurance liabilities and regulatory constraints.\n\n#### (3) ALM co-design: linkages among duration/cash flow/reinvestment/behavioral risks (10 points)\n- 9-10: Provides an executable ALM co-design framework (e.g., liability cash flows and behavioral assumptions -> duration gap and convexity -> reinvestment strategy and credit allocation -> linkage to dividend scale/product pricing -> capital and liquidity buffers), and clearly states key trade-offs during rate declines (yield erosion vs. capital stability, moving down credit quality vs. loss distribution, liquidity vs. illiquidity premium, dividend stability vs. intergenerational equity).\n- 6-8: Framework is largely complete, but lacks deep trade-off analysis and/or lacks implementable metrics (e.g., gap measurement, sensitivities, stress-test dimensions).\n- 0-5: No closed-loop ALM; assets and liabilities are disconnected.\n\n---\n\n### 4. Long-Term Positioning, Industry Benchmarking, and Risk Uncertainty (20 points)\n\n#### (1) Quality of market practice evidence and peer benchmarking (6 points)\n- 5-6: Cites representative U.S. participating players/product practices (e.g., operating characteristics of major mutual life insurers, dividend strategies, and asset-allocation tendencies) and explains 'why they do so' rather than merely listing facts. Data sources are clear (annual reports, regulatory filings, rating agency reports, industry associations, etc.).\n- 3-4: Includes cases/data, but they are fragmented; benchmarking dimensions and/or source citations are insufficient.\n- 0-2: Little to no practical evidence, or cases do not match participating products.\n\n#### (2) Judgment on the long-term role of participating products within savings/wealth management (8 points)\n- 7-8: Clearly answers 'what is the long-term role, for whom, and what does it substitute or complement': delivers a consistent conclusion across customer needs (stable growth, tax/estate planning, volatility tolerance, liquidity preference, etc.), distribution and competing products (fixed annuities, IUL/VUL, traditional guaranteed-rate life insurance, etc.), and insurer objectives (stable AUM, capital efficiency, dividend promise management). States applicability boundaries in a declining-rate period and under future cycle transitions.\n- 4-6: Provides a positioning conclusion, but does not connect the customer-product-insurer triad, and/or lacks boundary conditions from a cycle perspective.\n- 0-3: Positioning is vague or detached from the core 'ALM coordination' storyline.\n\n#### (3) Key risks and uncertainties: completeness and actionable mitigations (6 points)\n- 5-6: Risk coverage is comprehensive and tightly tied to participating mechanics: includes at least interest-rate and reinvestment risk, credit losses and spread risk, equity/alternative volatility and liquidity risk, behavioral risks such as surrender/policy loans, reputational and sales-compliance risks from dividend reductions, and regulatory/accounting change risk. Provides actionable mitigation ideas (stress-test dimensions, risk budgets, dividend policy governance, liquidity tiering, etc.).\n- 3-4: Lists major risks, but lacks impact pathways or mitigations remain principle-level.\n- 0-2: Risk discussion is perfunctory or mismatched to the key trade-offs of participating products.\n"
  },
  {
    "id": "14",
    "question": "Recently, amid a so-called “tariff tantrum,” the US Treasury cash-futures basis widened sharply, with implied repo rates on Treasury futures rising well above comparable SOFR rates, raising concerns about market liquidity and leveraged trading stability. Against this backdrop, please write a deep research report titled \"Treasury Basis Dislocations during the Tariff Tantrum: Liquidity Stress or Temporary Mispricing?\" and address the following tasks:\n\n1. Review of basis dislocations: Systematically examine the magnitude, duration, and cross-maturity patterns of the recent basis widening across 2-year to 10-year and ultra-long Treasury futures, and compare them with past stress episodes;\n2. Mechanisms and market participant behavior: Analyze the key transmission channels behind the sharp basis widening, focusing on hedge fund deleveraging, dealer intermediation capacity, and funding market dynamics, and explain the subsequent normalization;\n3. Assessment of market stress and policy implications: Evaluate whether the episode reflects a temporary breakdown in cash-futures arbitrage or signals broader liquidity and structural stress in the US Treasury market, and discuss the implications for rate market stability and risk management.",
    "classification": "Capital Markets-Fixed Income & Rates Research",
    "classification_code": "CAP-FIX",
    "report_type": "US Treasury Basis Dislocations and Rates Market Liquidity Stress Report",
    "report_type_zh": "美债基差异动与利率市场流动性压力研究报告",
    "language": "en",
    "expert_evaluation_criteria": "## Evaluation Criteria for \"U.S. Treasury Basis Dislocations Under Tariff Shocks\" (Research Report)\n\nThis rubric is used to conduct an actionable, quantitative evaluation of the research report \"U.S. Treasury Basis Dislocations Under Tariff Shocks\". It is compatible with both human review and LLM-based automated scoring. The framework includes **4 primary dimensions and 12 secondary dimensions**, with a total score of **100 points**. It emphasizes “consistent definitions, sufficient evidence, testable mechanisms, and implementable conclusions.”\n\n---\n\n### 1. Fact Reconstruction and Measurement Conventions (30 points)\n\n#### (1) Accuracy of definitions for basis and implied repo rate (IRR) (10 points)\n- 9–10: Clearly defines **Treasury basis** (price/yield differential between futures net price/dirty price and the deliverable bond/CTD, or an equivalent return differential). Provides an **IRR computation framework** (including coupon/accrued interest, financing cost, delivery date/holding period, conversion factor, and delivery-option effects). Explains the interpretation of **IRR vs SOFR/GC repo**. States the contracts used (e.g., TU/TY/US/Ultra 10Y) and the CTD selection rule.\n- 6–8: Definitions are broadly correct, but explanations of CTD/conversion factor/delivery option or accrued-interest treatment are incomplete, reducing reproducibility.\n- 0–5: Confuses or mis-defines basis, IRR, and repo rates; conventions are incorrect and cannot support subsequent comparisons or conclusions.\n\n#### (2) Completeness and comparability of cross-tenor reconstruction (10 points)\n- 9–10: Covers **2Y–10Y and the ultra-long end (Ultra 10Y)**. Reports widening **magnitude** (bp/price points/IRR minus SOFR spread), **duration**, and **peak/mean-reversion timing**. Explains differences across tenors (e.g., duration, deliverable-basket structure, hedging-demand differences). Maintains consistent conventions and a clear timeline throughout.\n- 6–8: Covers major tenors but omits Ultra 10Y or provides insufficient cross-tenor comparison; or relies mainly on qualitative statements without key statistics.\n- 0–5: Reconstruction is fragmented, focusing on only a few contracts or isolated dates; fails to form a coherent “event profile.”\n\n#### (3) Benchmarking against historical stress episodes and arguing “similarities/differences” (10 points)\n- 9–10: Benchmarks against 2–3 representative historical periods (e.g., **2019 repo stress, the 2020/03 Treasury liquidity crisis, basis disturbances driven by delivery periods/CTD switches**), ensuring consistent measurement conventions. Clearly contrasts this episode with history in terms of **funding conditions** (SOFR/GC, fails, haircut/margin), **intermediation capacity** (dealer balance sheet), and **levered-fund behavior** (HF net positions/open interest), supporting a judgment on whether the event is “systemic.”\n- 6–8: References historical episodes but uses few benchmark metrics, inconsistent conventions, or remains largely narrative.\n- 0–5: No historical benchmarking, or only subjective analogies such as “looks/doesn’t look like 2020.”\n\n---\n\n### 2. Mechanism Explanation and Participant Behavior (30 points)\n\n#### (1) Closed-loop transmission chain and testability (10 points)\n- 9–10: Presents a clear closed loop: **tariff shock → volatility/risk appetite shift → changes in funding conditions and margin/haircuts → crowded basis trades and deleveraging → constrained dealer intermediation → wider basis/higher IRR → mean reversion after stabilizing factors emerge**. Each link is supported by observable evidence or proxy variables (e.g., repo spreads, futures OI, CFTC levered-fund net positions, bid-ask spreads/market depth).\n- 6–8: Chain is mostly complete, but a key link lacks evidence (common gaps: how funding constraints transmit into futures pricing, or quantitative traces of dealer constraints).\n- 0–5: Merely lists possible reasons without causal structure or a verifiable pathway.\n\n#### (2) Hedge-fund deleveraging and basis-trade mechanism (10 points)\n- 9–10: Accurately describes the canonical **cash–futures basis trade** (long cash bonds, short futures, repo funding/leverage). Explains sensitivity to **IRR, margin, haircuts, volatility, and funding availability**. Uses data/facts (e.g., changes in levered-fund net positions, futures OI, repo volumes and term structure changes) to argue the marginal contribution of “deleveraging/crowded trades” to basis widening, and explains why the basis later repairs once conditions stabilize.\n- 6–8: Mechanism is correct but evidence is thin, or the “why it normalizes” trigger is not explained.\n- 0–5: Gets the deleveraging–basis relationship backwards/incorrect, or stays at slogan level.\n\n#### (3) Dealer intermediation capacity combined with funding-market dynamics (10 points)\n- 9–10: Jointly explains **dealer balance sheet costs/constraints** (e.g., balance sheet usage, directionality of regulatory constraints, risk limits) and **funding conditions** (SOFR, GC, term repo, specialness, fails, auction/settlement-related cash needs). Distinguishes **OTR vs off-the-run** effects and deliverable-basket liquidity differences on basis. Supports the “post-stabilization repair” with evidence consistent with improved intermediation/funding conditions.\n- 6–8: Mentions dealer and funding factors but lacks an integrated analysis, or omits key micro-mechanisms (specialness, fails, OTR/off-the-run).\n- 0–5: Over-generalizes to “poor liquidity/dealers step away” without specific constraints, indicators, or evidence.\n\n---\n\n### 3. Stress-Type Assessment, Policy Implications, and Risk-Management Takeaways (25 points)\n\n#### (1) An operational framework to distinguish “temporary dislocation” vs “structural liquidity stress” (10 points)\n- 9–10: Proposes an actionable diagnostic framework with indicators covering at least: **price layer** (basis, IRR–SOFR, abnormal swap spreads/curve spreads), **liquidity layer** (bid-ask, depth, volume/price impact), **funding layer** (repo spreads, specialness, fails, haircut/margin), and **positioning/intermediation layer** (OI, CFTC positioning, dealer-inventory signals). Provides an integrated conclusion and explicit uncertainty boundaries.\n- 6–8: Discusses both interpretations but lacks a systematic indicator set or thresholds/benchmarks; persuasion is moderate.\n- 0–5: Draws conclusions from a single symptom (e.g., “IRR above SOFR”) without cross-validation.\n\n#### (2) Professionalism of policy implications and market-stability discussion (8 points)\n- 7–8: When discussing policy, clearly specifies “tool → channel → side effects”: e.g., potential pathways involving **repo backstop facilities, clearing and trading infrastructure, dealer-making and capital constraints, Treasury issuance and liquidity supply**. Avoids treating policy as a panacea. Identifies concrete vulnerabilities in rates-market stability (e.g., procyclicality of levered trades, margin spirals).\n- 4–6: Offers several policy points but lacks transmission mechanisms or discussion of costs/constraints.\n- 0–3: Policy discussion is vague, disconnected from basis mechanisms, or clearly unprofessional.\n\n#### (3) Actionability of risk-management implications (7 points)\n- 6–7: Provides implementable risk controls covering at least: **funding and margin management (haircut/margin buffers)**, **liquidity reserves and stress-test metrics**, **scenario/sensitivity analysis for basis positions (CTD switches, delivery options, repo specialness)**, and **stop-loss/deleveraging triggers**. Differentiates guidance for hedge funds, dealers, and asset managers/hedgers.\n- 3–5: Directionally correct but too principles-based; lacks metric- or process-level specification.\n- 0–2: Only generic slogans such as “control risk, reduce leverage.”\n\n---\n\n### 4. Data and Methodological Rigor, Presentation, and Reproducibility (15 points)\n\n#### (1) Authority, transparency, and timeliness of data sources (6 points)\n- 6: Clearly lists data sources and frequencies (e.g., futures prices and contract specs, deliverables and CTD, SOFR/GC/term repo, trading and liquidity proxies, positioning data), the sample window, and key event dates. Distinguishes primary vs secondary data and potential biases.\n- 3–5: Data sources are broadly reliable, but some key fields are missing (e.g., CTD conventions, delivery-date handling) or citations are insufficiently transparent.\n- 0–2: Sources are unclear, timestamps are inconsistent, or the analysis relies heavily on unverified narrative.\n\n#### (2) Method disclosure, reproducibility, and robustness checks (5 points)\n- 5: Discloses key computation steps (basis/IRR, cashflow treatment, delivery/funding assumptions) and performs at least one robustness check (e.g., alternative funding-rate assumptions, sensitivity to CTD changes, removing extreme quotes/liquidity adjustment). Charts and conclusions are consistent.\n- 3–4: Provides partial logic, but key parameters/assumptions are not disclosed, making reproduction difficult.\n- 0–2: No method disclosure; conclusions are not verifiable.\n\n#### (3) Report structure, argumentative clarity, and professional writing standards (4 points)\n- 4: Clear structure (executive summary → facts → mechanisms → assessment → implications/risk management → appendix). Uses concepts rigorously; charts are fully labeled. Clearly distinguishes facts, inference, and assumptions; conclusions include conditions and limitations.\n- 2–3: Readable overall, but terminology is imprecise, chart–text linkage is weak, or conclusion boundaries are unclear.\n- 0–1: Disorganized structure; frequent misuse of key concepts undermines readability and credibility.\n"
  },
  {
    "id": "15",
    "question": "With the rapid expansion of fiat-backed stablecoins (e.g., USDT, USDC), whose total market capitalization has exceeded USD 200 billion, most commercial banks remain in a wait-and-see stance. The core question is: under what conditions will the growth of stablecoins pose a material challenge to the bank deposit system, and how should banks respond strategically?\n\nFocusing on the development of stablecoins in the United States and major international markets, and based on publicly available research and regulatory frameworks, conduct a systematic analysis from a financial system perspective, addressing the following tasks:\n\n1. Threshold analysis:\nWhat key conditions (e.g., scale, regulatory clarity, yield competitiveness) are required for stablecoins to transition from a niche tool to systemically important financial infrastructure? Is the United States currently on this trajectory? What lessons can be drawn from comparable historical episodes, such as the expansion of money market funds or the adoption of payment platforms like PayPal?\n\n2. Strategic dilemma for banks:\nBanks face a strategic paradox: ignoring stablecoins risks disintermediation, while actively engaging may expose them to regulatory and reputational constraints. In this context, what is the likely optimal strategy for large U.S. commercial banks, and what are the underlying logic and constraints?\n\n3. Regulatory calibration:\nProposed U.S. stablecoin legislation requires 1:1 reserve backing. How would this requirement interact with bank liquidity regulations (e.g., LCR and NSFR)? Would it effectively mitigate systemic risk, or merely redistribute risk within the financial system (e.g., concentrating risk in reserve asset markets)?",
    "classification": "Banking-Treasury & ALM",
    "classification_code": "BNK-TSY",
    "report_type": "Assessment Report on the Impact of Stablecoins on Bank Liquidity Management and Balance Sheet Structure",
    "report_type_zh": "稳定币发展对银行流动性管理与资产负债结构影响评估报告",
    "language": "en",
    "expert_evaluation_criteria": "## Evaluation Criteria for *Stablecoin Expansion, Bank Deposit Disintermediation, and Strategic Response*\n\nThis framework evaluates deep research reports on stablecoin growth, bank disintermediation risk, and regulatory calibration. The focus is whether the report identifies the tipping point at which stablecoins materially affect bank deposits, analyzes banks’ strategic choices and constraints, and assesses the interaction between stablecoin regulation and bank liquidity rules. Total: 100 points.\n\n---\n\n### 1. Tipping-Point Analysis of Stablecoin Impact on Banks (30 points)\n\n#### (1) Adoption conditions and transition mechanism (12 points)\n- **10–12**: Clearly identifies key conditions such as scale, regulatory clarity, yield competitiveness, payment use cases, and user adoption, and explains how stablecoins could evolve into systemically important infrastructure.\n- **7–9**: Covers major conditions, but the mechanism is incomplete.\n- **0–6**: Lists conditions without a clear logic chain.\n\n#### (2) Transmission path to bank deposit substitution (10 points)\n- **9–10**: Clearly explains how stablecoins may affect demand deposits, payments, and bank funding costs, distinguishing gradual substitution from sudden shocks.\n- **6–8**: Identifies the direction of impact, but the transmission mechanism is rough.\n- **0–5**: Lacks substantive analysis of bank liability-side impact.\n\n#### (3) Historical analogy and current-stage judgment (8 points)\n- **7–8**: Uses cases such as money market funds and PayPal effectively and gives a clear judgment on the current U.S. stage.\n- **4–6**: Provides analogy, but with limited relevance or weak conclusions.\n- **0–3**: Lacks historical reference or stage assessment.\n\n---\n\n### 2. Bank Strategic Response Analysis (30 points)\n\n#### (1) Identification of strategic options (10 points)\n- **9–10**: Clearly compares key paths such as waiting, defensive response, partnership, and in-house deployment, with relevant conditions for each.\n- **6–8**: Identifies major options, but comparison is not systematic.\n- **0–5**: Thin strategic analysis without a clear framework.\n\n#### (2) Optimal strategy and constraint analysis (12 points)\n- **10–12**: Explains the relatively optimal strategy for large U.S. banks from the perspectives of regulation, reputation, funding stability, customer relationships, and technical capability, while defining its limits.\n- **7–9**: Offers a clear view, but constraint analysis is incomplete.\n- **0–6**: Gives conclusions without sufficient support.\n\n#### (3) International comparison and lessons (8 points)\n- **7–8**: Draws on major international markets and extracts relevant implications for bank strategy.\n- **4–6**: Includes some international perspective, but comparison is shallow.\n- **0–3**: Lacks international comparison.\n\n---\n\n### 3. Regulatory Calibration and Systemic Risk Assessment (25 points)\n\n#### (1) Understanding of the stablecoin regulatory framework (8 points)\n- **7–8**: Accurately explains core requirements such as 1:1 reserves, reserve asset scope, and redemption rules.\n- **4–6**: Mostly correct, but incomplete.\n- **0–3**: Contains major omissions or misunderstandings.\n\n#### (2) Interaction with bank liquidity regulation (10 points)\n- **9–10**: Clearly analyzes the relationship between reserve rules and LCR/NSFR, and the implications for bank liquidity management and liability structure.\n- **6–8**: Identifies the main interaction, but lacks depth.\n- **0–5**: Fails to discuss the mechanism effectively.\n\n#### (3) Systemic risk reallocation judgment (7 points)\n- **6–7**: Assesses whether regulation truly reduces risk or merely shifts it toward reserve assets, liquidity concentration, or market structure.\n- **3–5**: Provides risk judgment, but not systematically.\n- **0–2**: Lacks a systemic-risk perspective.\n\n---\n\n### 4. Report Quality and Strategic/Policy Conclusions (15 points)\n\n#### (1) Structure and professionalism (5 points)\n- **5**: Clear structure, sound logic, and strong financial-system research quality.\n- **3–4**: Basically complete, but average in organization.\n- **0–2**: Weak structure and professionalism.\n\n#### (2) Actionability of conclusions (6 points)\n- **5–6**: Provides clear and actionable implications for bank strategy or regulation.\n- **3–4**: Clear conclusions, but only moderately actionable.\n- **0–2**: Vague conclusions.\n\n#### (3) Risk disclosure and boundary conditions (4 points)\n- **4**: Covers key risks such as regulatory change, weaker-than-expected adoption, credit events, and liquidity shocks, with clear transmission paths.\n- **2–3**: Lists major risks, but without depth.\n- **0–1**: Missing or formulaic risk disclosure."
  },
  {
    "id": "16",
    "question": "JPMorgan Chase’s San Francisco branch is conducting a new round of refined client coverage and comprehensive financial service assessment for its key corporate clients, including the technology industry leader Salesforce, Inc. and its legal representative, Marc Benioff.\n\nBased on this business context, please complete the following tasks:\n\n1. Using publicly available information, construct a comprehensive client profile of Salesforce and its legal representative Marc Benioff (including corporate attributes, industry characteristics, financial and operating features, and personal influence);\n2. Based on the client profile, analyze the client’s deeper potential needs across corporate banking services, cross-border finance, investment and financing, and high-end personal financial services;\n3. From the perspective of a major U.S. commercial bank’s San Francisco branch, propose actionable and implementable client coverage and product allocation recommendations.",
    "classification": "Banking-Customer & Marketing Management",
    "classification_code": "BNK-CMM",
    "report_type": "Corporate Client Profiling and Integrated Financial Services Solution",
    "report_type_zh": "对公客户画像与综合金融服务方案",
    "language": "en",
    "expert_evaluation_criteria": "## Evaluation Criteria for JPMorgan Chase San Francisco Branch — Key Corporate Client (Salesforce & Marc Benioff): Refined Account Management and Comprehensive Financial Services Solution\n\nThe framework aims to deliver a systematic, executable, and quantitative evaluation of (i) an open-source intelligence (public information) profile around Salesforce (the enterprise) and Marc Benioff (the key individual), (ii) needs insights, and (iii) an operations/product implementation plan from the branch perspective. The rubric contains **4 primary dimensions and 14 secondary dimensions, totaling 100 points**, and is designed to support consistency between human review and LLM-based automated scoring.\n\n---\n\n### 1. Overall Client Profiling Quality (30 points)\n\n#### (1) Completeness and Structure of the Company’s Baseline Profile (8 points)\n- **7–8 points**: Clearly covers company nature and governance (listing status, key points on ownership/board structure), core business and product portfolio, customer mix and business model (key features such as subscription/renewals/deferred revenue or contract liabilities), geographic footprint and organizational structure; information is well-layered and can be directly mapped into bank CRM fields.\n- **4–6 points**: Covers major elements but with loose structure; some key fields are missing (e.g., revenue composition, customer concentration, geographic/industry exposure).\n- **0–3 points**: Generic description or remains at “company overview” level; lacks structured, actionable points usable for client management.\n\n#### (2) Industry and Competitive Landscape Insights (7 points)\n- **6–7 points**: Based on the company’s arena (enterprise software/CRM/cloud ecosystem/AI overlays, etc.), provides verifiable industry characteristics, growth drivers, cycles, and risks; clearly identifies key competitors and substitution relationships, and explains implications for bank credit, cash flows, and financing/investment needs.\n- **3–5 points**: Industry description is broadly correct, but lacks competitive landscape, cyclicality, or mapping to financial needs.\n- **0–2 points**: Industry view is vague or inconsistent with the company’s reality.\n\n#### (3) Financial and Operating Characteristics (10 points)\n- **9–10 points**: Uses public financial disclosures (e.g., 10-K/10-Q, investor presentations, earnings call commentary) to characterize revenue growth and mix, profitability/cash-flow quality, balance-sheet highlights (cash and investments, debt structure, buyback/dividend policy, etc.), and key operating metrics (RPO/contract liabilities/retention and customer expansion, where applicable); identifies key trend changes and explains drivers.\n- **5–8 points**: Uses some financial data but definitions are unclear or lacks cash-flow/balance-sheet perspective; insufficient trend explanation.\n- **0–4 points**: Almost no quantitative support or sources are unclear; conclusions are not verifiable.\n\n#### (4) Key Individual (Marc Benioff) Profile and Influence Assessment (5 points)\n- **5 points**: Clearly states his role and influence pathways in governance/strategy (title, decision power, public persona, major philanthropy/social initiatives footprint, etc.), and builds an “explainable linkage” to potential private banking/family governance needs without over-inference.\n- **2–4 points**: Basic introduction is present, but influence assessment or linkage to financial needs is weak.\n- **0–1 point**: No meaningful key-individual profile, or heavy speculation/contradiction with public information.\n\n---\n\n### 2. Potential Needs Insights and Opportunity Map (30 points)\n\n#### (1) Identification of Core Corporate Banking and Transaction Banking Needs (10 points)\n- **9–10 points**: Starting from cash-flow and operating scenarios, systematically identifies needs and prioritizes them by layers: cash management/liquidity, collections & payments and account structure, centralized payables, supplier payments and receivables, payroll and benefits-related services, short-term investments and liquidity management; clearly specifies “demand triggers” (e.g., M&A, international expansion, interest-rate regime shifts, buyback cadence, etc.).\n- **5–8 points**: Covers several product points but lacks scenario-based decomposition or clear prioritization/triggers.\n- **0–4 points**: Only lists products; lacks linkage to operating facts.\n\n#### (2) Cross-Border Finance and Global Treasury Needs (8 points)\n- **7–8 points**: Based on geographic revenue/cost structure, global entity footprint, and potential currency exposures, identifies needs such as FX/hedging, cross-border cash pooling, global account structures, tax and fund transfer constraints, cross-border collections and local clearing; highlights compliance and operating frictions (e.g., sanctions/AML, data and privacy, cross-border path complexity).\n- **4–6 points**: Mentions cross-border needs but lacks specificity on exposure sources, currencies/regions, or compliance constraints.\n- **0–3 points**: Cross-border section is generic or inconsistent with the client’s globalization profile.\n\n#### (3) Deeper Needs in Financing/Investment, Capital Markets, and M&A (8 points)\n- **7–8 points**: From perspectives including capital structure, valuation/buybacks, potential M&A strategy, convertible/bond financing, equity incentives and employee liquidity, proposes actionable financing/investment opportunities; distinguishes “branch-executable leads” vs. “items requiring investment bank/head office coordination,” and provides trigger signals and information sources.\n- **4–6 points**: Identifies some financing/M&A needs but lacks linkage to financial facts or lacks executable triggers.\n- **0–3 points**: Only generic talk of “financing/IPO/M&A,” with no evidence chain or obvious inapplicability.\n\n#### (4) High-Net-Worth Personal Finance and Family Governance Needs (4 points)\n- **4 points**: Within compliance and information-boundary constraints, proposes potential needs matching the key individual’s characteristics (wealth preservation, tax and philanthropy structuring, trusts/family governance, concentrated holding risk management, concentration and liquidity planning, etc.), and clearly states boundaries of “what the bank can/cannot do” plus trigger conditions.\n- **2–3 points**: Provides directions but lacks boundaries, triggers, or verifiable signals.\n- **0–1 point**: Oversteps boundaries, involves inaccessible private information, or fully ignores individual-side needs.\n\n---\n\n### 3. Branch-Perspective Operating Strategy and Implementability of Product Configuration (25 points)\n\n#### (1) Client Coverage Model and Collaboration Mechanism Design (8 points)\n- **7–8 points**: Clearly defines the San Francisco branch coverage approach (RM/industry team/product specialists), coordination mechanisms with head office/investment bank/transaction banking/private bank, client engagement paths (executive-level dialogue, CFO, treasury/tax/legal counterparts, etc.), and explains role division and execution cadence.\n- **4–6 points**: Shows collaboration intent but remains conceptual; lacks role clarity or execution mechanism.\n- **0–3 points**: Only starts from “what the bank can offer,” with no organizational/collaboration design.\n\n#### (2) Product Portfolio and Solution Fit (10 points)\n- **9–10 points**: Provides a “combinable and packageable” solution (transaction banking + credit + cross-border + capital markets/risk management + properly bounded coordination with HNW/personal services), and explains for each product the value proposition, prerequisites, customer benefits, and bank economics; avoids unrealistic one-size-fits-all “full bundle.”\n- **5–8 points**: Product recommendations are broadly reasonable but lack bundling logic, prerequisites, or benefit/cost trade-offs.\n- **0–4 points**: Product piling; mismatch with the client profile; or clearly not implementable.\n\n#### (3) Execution Plan, Milestones, and Quantifiable KPIs (7 points)\n- **6–7 points**: Provides a 30/60/90-day or quarterly execution plan including key meetings, document checklist, pilot scenarios, and implementation pathways; sets quantifiable KPIs (e.g., cash management penetration, FX volume, credit utilization, number of cross-border accounts implemented, wallet share, etc.) and a review/feedback loop.\n- **3–5 points**: Has action items but timeline/KPIs are unclear, or lacks a closed-loop review mechanism.\n- **0–2 points**: Only directional suggestions; no plan or measurement standard.\n\n---\n\n### 4. Data Credibility, Compliance/Risk Control, and Communication Standards (15 points)\n\n#### (1) Reliability of Sources, Citations, and Consistency of Definitions (6 points)\n- **5–6 points**: Key facts are traceable to credible sources (SEC filings, financial statements, authoritative media/research institutions, etc.) with clear citations; metric definitions (GAAP vs. non-GAAP, TTM vs. fiscal year, etc.) are properly specified; avoids misattribution.\n- **3–4 points**: Sources are generally reliable but citations are incomplete, or there is minor risk of mixed definitions.\n- **0–2 points**: Large amount of unsourced data or clear mis-citations that undermine credibility.\n\n#### (2) Verifiability of Reasoning Chain and Awareness of “Public Information Boundary” (5 points)\n- **4–5 points**: Clearly distinguishes facts, inferences, and assumptions; provides supporting evidence and verifiable signals for inferences; does not use private/internal information to force conclusions; complies with research ethics and client information boundaries.\n- **2–3 points**: Some inferences lack support or boundary reminders are insufficient, but overall acceptable.\n- **0–1 point**: Heavy speculation; treats assumptions as facts; or clearly oversteps by using inaccessible information.\n\n#### (3) Identification of Compliance, Reputational Risk, and Conflicts of Interest (4 points)\n- **4 points**: Clearly identifies and addresses key risk points: KYC/AML, sanctions and cross-border compliance, MNPI and investment-banking wall, client privacy and data security, ESG/reputational risk, related-party transactions and conflicts of interest; explains constraints on the solution and mitigation measures.\n- **2–3 points**: Mentions compliance risks but lacks specificity or does not reflect practical impacts on the plan.\n- **0–1 point**: Ignores compliance/reputational risks, or recommendations contain obvious compliance issues."
  },
  {
    "id": "17",
    "question": "With the continuous advancement of banking digitalization and financial technology, digital operations have become a core capability connecting customers, products, channels, and services. Centered on the customer lifecycle, banks are gradually building digital operation systems characterized by data-driven decision-making, process automation, and refined management. However, the actual effectiveness and inherent constraints of these systems remain to be systematically evaluated.\n\nPlease conduct an analysis of the banking digital operation system and, based on concrete case studies, produce a structured research report addressing the following aspects:\n\n1. Current Practices and Implementation  \nSystematically review current digital operation practices across key stages such as customer acquisition, activation, conversion, and retention. Analyze how data-driven approaches, process automation, and refined management are applied in these contexts.\n\n2. Effectiveness and Challenges  \nEvaluate the effectiveness of these digital operation practices in improving customer value, operational efficiency, and service experience, and identify the main constraints and challenges they face.\n\n3. Optimization Pathways and Future Directions  \nConsidering technological advancements and evolving customer behavior, propose optimization directions and actionable pathways for future banking digital operation models from the perspectives of operational models, organizational mechanisms, and system capabilities.",
    "classification": "Banking-Customer & Marketing Management",
    "classification_code": "BNK-CMM",
    "report_type": "Banking Digital Operations Research Report",
    "report_type_zh": "银行数字化运营研究报告",
    "language": "en",
    "expert_evaluation_criteria": "## Evaluation Criteria for Research Report on a Bank’s Digital Operations System for Online Financial Business\n\nThis evaluation framework is used to provide a unified, executable, and quantitative assessment of analytical reports on a “bank’s digital operations system for online financial business,” balancing consistency between human review and large-model (LLM) automated scoring. The framework contains **4 primary dimensions and 12 secondary dimensions**, with a total score of **100 points**. Scoring focuses on: whether the end-to-end lifecycle closed loop is complete; whether evaluation is verifiable; whether solutions are implementable in real banking scenarios; and whether the optimization path has actionable implementation granularity and risk/compliance awareness.\n\n---\n\n### 1. Quality of Operations Framework and Current-Practice Mapping (30 points)\n#### (1) Completeness of Full Lifecycle and Closed-Loop Framework (10 points)\n- **9–10 points**: Builds a clear closed loop (e.g., “data collection/governance → insights & segmentation → outreach & journey orchestration → conversion & monetization/operations → retention & repeat purchase/cross-sell → review & iteration”), and maps it to typical online-finance business journeys (account opening, card binding, credit approval/loan drawdown & utilization, wealth product subscription, payment activation, etc.); identifies key inputs/outputs and metric definitions for each stage.\n- **6–8 points**: Framework is largely complete, but provides insufficient detail for stages such as review/iteration, post-investment/post-loan operations, and cross-channel collaboration; or mapping to online-finance business characteristics is weak.\n- **0–5 points**: Fragmented framework that remains at concept listing (e.g., “use big data/AI for operations”), lacking an end-to-end journey and an executable structure.\n\n#### (2) Depth of Practice Decomposition for Acquisition (10 points)\n- **9–10 points**: Covers and differentiates major acquisition methods and channels (owned app/mini-program, online account opening, payment scenarios, content/live streaming, external platform partnerships, offline-to-online traffic, etc.); explains the funnel from **spend/delivery → reach → account opening/KYC → first transaction**; discusses cost and quality (CAC, account-opening conversion rate, anti-fraud/anti-abuse) and segment differences.\n- **6–8 points**: Describes main tactics, but lacks funnel decomposition, channel differentiation, or quality controls (e.g., only discusses acquisition but not anti-fraud and “effective accounts”).\n- **0–5 points**: Only lists channel buzzwords or copies generic internet growth jargon, lacking banking constraints and key steps in acquisition.\n\n#### (3) Depth of Practice Decomposition for Activation, Conversion, and Retention (10 points)\n- **9–10 points**: Separately provides concrete operational actions and trigger mechanisms for activation (MAU/DAU, key behavior completion), conversion (first purchase/credit utilization, cross-sell, asset uplift), and retention (repeat purchase, churn warning, lifecycle operations), such as journeys/event triggers, contact frequency capping, benefits/loyalty system, and relationship-manager collaboration; reflects differences by segment/product line (deposits/loans/FX/wealth, credit cards, wealth management, funds, insurance, etc.).\n- **6–8 points**: Covers most links, but the linkage mechanism from “activation → conversion → retention,” product-specific strategies, or trigger rules are not concrete enough.\n- **0–5 points**: Incomplete coverage or generic content; lacks executable operational actions and scenario-based description.\n\n---\n\n### 2. Data-Driven and Fine-Grained Operations Effectiveness Evaluation (25 points)\n#### (1) Fit of Data Foundations, Tags/Profiles, and Data Governance (8 points)\n- **7–8 points**: Clearly defines data sources and boundaries (transactions/behavior/channels/marketing touchpoints/risk control/customer service, etc.); explains unified ID, data quality, master data management, real-time/offline pipelines; tags/profiles map to bank-accessible data and compliance requirements (data minimization, consent/authorization, audit trails).\n- **4–6 points**: Mentions tags/profiles and governance, but lacks key implementation details (e.g., ID stitching, consistent metric definitions, real-time requirements) or has weak compliance boundary discussion.\n- **0–3 points**: Unclear data sources; overly generic profiling; or proposes clearly unavailable/non-compliant data usage.\n\n#### (2) Verifiability of KPI System and Evaluation Methods (9 points)\n- **8–9 points**: Establishes a layered metric system (business outcome metrics + process metrics + quality/risk metrics) and provides verifiable methods: funnel analysis, cohort/segment control comparisons, A/B testing or quasi-experiments, incremental attribution (multi-touch/marketing attribution), long-term value metrics (LTV/CLV, net AUM growth, risk cost); clarifies metric definitions and observation windows.\n- **4–7 points**: Has metrics and methods, but metric definitions/windows/control-group setup are unclear; or uses correlation narratives instead of causal evaluation.\n- **0–3 points**: Lacks a metric system and evaluation methods; effectiveness judgments are mainly subjective.\n\n#### (3) Effectiveness and Reproducibility of Fine-Grained Strategies (8 points)\n- **7–8 points**: Explains how fine-grained strategies drive outcomes: tiering/segmentation, personalized recommendation/ranking, customer journey orchestration, frequency capping and fatigue management, benefit/cost control; discusses trade-offs between “short-term conversion” and “long-term retention/experience,” and proposes strategy modules that can be replicated across multiple product lines.\n- **4–6 points**: Has fine-grained ideas, but lacks strategy granularity (missing trigger conditions, frequency rules, cost constraints) or has weak reproducibility.\n- **0–3 points**: Stays at slogan level (e.g., “precision marketing/personalization for everyone”), lacking details and constraints.\n\n---\n\n### 3. Process Automation, System Capabilities, and Organizational Mechanisms (20 points)\n#### (1) Coverage and Business Value of Process Automation (8 points)\n- **7–8 points**: Clearly decomposes automatable objects (automated journey triggers, lead routing, outbound calling/message orchestration, operations dashboards and alerts, RPA handling, ticket/work-order routing, etc.), and evaluates impact on efficiency, timeliness, experience, and risk (productivity gains, shorter SLA, fewer complaints, etc.).\n- **4–6 points**: Lists automation scenarios, but lacks quantified business value, or ignores key nodes requiring human involvement (compliance review, relationship-manager service).\n- **0–3 points**: Equates automation with “fully unmanned,” ignoring real constraints of banking service and risk control.\n\n#### (2) Match Between System Architecture and Required Capabilities (7 points)\n- **6–7 points**: Breaks down system capabilities and dependencies: CDP (customer data platform), CRM, marketing automation/journey orchestration, recommendation engine, content platform, risk control and anti-fraud, data platform/real-time computing, permissions and audit; explains integration with legacy core systems and channels (app/mini-program/SMS/enterprise WeChat, etc.) and key challenges.\n- **3–5 points**: Mentions system modules, but architecture relationships, data flows/APIs, real-time and stability requirements are unclear.\n- **0–2 points**: Generic system description; lacks real banking IT constraints (transformation cycles, interfaces, data consistency, etc.).\n\n#### (3) Organization, Governance, and Collaboration (5 points)\n- **4–5 points**: Clearly defines operations-role division (HQ/branch; product/operations/data/technology/risk/compliance/customer service; relationship-manager collaboration), processes and access rights, approval and audit-trail mechanisms, KPIs and incentives (balancing scale and quality), and cross-line collaboration mechanisms.\n- **2–3 points**: Mentions organizational collaboration but remains general, without clarifying key mechanisms (e.g., access/approval/performance conflicts).\n- **0–1 point**: Ignores organizational governance and attributes issues entirely to technology or data.\n\n---\n\n### 4. Gap Diagnosis, Trend Integration, and Implementable Optimization Path (25 points)\n#### (1) Depth of Diagnosing Performance Gaps and Underlying Constraints (8 points)\n- **7–8 points**: Systematically identifies and explains sources of constraints and impact chains: data silos and inconsistent definitions, insufficient real-time capability, model non-interpretability/bias, outreach noise and customer fatigue, channel fragmentation, conflicts between operations and risk-control goals, compliance restrictions (privacy, consent, suitability), system debt and delivery cycles, etc.\n- **4–6 points**: Points out problems but lacks root-cause analysis or impact mechanisms; or only describes phenomena without constraint boundaries.\n- **0–3 points**: Overly generic diagnosis, lacking real constraints in a banking context.\n\n#### (2) Degree of Linking Tech Trends and Customer Behavior Changes (8 points)\n- **7–8 points**: Clearly ties trends to operations transformation: boundaries and risk control for AIGC in content production/customer service/investor education; real-time data and event-driven operations; privacy computing/federated learning for data collaboration; MLOps/feature platforms improving model iteration efficiency; effects of customer behaviors (cross-device, multi-touchpoint, low patience, preference for immediate feedback) on journey design and frequency capping.\n- **4–6 points**: Mentions trends but links weakly to specific operational steps, or lacks actionable landing points.\n- **0–3 points**: Lists trends without direct relevance to optimizing the operations system in this topic.\n\n#### (3) Executability of Optimization Directions and Implementation Roadmap (9 points)\n- **8–9 points**: Provides an implementable path and phased objectives (e.g., 0–3 months / 3–6 months / 6–12 months), milestone deliverables, resources and dependencies (data, systems, organization), cost-benefit and key risk controls; clarifies prioritization and the sequencing logic of “what first, what next” (e.g., unify ID and metric definitions → build journey orchestration and experimentation system → build intelligent recommendation and closed-loop optimization).\n- **4–7 points**: Proposes directions and some measures, but lacks a phased plan, assessment of resource constraints, or quantified benefit expectations.\n- **0–3 points**: Remains at vision-level recommendations, lacking implementable steps and constraint handling."
  },
  {
    "id": "18",
    "question": "Please assess how the leading U.S. global asset management firms—BlackRock, Vanguard, Fidelity Investments, State Street Global Advisors, and Capital Group—have implemented requirements related to the fair value measurement of financial instruments, in line with the U.S. accounting and regulatory framework, including ASC 820 (Fair Value Measurement) and relevant SEC and Federal Reserve guidance on valuation and disclosure.\n\nDrawing on publicly available information such as Form 10-K filings, regulatory disclosures, and authoritative ratings or research reports, conduct a systematic evaluation of each firm’s overall compliance and practice. On this basis, compare similarities and differences across firms in areas including valuation governance structures, valuation methodologies and fair value hierarchy classification, model usage and independent validation, and the adequacy and transparency of disclosures. Identify potential risk areas and gaps for improvement, and propose practical, actionable recommendations to enhance valuation governance and risk management.",
    "classification": "Banking-Risk Management",
    "classification_code": "BNK-RSK",
    "report_type": "Fair Value Measurement Compliance Assessment Report",
    "report_type_zh": "金融工具公允价值估值合规评估报告",
    "language": "en",
    "expert_evaluation_criteria": "## Scoring Rubric for Compliance and Disclosure Benchmarking Research on Fair Value Valuation of Financial Instruments\n\nThis rubric is designed to quantitatively assess the professional quality and executability of a research report on 'Fair value valuation practices of the top five U.S. global financial asset management institutions'. It supports both human review and LLM-based automated scoring. The rubric contains 4 primary dimensions and 13 secondary dimensions, with a total score of 100.\n\n### 1. Accuracy of Accounting Standards and Regulatory Framework Mapping (25 points)\n\n#### (1) Mastery of ASC 820 core requirements and accuracy of citations (10 points)\n- 9-10: Accurately covers and correctly applies key ASC 820 concepts and disclosure points (e.g., exit price, principal/most advantageous market, boundaries for unit of account, the three-level input hierarchy, the three categories of valuation techniques, calibration and consistency, Level 3 sensitivity and rollforward table disclosures, etc.), and can map them item-by-item to the sampled institutions' disclosures with corresponding report sections/note topics; no major conceptual errors.\n- 6-8: Covers major concepts and the hierarchy framework, but is not sufficiently precise on disclosure details, calibration/assessments of input observability, or Level 3 requirements; or contains a small number of correctable deviations.\n- 0-5: Demonstrates clearly incomplete understanding or critical misinterpretations (e.g., mechanically equating Levels 1/2/3 with asset classes, confusing measurement attributes with valuation methods, or failing to explain the logic for determining input observability).\n\n#### (2) Capability to map SEC disclosure requirements to valuation topics (8 points)\n- 7-8: Establishes a clear mapping between the SEC disclosure framework and valuation issues (e.g., fair value disclosures in financial statement notes, significant accounting policies, estimation uncertainty, MD&A, risk factors, audit- and control-related disclosures, etc.), and explains how inter-firm disclosure differences affect investor understandability and comparability.\n- 4-6: Identifies the general direction of SEC disclosures, but mapping is too high-level and does not reach a verifiable level of 'specific disclosure elements/tables/definition differences'.\n- 0-3: Over-generalizes SEC requirements as 'need sufficient disclosure', lacking actionable mappings and supporting evidence.\n\n#### (3) Applicability assessment and boundary clarification for FRB supervisory requirements (7 points)\n- 6-7: Explains differences in applicability of relevant FRB supervisory focus areas across the five institutions (e.g., whether an entity is a bank holding company or within the Federal Reserve supervisory perimeter; shared supervisory expectations around valuation controls/MRM/third-party valuation management), and for institutions where requirements are not applicable or only weakly applicable, provides reasonable alternative evidence paths and clear research boundaries.\n- 3-5: Mentions FRB supervision, but the applicability discussion is relatively generic and does not sufficiently address sample differences or supervisory chain differences.\n- 0-2: Applies FRB requirements indiscriminately to all institutions or fully ignores the prompt requirements.\n\n---\n\n### 2. Depth of Systematic Comparison of the Five Institutions' Valuation Practices (35 points)\n\n#### (1) Comparison of valuation governance structure and division of responsibilities (9 points)\n- 8-9: Clearly compares each institution's valuation governance (board/audit committee/valuation committee, management responsibilities, three lines of defense, segregation between investment teams and valuation/finance/risk functions, use and oversight of external pricing services, etc.), and explains the business/organizational drivers behind differences and potential control implications.\n- 5-7: Compares major governance elements, but lacks depth on responsibility boundaries, independence safeguards, and conflict management.\n- 0-4: Stays at descriptive statements such as 'has a committee/process', lacking substantive comparison and discussion of control effectiveness.\n\n#### (2) Comparability analysis of valuation methods and ASC 820 level classification (9 points)\n- 8-9: Performs verifiable comparisons of valuation techniques and level-classification logic for key asset/instrument types (e.g., trading/available-for-sale investments, derivatives, private/illiquid investments, fund interests, structured products, etc.); identifies key differences such as 'same asset type classified into different levels', 'differences in input observability', and 'level transfers driven by changes in market activity'.\n- 5-7: Covers main methods and levels, but lacks detailed comparison of classification basis, reasons for level transfers, or differences by asset type.\n- 0-4: Only lists Level 1/2/3 proportions or concepts, lacking a causal chain linking method -> inputs -> level.\n\n#### (3) Model use, key assumptions, and independent validation (9 points)\n- 8-9: Identifies model-use scenarios (DCF, option pricing, credit spread/volatility curves, correlation, liquidity discounts, etc.), governance over key unobservable inputs and assumptions; compares independent validation mechanisms (validation frequency, back-testing/benchmarking, price challenge processes, valuation adjustments/reserves, involvement of external auditors/third-party valuation specialists, etc.), and highlights model risk points.\n- 5-7: Describes general model and validation practices, but lacks cross-institution comparison at the 'key assumptions/unobservable inputs' level or has a weak evidence chain.\n- 0-4: Treats 'using models' in a generic way without explaining validation, back-testing, challenge, and adjustment mechanisms.\n\n#### (4) Third-party pricing, price verification, and valuation adjustment mechanisms (8 points)\n- 7-8: Compares differences in pricing sources (exchange quotes, broker quotes, pricing vendors, internal models), price verification, handling of anomalous prices, quote tiering/credibility assessment, valuation adjustments (e.g., liquidity/credit valuation adjustments), and escalation/approval workflows; explains implications for Level 2/3 risk.\n- 4-6: Mentions third-party pricing and verification, but lacks process details, control points, and comparative analysis.\n- 0-3: Missing analysis of pricing sources and verification mechanisms, insufficient to support conclusions on valuation reliability.\n\n---\n\n### 3. Evidence Quality, Data Processing, and Research Verifiability (20 points)\n\n#### (1) Authority of information sources, coverage, and traceable citations (10 points)\n- 9-10: Main conclusions are traceable to public disclosures (10-K, relevant regulatory filings/fund disclosure documents, key audit report points, authoritative ratings/research reports, etc.), with clear citation labels (document type, year, section/note topic); flags differences in disclosure conventions across institutions; avoids substituting secondary interpretations for primary evidence.\n- 6-8: Sources are generally sufficient, but citations for some key comparison points are not precise enough, or there are cases of 'research report paraphrase replacing original disclosure'.\n- 0-5: Sources are unclear, citations are not verifiable, or reliance on non-authoritative materials undermines credibility.\n\n#### (2) Multi-year trend analysis and consistency checks (5 points)\n- 5: Conducts trend analysis across years (e.g., changes in Level 3 scale, level transfers, valuation technique changes, major accounting policy changes, disclosure expansion/contraction) and explains drivers and links to regulatory/market events.\n- 3-4: Includes year-over-year comparisons, but explanations/attribution are weak or cover only a few institutions/years.\n- 0-2: Primarily single-year cross-sectional description, failing to reflect multi-year implementation.\n\n#### (3) Method transparency and reproducible comparison framework (5 points)\n- 5: Clearly states the comparison framework and approach to definition/measurement alignment (e.g., how to handle entities without a 10-K, how to standardize terminology and table definitions, how to choose comparable business/entity scope, how to score or bucket qualitatively), enabling third parties to verify main conclusions.\n- 3-4: Provides partial methodological explanation, but does not clearly justify key definition choices and comparability handling.\n- 0-2: Lacks methodological disclosure; conclusions read as opinion-based and are hard to verify.\n\n---\n\n### 4. Disclosure Adequacy Assessment, Risk Identification, and Actionability of Recommendations (20 points)\n\n#### (1) Assessment of disclosure adequacy and investor understandability (8 points)\n- 7-8: Evaluates 'sufficiency, clarity, and comparability' against ASC 820/SEC disclosure elements, and identifies specific gap types (e.g., insufficient Level 3 input sensitivity disclosures, boilerplate valuation technique descriptions, unclear reasons for level transfers, insufficient disclosure of key estimation uncertainty, significant judgments not echoed in MD&A/risk factors, etc.).\n- 4-6: Identifies strengths/weaknesses, but evaluation criteria and evidence linkage are not tight enough, or lacks specific discussion of understandability/comparability.\n- 0-3: Only provides generic assessments (e.g., 'disclosure is adequate/needs strengthening') without verifiable support.\n\n#### (2) Identification of potential risk points and impact pathways (6 points)\n- 6: Identifies key valuation-related risks (model risk, unobservable inputs, liquidity dry-ups, failure of third-party pricing, conflicts of interest/insufficient independence, control deficiencies leading to misstatement, regulatory review and reputational risk, etc.) and explains impact pathways to financial statements, fair value level distribution, fees/performance, or investor decision-making.\n- 3-5: Identifies major risks but impact mechanisms are incomplete, or not tied to inter-institution differences.\n- 0-2: Risk discussion is a checklist without mechanisms or prioritization.\n\n#### (3) Executability and prioritization of recommendations (6 points)\n- 6: Recommendations are specific and implementable (accountable owners, processes/policies, data and systems, validation frequency and metrics, disclosure templates and examples), and are prioritized (e.g., near-term actions/mid-term optimization/long-term build-out or by risk priority); balances compliance, cost, and organizational feasibility.\n- 3-5: Directionally correct but too principle-based; lacks implementation levers or prioritization.\n- 0-2: Vague recommendations (e.g., 'strengthen governance/improve disclosure') that do not guide execution.\n"
  },
  {
    "id": "19",
    "question": "As R&D investment in the U.S. high-technology sector continues to rise, patents and intellectual property are increasingly evolving from defensive assets into financeable instruments. IP-backed financing, centered on patent collateralization, has begun to emerge among selected technology firms.\n\nFocusing on IP-backed financing practices among U.S. high-technology companies, select NVIDIA, Qualcomm, and Palantir Technologies as representative cases and conduct a systematic analysis of their core technologies and patent assets. The study should address the following tasks:\n\n1. Patent Assets and Financing Foundations\nExamine core technology areas and patent portfolio characteristics, and assess how patents support financing feasibility through technological barriers, industry positioning, and commercialization potential;\n\n2. Patent Valuation Methods and Applicability\nCompare valuation approaches commonly applied to high-technology patents in the U.S. market, including income-based, market-based, and option-based models, and analyze their applicability under technological uncertainty, market volatility, and differing patent life cycles;\n\n3. Collateralization Parameters and Risk Constraints\nAssess patent loan-to-value ratios, monetization potential, and key sources of uncertainty, and identify major risks related to technological substitution, legal protection, valuation volatility, and disposal efficiency;\n\n4. Financing Structure and Mechanism Design\nIn light of R&D intensity, technology iteration cycles, and capital market conditions, explore differentiated credit and financing structures, such as dynamic credit arrangements linked to R&D intensity or patent quality, as well as institutional mechanisms for patent registration, transfer, and disposal through IP exchanges, technology transfer platforms, and legal service providers, to enhance the feasibility and risk controllability of IP-backed financing.",
    "classification": "Banking-Assets",
    "classification_code": "BNK-AST",
    "report_type": "Patent-Backed Financing Assessment and Credit Structuring Report for High-Tech Enterprises",
    "report_type_zh": "高新技术企业专利质押融资评估与授信方案研究报告",
    "language": "en",
    "expert_evaluation_criteria": "## Evaluation Criteria for Research Reports on IP-Backed Financing (Technology Patent Pledge Financing) of U.S. High-Tech Companies\n\nThis scoring rubric is designed to provide an actionable, reproducible, quantitative evaluation of research reports on patent pledge financing (IP-backed financing) for U.S. high-tech companies. The rubric includes 4 primary dimensions and 12 secondary dimensions, with a total score of 100 points. It is compatible with both human review and large language model (LLM) automated scoring.\n\nThe rubric emphasizes: (i) the logic by which patent assets support financing, (ii) the rigor and boundaries of valuation methodologies, (iii) the verifiability of pledge parameters and risk constraints, and (iv) the practicality and compliance of financing structures and enforcement/disposition mechanisms.\n\n---\n\n### 1. Patent Assets and Financing Foundation (30 points)\n\n#### (1) Quality of Mapping Core Technology Domains and Patent Portfolio (10 points)\n- 9–10: Clearly decomposes the core technology domains of NVIDIA / Qualcomm / Palantir (e.g., GPU/CUDA ecosystem; wireless communications and SEPs; data analytics platforms and software/algorithm-related patents) and provides a verifiable portrait of the patent portfolio (counts/families, technical classifications, geographic coverage, maturity/expiration structure, citations and being-cited metrics, litigation or licensing history, etc.). Explains the mapping from “patents → products/revenue lines → competitive positioning.”\n- 6–8: Covers major technologies and a high-level patent overview, but uses fewer portfolio metrics or lacks verifiable sources; the “patent → business” mapping has gaps.\n- 0–5: Primarily company introductions or technical concepts; patent portfolio features are generic; lacks verifiable data and structured breakdown.\n\n#### (2) Argumentation on How Patent Quality, Moats, and Commercialization Potential Support Financing Feasibility (10 points)\n- 9–10: Uses interpretable indicators or frameworks to assess patent quality and moats (e.g., difficulty of technological substitution, ecosystem lock-in, standard essentiality, separability of licensing, remaining life, enforceability/litigability) and translates them into financing language (predictable cash flows, disposability/enforceability, discount and haircut logic).\n- 6–8: Discusses moats and commercialization but is largely qualitative; insufficient translation into “how it affects LTV/spread/covenants.”\n- 0–5: Vaguely states “patents are important/high barriers” without explaining concrete channels through which feasibility and terms are affected.\n\n#### (3) Rationale for Case Selection and a Cross-Company Comparability Framework (10 points)\n- 9–10: Explains structural differences across the three companies along “patent type (hardware/SEP/software) → monetization path (licensing/cross-licensing/product premium) → difficulty of disposition → valuation volatility,” and establishes a unified comparability basis (consistent patent quality metrics, a consistent valuation framework, and a consistent risk classification table).\n- 6–8: Provides comparisons, but bases/definitions are inconsistent or remains at conclusion-level only.\n- 0–5: Stacks cases without a comparability framework; comparison conclusions are not verifiable.\n\n---\n\n### 2. Patent Valuation Methods and Applicability Boundaries (25 points)\n\n#### (1) Completeness of Valuation Method Coverage and Boundary Discussion (9 points)\n- 8–9: Systematically compares the income approach, market approach, and option-based approaches in terms of applicable conditions, data requirements, and failure modes; discusses how high-tech patent features (technology uncertainty, market volatility, patent life cycle, SEP vs. non-SEP differences) affect method choice.\n- 5–7: Covers major methods, but boundary conditions are insufficient or does not reflect differences in patent types across cases.\n- 0–4: Merely lists methods; lacks actionable conclusions on “when to use / when not to use.”\n\n#### (2) Transparency of Key Assumptions, Parameters, and Calculation Process (8 points)\n- 7–8: Discloses key parameters and definitions (royalty rate/hypothetical royalty, attributable revenue base, discount rate and risk premium, remaining life, invalidation probability/litigation costs, growth rate, volatility, etc.), explains sources and reasonableness, and is reproducible (provides calculation steps or tabular definitions).\n- 4–6: Includes parameters but explanations are insufficient, or missing key definitions makes verification impossible.\n- 0–3: Does not disclose parameters; calculations are not verifiable.\n\n#### (3) Capability to Handle Uncertainty (Scenarios / Sensitivity / Practical Optionization) (8 points)\n- 7–8: Provides at least two uncertainty treatments (e.g., scenario analysis + sensitivity analysis; or option-based valuation with defined volatility/trigger conditions) and uses results to design financing terms (haircut, LTV, trigger-based top-up collateral/margin).\n- 4–6: Conducts uncertainty analysis but does not link it to financing terms, or analysis is one-dimensional.\n- 0–3: Does not handle uncertainty; directly uses point estimates for pledging, causing risk spillover.\n\n---\n\n### 3. Pledge Parameters and Risk Constraints (25 points)\n\n#### (1) LTV (Pledge Ratio), Liquidity/Monetizability, and Pricing Logic (9 points)\n- 8–9: Provides derivation logic for LTV/pledge ratio (valuation → discount → disposition costs → time discounting → recovery rate), sets ranges reflecting patent type and disposition path differences, and explains the matching of spread/fees, tenor, amortization/maturity structure, and collateral revaluation frequency.\n- 5–7: Provides ranges but lacks derivation, or is inconsistent with disposition assumptions.\n- 0–4: Only provides “rule-of-thumb LTV” without verifiable justification.\n\n#### (2) Completeness of Key Risk Identification and Impact Channels (10 points)\n- 9–10: Covers and decomposes impact channels including: technology substitution (generation shifts, architecture migration), legal validity and title/perfection (chain of title, joint inventorship/employee inventions, invalidity challenges/IPR), valuation volatility (market/volatility/growth expectations), disposition efficiency (buyer universe; antitrust/export controls/national security constraints), concentration and cross-licensing (especially Qualcomm-like SEP/FRAND risks), reputation and litigation spillovers; risks are tightly tied to case characteristics.\n- 6–8: Covers major risks, but mechanisms are generic or insufficiently linked to cases.\n- 0–5: Risk listing without key categories or without explanation of impact channels.\n\n#### (3) Risk Mitigation Tools and Covenant Design (6 points)\n- 6: Proposes implementable covenants and monitoring plans (revaluation/top-up triggers, patent maintenance fee and litigation obligations, oversight of licensing revenues and cash sweeps, negative covenants, information disclosure, key patent schedules and substitution mechanisms, insurance/litigation funding arrangements, etc.) and explains how they improve recovery and reduce volatility.\n- 3–5: Proposes some tools but lacks trigger conditions or responsible-party/execution arrangements.\n- 0–2: Nearly no mitigation design, or only slogans such as “strengthen risk control.”\n\n---\n\n### 4. Financing Structure and Mechanism Design (20 points)\n\n#### (1) Differentiated Credit Underwriting and Financing Scheme Design (10 points)\n- 9–10: Incorporates R&D spending structure, technology iteration cycles, and capital market conditions to propose at least two executable structures (e.g., revolving credit with periodic review; tiered collateral pools; dynamic limits based on patent quality scoring; cashflow-linked structures tied to licensing revenues; boundary discussion of SPV/ring-fencing arrangements, etc.) and clearly specifies fit-for-purpose targets (differentiated fit for NVIDIA/Qualcomm/Palantir).\n- 6–8: Has structural ideas but insufficient parameterization (limit adjustment rules, triggers, tenor matching not specified).\n- 0–5: Generic方案; cannot be implemented into credit terms or differentiated fit.\n\n#### (2) Closed-Loop Design for Registration, Transfer, and Disposition/Enforcement Mechanisms (6 points)\n- 6: Clearly describes how to create and perfect security interests in the U.S. context (e.g., UCC perfection; relationship to USPTO recordation and practical differences), patent schedule management, third-party valuation and escrow/custody, and the process and timing assumptions for disposition via IP marketplaces/technology transfer intermediaries, considering disposition costs and discounts.\n- 3–5: Mentions platforms/intermediaries, but the process and key steps (title verification, perfection, disposition) are incomplete or not executable.\n- 0–2: Lacks closed-loop mechanisms; remains at principle-level statements.\n\n#### (3) Compliance, Regulatory, and Disclosure Awareness (4 points)\n- 4: Identifies and discusses the impact of at least two constraints on the structure (e.g., export controls/national security review limiting buyer universe; antitrust and SEP/FRAND constraints; rights boundaries for data/software-related IP; uncertainty of claims and priority in bankruptcy scenarios; impacts of disclosure and accounting treatment on financing availability).\n- 2–3: Mentions compliance but does not explain how it affects structure and disposition.\n- 0–1: Ignores key compliance/regulatory constraints with obvious real-world bias.\n"
  },
  {
    "id": "20",
    "question": "In the U.S. existing home mortgage market, competition between large national banks and regional banks continues to evolve. As a major market participant, JPMorgan Chase exhibits strong industry representativeness in terms of business footprint, product capabilities, and risk management practices.\n\nConduct a systematic assessment of JPMorgan Chase’s competitive position in the U.S. existing home mortgage market. Based on a review of peer banks’ mortgage lending activities—including both large national banks and regional institutions—evaluate JPMorgan Chase’s relative strengths and weaknesses across product design, pricing and interest rate strategies, approval and funding efficiency, channel deployment, and risk management and compliance capabilities.\n\nTaking into account shifts in the U.S. housing market cycle and the evolution of mortgage-related regulatory frameworks, propose actionable directions for business optimization and competitive strategy.",
    "classification": "Banking-Assets",
    "classification_code": "BNK-AST",
    "report_type": "Competitive Landscape and Operating Strategy Study of Residential Mortgage Finance",
    "report_type_zh": "住房按揭金融业务竞争格局与经营策略研究报告",
    "language": "en",
    "expert_evaluation_criteria": "## Evaluation Criteria for Research Reports on JPMC's Competitive Position in the U.S. Existing-Home Mortgage Market\n\nThis rubric provides a systematic, actionable, and quantitative framework to evaluate research reports focused on JPMC's competitive position in the U.S. existing-home mortgage lending market. Scoring emphasizes: rigor of peer benchmarking, depth of professional decomposition of the mortgage value chain, explicit integration of cycle and regulatory factors, and the executability plus risk/compliance completeness of strategic recommendations. The framework includes 4 top-level dimensions and 14 second-level dimensions, totaling 100 points, and is designed to support consistent use in both human review and LLM-based automated scoring.\n\n---\n\n### 1. Market Definition and Peer Benchmarking Framework (25 points)\n\n#### (1) Clarity of problem boundaries and business/value-chain definition (8 points)\n- 7-8: Clearly defines the 'Existing Home' context (primarily purchase mortgages, distinguished from new-home/construction lending), and breaks down the mortgage value chain (origination/underwriting/closing/sell or hold-on-balance-sheet/servicing/MSR/risk hedging). Provides explicit definitions for key segmentations such as purchase vs. refi, agency vs. non-agency, and retail vs. correspondent/broker.\n- 4-6: Explains the basic scope of existing-home mortgages, but key segmentation definitions are incomplete, or the distinctions between origination and servicing and/or hold vs. sell are insufficiently clarified.\n- 0-3: Scope is vague; mixes existing-home with the overall mortgage market, or lacks a structured value-chain decomposition.\n\n#### (2) Reasonableness of peer set selection and control of comparability (8 points)\n- 7-8: Covers both national banks and regional banks, and explains selection rationale (asset size, mortgage activity level, channel model, geographic footprint, agency mix, etc.). Clearly states what is comparable and how non-comparable items are handled (e.g., accounting treatments, differences in sell vs. hold strategies).\n- 4-6: Peer coverage is broadly adequate, but lacks comparability explanation, or the regional-bank sample is not sufficiently representative.\n- 0-3: Peer selection is arbitrary and poorly justified, weakening the credibility of benchmarking conclusions.\n\n#### (3) Depiction of market structure and competitive landscape (9 points)\n- 8-9: Uses verifiable metrics to describe the landscape (e.g., HMDA origination share, purchase mix, average rate/points, closing cycle time, denial-rate decomposition, channel mix, servicing scale/MSR, etc.), and distinguishes competitive factors across different cycle phases (high-rate vs. low-rate, tight vs. ample inventory).\n- 5-7: Provides some data/fact support, but the metric set is incomplete, or lacks segmentation (channel/region/customer segment), resulting in an overly coarse competitive picture.\n- 0-4: Mainly qualitative description; lacks structured metrics and verifiable support.\n\n---\n\n### 2. Comparative Assessment of Competitive Drivers (Product, Pricing, Efficiency, Channels, Experience) (35 points)\n\n#### (1) Product design and fit to target segments/scenarios (7 points)\n- 6-7: Compares differences between JPMC and peers in product suite and policy (e.g., conforming/jumbo, ARM/FRM, down-payment structures, relationship pricing, first-time buyer programs, low-down-payment/affordability solutions, etc.), and links them to key pain points in the existing-home transaction chain (rate lock, closing window, appraisal/underwriting).\n- 3-5: Covers major product types but is mostly a list; lacks fit-to-scenario analysis for the existing-home purchase context.\n- 0-2: Product analysis is generic and does not reflect differentiation or business implications.\n\n#### (2) Pricing and rate strategy comparison (including points/fees/lock/hedging) (8 points)\n- 7-8: Compares rate pricing mechanisms and strategies (rate/points/fees, impacts of LLPA/credit score/LTV, relationship discounts, rate-lock strategy) and explains underlying constraints (funding costs, capital usage, GSE/secondary-market execution, pipeline hedging). Connects trade-offs across cycles (volume capture vs. margin protection).\n- 4-6: Compares headline rates or fees but lacks explanation of funding costs/hedging/secondary-market mechanics.\n- 0-3: Only concludes 'higher/lower rates' without mechanisms or evidence.\n\n#### (3) Approval-to-funding efficiency (underwriting, closing, operating capability) (7 points)\n- 6-7: Uses comparable metrics/process mapping to characterize efficiency (application-to-clear-to-close, closing timeline, automated underwriting share, number of document re-requests, denial-reason structure, etc.) and explains drivers of differences (systems, risk policy, staffing, coordination with third-party appraisal/title/insurance).\n- 3-5: Discusses 'fast/slow' but lacks metrics or fails to pinpoint process bottlenecks.\n- 0-2: Efficiency assessment is subjective and lacks verifiable basis.\n\n#### (4) Channel footprint and customer acquisition cost structure (7 points)\n- 6-7: Compares JPMC and peers across branch retail, online direct, broker/third-party, correspondent, and partner ecosystems (realtors/builders/platforms), and discusses implications for existing-home conversion (lead quality, conversion rate, per-loan acquisition cost, regional penetration).\n- 3-5: Mentions channel differences but lacks cost/conversion/KPI perspective or linkage to the existing-home context.\n- 0-2: Channel analysis stops at 'has branches/has online' without business meaning.\n\n#### (5) Customer experience and cross-sell/relationship synergy (6 points)\n- 5-6: Evaluates key customer-journey touchpoints (pre-approval, rate lock, closing coordination, post-close servicing) and, leveraging large-bank advantages, discusses how to coordinate with deposits, wealth management, credit cards, etc., including boundaries and compliance constraints.\n- 3-4: Mentions experience or synergy but lacks actionable decomposition or remains conceptual.\n- 0-2: Does not reflect linkage between mortgage business and relationship management, or conclusions are empty.\n\n---\n\n### 3. Cycle and Regulatory Integration; Risk and Compliance Capability (20 points)\n\n#### (1) Housing/interest-rate cycle framework and scenario analysis (7 points)\n- 6-7: Builds a transmission chain from 'rates -> affordability -> transaction volume -> home prices/inventory -> defaults/loss severity -> mortgage profitability'. Provides at least 2 scenarios (e.g., rates decline and refinancing window reopens; rates stay high and purchase volumes stagnate) and derives implications for JPMC vs. peers.\n- 3-5: Describes cycle impacts but the transmission chain is incomplete, or lacks scenario quantification/trigger conditions.\n- 0-2: Barely integrates cycle variables, or reduces cycle impact to slogans.\n\n#### (2) Comparative risk management capability (credit/market/operational/model/hedging) (8 points)\n- 7-8: Compares JPMC and peers on credit risk (FICO, DTI, LTV policies and denial structure), interest-rate and pipeline risk (rate lock/hedging), MSR valuation and hedging, and operational risk (fraud, documentation, third parties). Identifies trade-offs among 'risk appetite -> scale growth -> capital usage'.\n- 4-6: Covers some risk points but misses key links such as hedging/MSR/operational risk, or the comparison is not systematic.\n- 0-3: Risk discussion is generic (only 'strong/weak risk control') and misses mortgage-specific risk drivers.\n\n#### (3) Regulatory and compliance capability (QM/ATR, HMDA, Fair Lending, UDAAP, etc.) (5 points)\n- 4-5: Clearly identifies core mortgage regulatory constraints and how they affect product, pricing, underwriting, and channels (e.g., fair lending, explainability of pricing differences, denial reason disclosures, consumer protection, data reporting and audit readiness). Proposes process and monitoring controls that embed compliance into operations.\n- 2-3: Mentions regulatory terms but does not explain impacts on operating decisions, or lacks specific controls.\n- 0-1: Ignores regulation/compliance, or proposes actions with obvious compliance/reputational risk.\n\n---\n\n### 4. Executability of Strategic Recommendations and Value Realization (20 points)\n\n#### (1) Targeted optimization directions and competitive strategy (8 points)\n- 7-8: Recommendations directly address weaknesses/opportunities identified in the benchmarking (e.g., pricing strategy, products for specific segments, channel redesign, closing-efficiency uplift, MSR strategy, etc.) and clearly specify differentiated approaches against national banks vs. regional banks.\n- 4-6: Directions are reasonable but weakly connected to prior comparative findings, or lacks differentiated playbooks by competitor type.\n- 0-3: Recommendations are generic (e.g., 'strengthen risk control/improve service') and not targeted.\n\n#### (2) Implementation path (org, systems, processes) and KPI design (7 points)\n- 6-7: Provides an executable roadmap (phases, accountable owners, system changes/data needs, partner management) and defines measurable KPIs (e.g., pull-through rate, rate-lock conversion, closing time, acquisition cost per loan, denial-rate structure, delinquency/loss rates, customer satisfaction).\n- 3-5: Includes partial implementation detail or KPIs, but not systematic; lacks resources/organization and timeline.\n- 0-2: Lacks an implementation plan and quantitative metrics.\n\n#### (3) Risk-return trade-offs and constraints (capital, liquidity, compliance, reputation) (5 points)\n- 4-5: Compares sources of upside and costs/risks (spread, fees, share gains, cross-sell vs. capital usage, hedging cost, compliance cost, reputational risk) and states boundaries/trigger thresholds for 'do vs. not do'.\n- 2-3: Mentions risk-return but lacks a quantified framework or clear boundary conditions.\n- 0-1: Talks only about growth without constraints, or ignores hard constraints such as capital and compliance.\n"
  },
  {
    "id": "21",
    "question": "在金融强监管背景下，保险资产负债管理已从以偿付能力为核心的静态约束，转向覆盖久期匹配、成本收益匹配与现金流匹配等多维度的系统性监管框架。监管部门全面实施《保险资产负债管理监管规则（1–5号）》，对保险公司资产配置与投资行为提出了更高的合规与稳健性要求。\n\n请围绕“强监管框架下保险公司的最优资产配置决策”这一主题，撰写一份研究型分析报告，系统分析以下问题：\n\n1. 概述《保险资产负债管理监管规则（1–5号）》的核心监管目标与主要约束机制，并说明其相较于以偿付能力为核心的传统监管体系，在资产配置层面的关键差异。\n2. 从资产负债管理视角，分析期限结构匹配、成本收益匹配与现金流匹配等监管要求，如何通过约束资产久期、收益特征与流动性结构，影响保险公司的资产配置决策。\n3. 结合人身险（如分红险）负债的长期性与刚性特征，分析在强监管约束下保险公司资产配置结构的可能变化方向，并重点讨论固定收益类资产配置比例上升的内在逻辑。\n4. 综合评估资产负债管理监管规则对保险公司投资收益、风险暴露、久期缺口及长期经营稳健性的影响，并在合规前提下提出优化资产配置与风险管理的政策建议。",
    "classification": "Insurance-Actuarial & Reserving",
    "classification_code": "INS-RSV",
    "report_type": "Asset-Liability Management and Asset Allocation in Insurance",
    "report_type_zh": "保险资产负债管理与资产配置研究",
    "language": "zh",
    "expert_evaluation_criteria": "## 保险公司“强监管框架下最优资产配置决策”研究报告评分标准\n\n本评分体系用于对围绕《保险资产负债管理监管规则（1–5号）》的研究型分析报告进行可执行量化评估。总分100分，设4个一级维度、12个二级维度。评分重点关注：监管理解的准确性、ALM传导机制的专业深度、对长期负债/分红险特征的适配性、以及合规前提下的配置优化建议的可落地性。支持人工评审与大模型（LLM）一致性评分。\n\n---\n\n### 1. 监管框架理解与差异化解读（共 25 分）\n\n#### （1）监管目标与规则要点覆盖完整性（10分）\n- 9–10分：准确概括规则（1–5号）核心目标（如强化资产负债匹配、抑制期限错配与流动性风险、约束激进投资行为、强调治理与过程管理等）与主要约束机制（匹配指标/量化评估、压力测试与情景分析要求、信息披露或报送与穿透管理、内部治理与责任机制等），覆盖全面且表述无明显偏差。\n- 6–8分：总体方向正确，能覆盖主要目标与部分机制，但对关键机制的表述偏概念化或遗漏一到两个关键约束点。\n- 0–5分：对规则理解明显不准确、要点缺失较多，或以行业传闻代替规则逻辑。\n\n#### （2）与传统“以偿付能力为核心”监管的关键差异（8分）\n- 7–8分：清晰指出差异：从资本充足率等“结果型/静态”约束，转向“过程型/动态”ALM约束；从单一风险资本视角，扩展到久期、现金流、成本收益匹配与流动性结构等多目标约束；并能说明这些差异如何改变资产配置目标函数与约束集合。\n- 4–6分：能指出部分差异，但缺少对“为什么会改变资产配置决策”的机制说明。\n- 0–3分：差异描述笼统（如仅写“更严格了”），或把ALM规则等同于偿付能力规则的简单加严。\n\n#### （3）“监管条款 → 配置约束条件”的转译能力（7分）\n- 6–7分：能将监管要求转译为可执行的配置约束（如目标久期区间、久期缺口限额、现金流缺口/期限桶覆盖要求、收益形态与负债成本匹配、流动性资产比例与变现折扣考虑等），并能落到资产类别与期限结构层面。\n- 3–5分：有转译尝试，但缺少关键指标口径、边界条件或可操作表达。\n- 0–2分：停留在条文复述或口号式表达，无法指导配置决策。\n\n---\n\n### 2. ALM约束传导机制与分析深度（共 30 分）\n\n#### （1）期限结构/久期匹配的机理与量化表达（10分）\n- 9–10分：说明监管如何约束资产久期、利率敏感性与再投资风险；能使用专业指标/框架（如久期缺口、关键期限久期KRD、凸性、利率情景下资产负债PV敏感性、免疫/LDI思路等）推导配置倾向，并给出清晰的逻辑链（负债久期与现金流 → 资产久期配置 → 缺口管理与再平衡）。\n- 6–8分：能讲清基本传导逻辑，但量化指标使用不充分或仅定性描述。\n- 0–5分：对久期匹配理解偏差（如把“久期越长越好”绝对化），或缺少可验证的分析链条。\n\n#### （2）成本收益匹配：负债成本、保证与分红机制的约束（10分）\n- 9–10分：明确负债成本的构成与约束含义（保证利率/最低利益、分红或结算机制、费用与退保等对实际负债成本的影响），并分析资产端收益形态（票息、久期溢价、信用利差、权益波动收益）如何在监管下与负债成本匹配；能讨论利差损风险、收益平滑与会计/估值口径对行为的影响（不要求拘泥具体会计准则条文，但逻辑需自洽）。\n- 6–8分：能覆盖“负债成本约束资产收益”的主线，但对分红险等产品机制拆解不够，或缺少对收益来源与波动的结构化分析。\n- 0–5分：将“高收益资产=更优”简单化，忽视负债成本刚性与收益波动/兑现的不匹配。\n\n#### （3）现金流匹配与流动性约束：期限桶、退保与压力情景（10分）\n- 9–10分：能以现金流/流动性视角解释监管约束：期限桶现金流缺口、流动性资产配置、变现能力与折价、集中度与非标/长久期资产的流动性代价；纳入退保、保费增长放缓、信用事件等压力情景，说明对配置与备付的影响路径。\n- 6–8分：能说明流动性约束的重要性与基本影响，但缺少情景化分析或对“可变现性/折价”的讨论不足。\n- 0–5分：把现金流匹配等同于“多配货币基金”等简单结论，或忽略退保与压力情景。\n\n---\n\n### 3. 人身险（分红险等）负债特征下的配置结构推演（共 25 分）\n\n#### （1）长期性与刚性特征刻画准确度（8分）\n- 7–8分：准确描述长期负债的期限结构、利益刚性来源（保证/最低利益、刚性兑付预期）、行为假设敏感项（退保、保费续期、分红期望）对负债现金流与久期的影响，并能指出哪些是“可调”、哪些是“不可调”的负债端变量。\n- 4–6分：描述基本正确，但对分红险等机制刻画较粗，或忽视关键行为假设。\n- 0–3分：负债特征描述与实际产品机理明显不符或过度泛化。\n\n#### （2）配置结构变化方向与“固收占比上升”的内在逻辑（10分）\n- 9–10分：在强监管约束下，推导资产结构变化（期限结构拉长/阶梯化、固收核心仓位提升、信用与利率风险取舍、权益与另类的边界条件与使用方式）；对“固收占比上升”给出多维原因链条（久期匹配、现金流可预测性、收益兑现确定性、监管量化评估与压力测试下的资本与波动约束、流动性与声誉风险等），同时指出代价与反例情景（如利率下行后的再投资风险、信用下沉约束）。\n- 6–8分：方向判断合理，但论证链条较短，缺少“为什么在规则下更优”的约束驱动解释。\n- 0–5分：仅凭经验下结论（如“监管严了所以只能买债”），缺乏机制推导。\n\n#### （3）“最优”定义与配置方法适配性（7分）\n- 6–7分：对“最优资产配置”给出可解释的目标函数与约束（如在满足久期/现金流/成本收益匹配与合规限额下，最大化风险调整后收益或最小化缺口波动；可讨论SAA/TAA、LDI、再平衡规则等），方法与保险负债特征匹配。\n- 3–5分：提出优化框架但较抽象，缺少约束集合或目标指标的明确表达。\n- 0–2分：没有“最优”定义，或完全照搬公募/券商配置框架而不适配保险负债与监管约束。\n\n---\n\n### 4. 综合影响评估与合规可落地建议（共 20 分）\n\n#### （1）对收益、风险暴露、久期缺口与稳健性的综合评估（8分）\n- 7–8分：对规则实施后可能带来的变化进行平衡评估：投资收益（票息/利差/久期溢价）、风险暴露（利率、信用、流动性、权益波动、集中度等）、久期缺口变化与经营稳健性；能给出情景对比或至少结构化的权衡结论，而非单向度判断。\n- 4–6分：覆盖主要影响项，但缺少权衡分析或影响路径较粗。\n- 0–3分：只谈收益或只谈风险，缺乏综合评估。\n\n#### （2）优化资产配置与风险管理建议的可操作性（8分）\n- 7–8分：建议具体、可执行且合规，例如：久期与现金流分桶管理、缺口限额与预警阈值、压力测试与敏感性分析框架、负债端假设治理、资产端信用分层与集中度管理、流动性备付与应急融资预案、再平衡与止损机制、投研与交易流程改造等，并能说明落地条件与组织协同。\n- 4–6分：建议方向正确但偏原则化，缺少工具、阈值、流程或责任机制。\n- 0–3分：建议泛泛而谈，或提出明显不合规/不可执行的做法。\n\n#### （3）合规与风险提示（含声誉风险）完整性（4分）\n- 4分：明确识别合规边界与潜在监管关注点（穿透、非标与关联交易、期限错配、信息披露/报送、估值与流动性折价等），并在建议中体现“合规优先”的约束；风险提示与正文观点形成制衡。\n- 2–3分：提及合规与风险，但较形式化。\n- 0–1分：缺少合规意识，或建议触碰明显监管红线。"
  },
  {
    "id": "22",
    "question": "在低利率长期化与监管趋严的背景下，寿险行业正围绕负债属性与资产负债管理展开深度调整。分红型寿险因兼具保障属性与收益弹性，被视为当前环境下重新具备战略价值的重要产品形态，并呈现出明显的公司间分化。\n\n请围绕“低利率环境下分红险的复兴与分化”这一主线，从战略、精算与资产负债管理的综合视角，撰写一份寿险行业深度研究报告，重点分析以下问题：\n\n1. 结合利率周期、监管约束与客户风险偏好变化，分析分红型寿险相较传统型寿险重回主流的核心原因，并说明其“固收+”属性的环境适配性。\n2. 比较现金分红与保额分红、高保证与低保证分红险在收益结构、负债稳定性及客户适配性方面的差异，并评估其对寿险公司长期经营的影响。\n3. 结合 IFRS 17 下浮动收费法（VFA），分析分红险在资产配置灵活性、风险分担结构及资产负债匹配方面相较传统险的关键变化。\n4. 以中国平安、中国太保或友邦保险等代表性寿险公司为例，分析其分红险战略选择（稳健型 vs 进取型）的内在逻辑及其在投资收益、负债成本与经营风险之间的权衡。\n5. 在监管趋严与投资回报不确定性上升的背景下，总结分红险发展中需要重点防范的主要风险，并提出对寿险公司产品结构与资产配置策略的启示。",
    "classification": "Insurance-Life Insurance",
    "classification_code": "INS-LIF",
    "report_type": "Research Report on Participating Life Insurance Products and Asset-Liability Management",
    "report_type_zh": "分红型寿险产品与资产负债管理研究报告",
    "language": "zh",
    "expert_evaluation_criteria": "## 寿险行业深度研究报告评分标准\n\n本评价体系适用于评估以 低利率环境下分红险的复兴与分化 为主题的研究报告。评分强调：战略洞察的增量性、精算与产品机制的准确性、IFRS 17/VFA与ALM分析的专业度、案例与数据的可验证性，以及风险与建议的可操作性。体系共设 **4 个一级维度、12 个二级维度**，总分 **100 分**，兼容人工评审与LLM自动评分的一致性应用。\n\n### 1. 核心观点与行业环境适配（共 25 分）\n\n#### （1）利率周期与监管约束刻画准确性（10 分）\n- **9–10 分**：清晰刻画低利率长期化对寿险“利差/久期/再投资风险”的影响机制；准确覆盖关键监管约束（如定价利率/演示利率约束、报行合一/费用管控、投资端穿透与资管新规相关约束、偿付能力/资本约束等）及其对产品形态与经营策略的传导路径；能明确“约束—行为—结果”的链条。\n- **6–8 分**：能覆盖主要利率与监管因素，但传导逻辑略粗；对监管条款的适用边界、时间点或影响方向存在遗漏。\n- **0–5 分**：停留在口号式描述（“低利率、强监管”），缺少机制解释；或出现明显方向性错误（例如将监管影响与产品收益结构混淆）。\n\n#### （2）分红险“重回主流”的核心原因与“固收+”适配性论证（8 分）\n- **7–8 分**：能从“客户风险偏好（保守化/追求稳健确定性）—公司负债成本管理—投资收益不确定性”三端同时论证分红险复兴；明确解释其“保证 + 参与分享”的**固收+**属性如何在低利率下改善传统险的刚性负债压力（如通过降低保证、引入风险共担/收益弹性来缓释利差风险）。\n- **4–6 分**：能说明分红险更适配低利率，但论证主要停留在结论层或仅从单一主体（只讲客户或只讲公司）展开。\n- **0–3 分**：将分红险简单等同高收益替代品，或忽略保证/分红非保证的关键前提，导致论证失真。\n\n#### （3）“复兴与分化”的竞争格局判断与前瞻性（7 分）\n- **6–7 分**：明确提出公司间分化的核心驱动（资产能力、渠道结构、风险偏好、资本充足度、红利政策与平滑能力、会计与估值管理等），并给出可验证的前瞻判断或情景（如未来1–2年产品供给/红利实现率分化的条件）。\n- **3–5 分**：指出分化现象与部分原因，但缺少可检验的判断框架或时间条件。\n- **0–2 分**：只描述“公司策略不同”，缺乏结构化解释与可验证预测。\n\n---\n\n### 2. 产品结构与精算机制分析（共 25 分）\n\n#### （1）现金分红 vs 保额分红：收益结构与客户体验机制（12 分）\n- **11–12 分**：准确拆解两类分红方式在**红利形态（现金领取/累积生息 vs 增加保额/复归红利）**、收益呈现路径、长期复利效应、领取灵活性、退保价值影响、红利平滑与分配政策敏感性等方面的差异；能结合典型条款要点（如红利来源、分配规则、非保证声明、特别储备/平滑机制概念）解释客户侧“确定性/波动性/可理解性”差异。\n- **7–10 分**：能区分两者主要差异，但对关键机制（如复归红利对长期保障与现价的影响、现金红利的再投资假设）解释不完整。\n- **0–6 分**：将两者差异仅归结为“给现金/不给现金”，缺乏机制；或出现概念性错误（把非保证红利当作保证收益）。\n\n#### （2）高保证 vs 低保证分红险：负债稳定性与经营影响评估（8 分）\n- **7–8 分**：从精算视角说明保证水平变化对**负债久期与刚性、利差风险、退保行为与资产端匹配难度、费用摊销与新单价值、资本占用/偿付能力压力**的影响；能讨论不同保证结构在不同利率情景下的“尾部风险”与利润波动。\n- **4–6 分**：能说明高保证更“刚性”、低保证更“弹性”，但缺少对负债稳定性与经营指标的量化或机制化解释。\n- **0–3 分**：仅做价值判断（“高保证更好/更差”），缺少精算与经营传导。\n\n#### （3）关键假设、利润来源与敏感性分析（5 分）\n- **5 分**：明确列示并解释关键假设（投资收益率路径、费用率、退保率/持续率、死亡/发生率、分红政策等）与利润来源拆分（利差/死差/费差/退保差等框架之一即可）；至少提供1个以上敏感性或情景对比（如利率下行100bp对红利能力/CSM/利润波动的影响方向）。\n- **3–4 分**：提到假设或利润来源，但缺少结构化列示或缺少敏感性对比。\n- **0–2 分**：不披露关键假设，结论不可检验。\n\n---\n\n### 3. 资产负债管理与 IFRS 17/VFA 专业度（共 30 分）\n\n#### （1）IFRS 17 下 VFA 的适用条件与会计机制阐释（10 分）\n- **9–10 分**：准确说明VFA适用逻辑（参与型合同、与基础项目回报高度相关、实体收取“可变收费”等核心条件的表述到位即可），并讲清对关键会计要素的影响（如CSM形成与释放、金融假设变动的处理思路、风险分担如何影响利润波动呈现）；避免将VFA与GMM/BBA或PAA混淆。\n- **6–8 分**：能描述VFA大意与方向性影响，但对适用边界或关键机制（例如“基础项目/Underlying Items”“可变收费”的经济含义）解释不够清晰。\n- **0–5 分**：概念混乱或出现实质性错误（例如把VFA当作“直接按市值计量利润”之类的误解）。\n\n#### （2）分红险的资产配置灵活性、风险分担与ALM匹配分析（12 分）\n- **11–12 分**：能从“风险共担结构”出发，解释分红险相对传统险在资产端的可承受波动、配置边界与策略空间（如久期匹配、信用风险暴露、权益/另类配置的审慎边界、再投资风险管理工具等）；清晰呈现“资产收益—红利/保证—公司利润/资本”的传导，并指出ALM关键指标（久期缺口、现金流匹配、凸性/期权性、流动性压力来源等）。\n- **7–10 分**：讨论了资产配置与匹配，但缺少指标化表达或缺少对风险分担结构的穿透解释。\n- **0–6 分**：仅泛谈“提高权益配置/多元投资”，缺少ALM约束与负债特性连接。\n\n#### （3）监管/资本约束下的稳健性检验（8 分）\n- **7–8 分**：能将产品策略与资本/偿付能力约束联动分析（可用“资本占用、利率/信用/权益风险资本、压力测试、流动性与退保冲击”任意一套一致框架），并给出至少一个压力或情景下的风险点与应对。\n- **4–6 分**：提及偿付能力或资本约束，但缺少情景化检验或缺少与产品结构的对应关系。\n- **0–3 分**：忽略资本与监管约束，或将其与会计口径混为一谈。\n\n---\n\n### 4. 案例研究、证据质量与可操作建议（共 20 分）\n\n#### （1）代表性公司案例：战略选择逻辑与权衡（10 分）\n- **9–10 分**：以平安/太保/友邦等为例，清晰界定其分红险策略属于“稳健型 vs 进取型”（或自定义但需一致），并用可核验证据支撑（如产品保证水平、分红实现与平滑方式、投资组合特征、渠道结构、负债成本管理取向、风险偏好与资本充足度等）；明确阐释其在**投资收益—负债成本—经营风险**之间的取舍逻辑，避免只做定性贴标签。\n- **6–8 分**：有公司对比与方向判断，但证据不足或缺少“权衡”的结构化拆解。\n- **0–5 分**：案例停留在新闻复述/口碑评价，缺少可验证事实与机制解释。\n\n#### （2）主要风险识别与防范建议（6 分）\n- **6 分**：覆盖并区分至少三类关键风险，且说明影响路径与触发条件（示例：利率下行与再投资风险、信用利差与资产减值、权益波动与红利能力/客户预期管理、退保与流动性风险、模型与假设风险、合规与销售误导/声誉风险、会计利润波动与管理层行为扭曲等）；给出对应防范抓手（条款设计、红利政策、资产久期与质量、流动性管理、信息披露与销售管理等）。\n- **3–5 分**：提及若干风险，但缺少触发条件/影响链条或防范措施较空泛。\n- **0–2 分**：风险提示模板化罗列，无法指导实际管理。\n\n#### （3）对产品结构与资产配置策略的启示：可执行性（4 分）\n- **4 分**：给出可落地建议，并能与前文分析闭环对应（例如：保证水平梯度设计、红利机制与客户沟通策略、目标久期与资产篮子、分账户/分策略管理、情景阈值触发的再平衡规则等）；建议具备优先级或适用条件。\n- **2–3 分**：建议方向正确但偏宏观，缺少实施路径或边界条件。\n- **0–1 分**：仅给原则性口号，无法执行。"
  },
  {
    "id": "23",
    "question": "多因子选股模型中，基本面因子往往面临价格信息回流、财务披露滞后等问题，导致因子表观有效性与真实可交易性出现偏差。请以“基本面因子的重构与提纯”为主题，撰写一份面向 A 股市场的深度量化研究报告，至少包含以下内容：\n\n1、基本面因子实盘表现分析：请结合现有文献及研报，分析基本面因子在股票市场中的时序表现，分析部分时间段基本面因子失效原因；\n2. 传统因子痛点分析：分析常用基本面因子（如 BP、ROE、净利润增长率等）在直接使用时的主要问题（例如财务数据发布滞后带来的信息失效或前视风险等）；\n3. 因子重构方法论：提出可落地的因子重构逻辑，详细说明如何通过“财务数据对齐（as-of/滞后处理）”、“结合分析师基本面预测”等方式提升因子有效性与稳健性。",
    "classification": "Capital Markets-Financial Engineering & Quantitative Research",
    "classification_code": "CAP-QUA",
    "report_type": "Fundamental Factor Performance Analysis and Reconstruction Research Report",
    "report_type_zh": "基本面因子表现分析与重构研究报告",
    "language": "zh",
    "expert_evaluation_criteria": "## A 股“基本面因子的重构与提纯”量化研究报告评价标准\n\n本评价体系用于对多因子选股场景下“基本面因子重构与提纯”研究报告进行结构化量化评分，强调**机制解释—方法可落地—数据无前视—实证可复验—结果可交易**的一致性。体系共 4 个一级维度、14 个二级维度，总分 100 分，兼容人工评审与 LLM 自动评分的一致性应用。\n\n---\n\n## 1. 传统因子痛点分析质量（共 25 分）\n\n### （1）痛点覆盖完整性与 A 股适配性（10 分）\n- 9–10 分：系统覆盖并结合 A 股特性解释至少 3 类核心污染源：  \n  `行业暴露/结构性行业 beta`、`市值/估值风格暴露`、`价格信息回流（基本面指标与价格/动量/反转的相关）`、`财务披露滞后与信息可得性（公告日/报告期）`、以及常见工程陷阱（停牌、ST、涨跌停、退市与幸存者偏差等）。并能指出每类问题对 IC/分组收益/可交易性的具体影响路径。\n- 6–8 分：覆盖主要问题点，但对 A 股细节约束（披露节奏、交易制度、样本过滤）讨论偏泛或停留在现象描述。\n- 0–5 分：仅罗列概念或“因子失效”结论，缺少机制与 A 股落地背景。\n\n### （2）“表观有效性 vs 真实可交易性”区分与证据（8 分）\n- 7–8 分：明确区分研究中常见的“假有效”：如行业集中导致的伪单调、用报告期数据/公告日前视导致的虚高 IC、忽略涨跌停/停牌导致的不可成交收益；能用示例、分解或诊断图表（如暴露分解、公告日前后 IC 断点）支撑。\n- 4–6 分：能提出区分，但证据链不足或未能落到可量化诊断。\n- 0–3 分：混淆概念，把回测收益等同可交易收益，未识别前视/不可成交来源。\n\n### （3）问题量化诊断与归因方法（7 分）\n- 6–7 分：给出清晰可复现的诊断方法：如对因子做行业/市值暴露回归、对因子与收益/动量相关做偏相关或回归、对公告对齐前后 IC 对比、分行业分市值分组的稳定性检验等，并解释判据。\n- 3–5 分：有诊断动作但方法/判据含糊（例如只给相关系数不解释控制变量与显著性）。\n- 0–2 分：缺少诊断框架，仅凭经验判断归因。\n\n---\n\n## 2. 因子重构与提纯方法论（共 30 分）\n\n### （1）去行业/市值中性化方案的严谨性与可执行性（10 分）\n- 9–10 分：明确给出中性化定义与实现细节（如截面回归残差法/分组去均值法），包含：行业分类口径（申万/中信及版本）、市值项（`log(mcap)` 等）、权重选择、极值处理与缺失值处理；能讨论“是否需要同时做行业与市值”“顺序/一次回归”的差异与影响。\n- 6–8 分：方法方向正确，但关键实现细节缺失（行业口径、回归设定、样本过滤、稳健回归/加权等）。\n- 0–5 分：仅口头描述“做中性化”，无明确可复现操作，或中性化设置明显不合理。\n\n### （2）剥离价格影响/回流的定义与处理（8 分）\n- 7–8 分：清晰定义“价格影响”来源（估值分母含价格、基本面与动量/反转共振等），并给出可落地剥离方式（如对因子对近期收益/动量/波动/换手等做控制回归取残差；或将估值类因子拆分为基本面与价格成分并分别处理），说明为何能减少回流而不误删有效信号。\n- 4–6 分：提出剥离想法但实现路径不够清楚，或仅做简单相关性过滤缺乏控制变量框架。\n- 0–3 分：未处理价格回流，或用不恰当方式导致信号本质被破坏且无论证。\n\n### （3）财务数据对齐（as-of/滞后处理）与防前视设计（8 分）\n- 7–8 分：明确采用 point-in-time / as-of 数据理念，至少说明：  \n  `可得性时间戳（公告日/披露日）`、`生效规则（T+1/若干交易日滞后）`、`报告期 vs 公告日的区别`、`修正/追溯披露处理`；并能解释为何该规则在 A 股环境下合理，且不会引入隐性前视。\n- 4–6 分：认识到披露滞后，但仅做粗糙固定滞后（如统一滞后 N 天）且未说明公告口径与异常情况处理。\n- 0–3 分：仍使用报告期日期直接入库、或未说明对齐方式，存在明显前视风险。\n\n### （4）重构后因子定义清晰、可复现、可扩展（4 分）\n- 4 分：给出明确数学表达/伪代码/流程图，说明标准化（rank/z-score）、截面频率（月/周）、样本空间、缺失处理；并能说明如何扩展到 ROE、利润增速等多类基本面因子。\n- 2–3 分：定义基本清楚但缺少关键工程细节（更新频率、缺失/极值处理）。\n- 0–1 分：重构因子定义模糊，难以复现。\n\n---\n\n## 3. 绩效实证与稳健性验证（共 30 分）\n\n### （1）回测框架可信度与 A 股交易约束处理（10 分）\n- 9–10 分：回测设计完整且可信：清晰给出股票池口径（全 A/剔除 ST/上市时长）、调仓频率、信号生成与交易执行时序、T+1、涨跌停/停牌不可成交处理、复权口径、基准与对冲方式（如市场中性/指数对冲）等；能明确避免数据泄露。\n- 6–8 分：框架基本完整，但对关键可交易约束处理不充分（如涨跌停/停牌、T+1、样本过滤）。\n- 0–5 分：回测设定缺失或存在显著不可信假设（如用未来可得数据、忽略不可成交）。\n\n### （2）指标体系完整性：IC/ICIR/单调性/多空收益（8 分）\n- 7–8 分：按题目要求完整报告并解释：`IC 均值`、`ICIR`、`分组回测单调性（含统计显著性/置信区间更佳）`、`多空组合年化收益`；并对“重构前 vs 重构后”做同口径对比（同股票池、同调仓、同成本）。\n- 4–6 分：指标基本齐全但解释不足，或对比口径不一致影响结论可信度。\n- 0–3 分：关键指标缺失或只展示对自己有利的单一指标。\n\n### （3）稳健性检验与异象排查（8 分）\n- 7–8 分：提供多维稳健性：不同时期/牛熊、不同市值段、分行业、不同调仓频率、不同中性化设定、不同成本假设；并检查常见异象：行业集中度、因子漂移、样本外衰减、极端行情崩坏等。\n- 4–6 分：有部分稳健性，但维度有限或缺少对失败情形的解释。\n- 0–3 分：几乎无稳健性检验，结论脆弱。\n\n### （4）归因与对照实验（4 分）\n- 4 分：能做“为什么提升/为什么不提升”的归因（如暴露分解、与 size/value/momentum 的相关与残差解释、收益来源拆解）；对照组设计合理（仅改变一个环节：中性化/价格剥离/as-of）。\n- 2–3 分：有对照但不够严格（多处同时改动导致无法归因）。\n- 0–1 分：无对照与归因，只有结果展示。\n\n---\n\n## 4. 可交易性、风险与研究表达（共 15 分）\n\n### （1）交易成本、换手、回撤与敏感性分析（8 分）\n- 7–8 分：除年化收益外，系统给出 `换手率`、`最大回撤/回撤持续期`、`成本后收益`，并做成本敏感性（佣金/冲击/滑点不同假设）与容量/拥挤度讨论（A 股流动性分层）。\n- 4–6 分：提到交易成本或回撤，但缺乏系统敏感性，或成本模型过于理想化且未说明。\n- 0–3 分：忽略成本与换手，或只给毛收益结论。\n\n### （2）风险控制与组合构建可落地性（4 分）\n- 4 分：说明组合构建方式（分层/打分/优化），并考虑风险约束（行业偏离、市值暴露、beta、波动、个股权重上限等）；若能说明与多因子框架的融合方式更佳。\n- 2–3 分：有组合构建描述，但风险约束与实盘可行性不足。\n- 0–1 分：仅停留在“买最高卖最低”理想化描述。\n\n### （3）复现透明度、假设披露与合规意识（3 分）\n- 3 分：清晰披露数据口径与关键假设（财报字段、公告日来源、复权、剔除规则），给出可复现流程（伪代码/关键公式/参数表）；同时提示数据合规与使用边界（商业数据库、公告数据许可等）。\n- 1–2 分：披露不完整但基本可理解。\n- 0 分：关键口径与假设缺失，难以复现且存在潜在合规风险。"
  },
  {
    "id": "24",
    "question": "请以东山精密（002384.SZ）为研究对象，围绕“消费电子基本盘稳固，汽车再造东山”这一主题，完成一份公司深度研究报告，重点分析以下问题：\n\n1. 消费电子“基本盘”分析： 复盘公司在消费电子领域（重点包括 FPC 软板与 PCB 硬板）的市场地位，分析其在核心大客户供应链中的份额变化与竞争优势，并论证该业务如何支撑稳定现金流与技术积累；\n2. 汽车业务“再造”逻辑拆解： 分析公司新能源汽车战略路径，重点讨论其通过并购（如 Multek）与内生发展切入三电系统、车身轻量化与车载显示等领域的逻辑，并进一步评估：并购定价与估值合理性、整合与协同效应（synergy）兑现情况、业务增长与客户拓展进展，以及“外延并购 vs 内生发展”的策略取舍，并判断其在核心车企中的单车价值量提升空间；\n3. 财务预测与估值重构： 基于“消费电子 + 新能源汽车”双轮驱动框架，预测未来三年收入与利润结构变化，结合可比公司（如立讯精密、鹏鼎控股）的估值水平，分析汽车业务占比提升对估值中枢（PE）的影响，并给出投资评级与关键风险提示。",
    "classification": "Capital Markets-Company & Equity Research",
    "classification_code": "CAP-COM",
    "report_type": "In-Depth Equity Research Report on Electronics Manufacturing Company",
    "report_type_zh": "电子制造公司深度研究报告",
    "language": "zh",
    "expert_evaluation_criteria": "## 东山精密（002384.SZ）公司深度研究报告评价标准\n\n本评价体系用于对“公司深度研究报告（消费电子+汽车双主线）”进行可执行的量化评估，兼容人工评审与大模型自动评分的一致性。评分强调：产业与公司逻辑的可证伪性、关键数据的可追溯性、财务预测可复现性、估值重构的严谨性，以及投资结论的可操作性。体系共设4个一级维度、15个二级维度，总分100分。\n\n---\n\n### 1. 核心观点与研究增量（20分）\n\n#### （1）投资主线清晰度与可证伪性（8分）\n- 7–8分：明确给出“基本盘稳 + 汽车再造”的结论式观点，并拆成可检验命题（如份额变化、利润/现金流贡献、汽车业务放量条件与时点）；时间维度清晰（未来1–3年/三年预测口径一致），能指出关键观测指标与触发条件。  \n- 4–6分：主线明确但偏口号化，缺少可量化的检验指标或时间指引；论证更多停留在描述层。  \n- 0–3分：观点模糊或前后不一致，难以判断结论成立条件与证伪路径。\n\n#### （2）信息/认知增量与差异化（6分）\n- 5–6分：提供相对稀缺或高价值增量（如更细颗粒度的客户/产品/产能/ASP/良率/份额变化拆解、产业链验证、竞争对手对比的关键差异点），并用于修正市场共识或形成差异化判断。  \n- 3–4分：有一定增量（补充细分数据或更细的结构拆分），但对结论的边际贡献有限。  \n- 0–2分：主要复述公开材料与常识性判断，缺乏有效新增信息与独立观点。\n\n#### （3）分析框架完整性与一致性（6分）\n- 5–6分：形成并实际应用“行业→竞争格局→客户/产品→产能/成本→财务→估值”的闭环框架；各章节口径一致（指标、期间、分部定义一致），逻辑链无明显跳跃。  \n- 3–4分：框架基本齐全，但部分环节（如竞争格局到财务传导、或汽车放量到估值重构）论证偏弱。  \n- 0–2分：结构散乱，章节之间缺乏有效衔接，结论与论证脱节。\n\n---\n\n### 2. 消费电子“基本盘”分析质量（25分）\n\n#### （1）FPC/PCB细分定位与行业坐标（8分）\n- 7–8分：清晰界定公司在FPC软板、PCB硬板等细分的产品结构与能力边界（工艺/层数/应用场景等以报告口径表述），并用可追溯数据刻画行业空间、竞争格局与公司相对位置（份额/排名/结构性强弱项）。  \n- 4–6分：能描述所处赛道与大致位置，但细分拆解不够、对比维度有限或数据颗粒度偏粗。  \n- 0–3分：停留在“行业景气/公司龙头”式表述，缺少细分定位与量化坐标。\n\n#### （2）核心大客户供应链份额变化与粘性（7分）\n- 6–7分：围绕“核心客户—核心产品—份额/订单变化—原因—可持续性”展开，能够区分结构性因素（技术/良率/交付/认证周期）与周期性因素（终端销量波动/库存周期），并给出验证证据与替代解释。  \n- 3–5分：讨论客户与份额变化，但证据链不完整（缺少数据、缺少时间序列或缺少竞争对手对照），可持续性判断偏经验。  \n- 0–2分：仅列举客户名单或泛谈绑定关系，无法支撑“基本盘稳固”。\n\n#### （3）竞争优势拆解的可落地性（5分）\n- 5分：能把“优势”拆到可衡量要素（成本、良率、交期、认证、产能爬坡、工艺壁垒、客户协同等），并说明这些要素如何转化为份额/盈利能力/现金流稳定性。  \n- 3–4分：能提出优势点，但缺少量化或缺少与竞争对手的对照与传导机制。  \n- 0–2分：优势表述空泛（如“技术领先、管理优秀”），缺乏证据与机制。\n\n#### （4）现金流与技术积累的论证强度（5分）\n- 5分：明确说明消费电子业务如何贡献稳定现金流（营运资本、资本开支、利润质量等维度可自洽），以及如何沉淀可迁移能力（工艺、客户体系、质量体系等）并对汽车业务形成支撑或协同。  \n- 3–4分：提到现金流/技术积累，但缺少财务指标支撑或协同路径不清。  \n- 0–2分：未能证明“基本盘”对公司整体的稳定器作用。\n\n---\n\n### 3. 汽车业务“再造东山”逻辑与兑现路径（25分）\n\n#### （1）新能源战略执行路径与业务版图映射（7分）\n- 6–7分：把“并购+内生”的路径拆为阶段性里程碑（进入领域、认证导入、放量、盈利改善等），并清晰映射到具体产品线/客户/产能布局；能够说明选择这些赛道的产业逻辑与公司能力匹配点。  \n- 3–5分：战略方向描述清楚，但路径颗粒度不够或缺少里程碑与验证指标。  \n- 0–2分：仅宏观叙事，缺乏执行层面的路径拆解。\n\n#### （2）并购（如 Multek）协同、整合与风险评估（6分）\n- 5–6分：明确并购标的带来的能力补足（客户、技术、资质、产能等）与协同路径，并对整合风险（盈利修复、客户稳定性、管理整合、商誉/减值等）给出可验证指标或情景分析。  \n- 3–4分：提到协同与风险，但分析偏概括，缺少量化或缺少跟踪指标。  \n- 0–2分：仅把并购作为利好叙述，忽略整合与风险约束。\n\n#### （3）切入领域逻辑与“单车价值量”测算可信度（8分）\n- 7–8分：围绕三电系统相关部件、轻量化精密制造、车载显示等（以题目要求为核心范围）分别回答：公司提供什么、竞争对手是谁、进入壁垒是什么、在客户车型中的价值量构成；单车价值量测算口径清晰（BOM/ASP/渗透率/份额/车型生命周期），并给出敏感性或区间。  \n- 4–6分：有价值量提升的方向判断，但测算口径不够透明或关键参数缺少依据。  \n- 0–3分：仅给出“价值量提升空间大”的结论，缺乏可复现测算。\n\n#### （4）核心车企客户导入与放量可行性（4分）\n- 4分：明确客户导入流程与放量约束（认证周期、定点节奏、产能/良率、价格与降本、供应链安全等），并与预测节奏一致；能区分“进入供应链”与“放量/提升份额”。  \n- 2–3分：提到客户拓展，但对放量约束与节奏把握不足。  \n- 0–1分：客户放量假设跳跃，与行业常识或自身论证不匹配。\n\n---\n\n### 4. 财务预测、估值重构与投资结论（30分）\n\n#### （1）三年预测模型结构与关键假设透明度（10分）\n- 9–10分：模型可复现：分部拆分（至少消费电子/汽车）、收入驱动项（量价、产能、份额、客户放量）、成本与费用逻辑（毛利、费用率、折旧摊销、研发等）、营运资本与资本开支假设清晰；关键假设有来源或对标，并做一致性校验（如产能—出货—收入匹配）。  \n- 5–8分：有三表或预测框架，但关键驱动假设不够透明、或分部拆分偏粗，复现性一般。  \n- 0–4分：预测更像“拍数”，缺少清晰假设与推导链条。\n\n#### （2）收入结构演变与盈利弹性解释（6分）\n- 5–6分：能把“消费电子+汽车”结构变化量化到收入占比、毛利结构、利润弹性与现金流影响，并解释驱动因素与边界条件（如价格压力、稼动率、规模效应、产品结构升级）。  \n- 3–4分：描述了结构变化趋势，但对利润与现金流的传导解释不足。  \n- 0–2分：仅给趋势结论，无法说明盈利弹性从何而来。\n\n#### （3）估值方法选择与“PE中枢重构”论证（8分）\n- 7–8分：估值方法与公司阶段匹配（PE/PEG、SOTP、DCF等选择有理由）；可比公司选择合理且口径可比（业务结构、成长性、盈利质量）；能够明确说明“汽车业务占比提升→市场给予更高/不同PE”的机制，并用情景或SOTP展示估值中枢变化。  \n- 4–6分：给出可比估值与结论，但口径对齐不足，或“PE中枢变化”的机制阐释偏弱。  \n- 0–3分：估值对比随意、口径混乱，或无法支撑目标估值/中枢变化。\n\n#### （4）投资评级、催化剂、风险与情景分析（6分）\n- 5–6分：给出明确投资评级与逻辑闭环（为何买/为何不买）；列出可跟踪催化剂与兑现时点（订单、定点、产能爬坡、利润拐点等）；风险提示具体且有影响路径，并提供情景/敏感性（关键假设变化对EPS/估值的影响）。  \n- 3–4分：有评级与风险，但催化剂不够可跟踪或情景分析不足。  \n- 0–2分：评级结论与正文脱节，风险提示流于模板化。"
  },
  {
    "id": "25",
    "question": "为支持银行系统研判海南自由贸易港封关运作背景下的金融发展环境与业务布局方向，需要基于权威政策文件对相关制度安排与政策导向进行系统性研究。\n\n请结合《海南省委关于国民经济和社会发展第十五个五年规划的建议》等公开政策文件，撰写一份结构化研究报告，重点完成以下分析：\n\n1. 关键政策方向与发展目标梳理\n系统梳理海南“十五五”规划中与自由贸易港建设相关的核心政策方向、发展目标与重点任务，重点关注封关运作背景下的制度设计与政策取向；\n\n2. 自贸港封关运作的制度安排分析\n分析自贸港封关后在贸易、投资、跨境资金流动、金融监管等方面可能发生的制度变化及其对金融体系的影响；\n\n3. 银行业务机遇、创新方向与挑战评估\n结合银行业务实践，重点分析封关运作后银行在跨境金融、贸易金融、资金管理及金融服务创新等领域可能面临的业务机遇、创新空间与潜在风险挑战。",
    "classification": "Banking-Customer & Marketing Management",
    "classification_code": "BNK-CMM",
    "report_type": "Free Trade Port Policy Analysis and Banking Opportunity Assessment Report",
    "report_type_zh": "自贸港政策解读与银行业务机遇分析报告",
    "language": "zh",
    "expert_evaluation_criteria": "## 海南自贸港封关运作政策研究报告评价标准\n\n本评价体系用于量化评估“海南自由贸易港封关运作背景下金融发展环境与银行业务布局”研究报告质量，强调**政策引用权威性与可追溯性、制度推演的逻辑闭环、银行策略的可落地性、以及风险合规意识**。体系共设4个一级维度、12个二级维度，总分100分，适配人工评审与LLM自动评分的一致性应用。\n\n---\n\n### 1. 关键政策方向与发展目标梳理质量（30分）\n\n#### （1）政策覆盖完整性与层级识别（10分）\n- 9–10分：覆盖“十五五”相关文件中与自贸港/封关运作/金融开放密切相关内容，并能区分**中央—地方、法律—规划—配套细则**的层级关系；关键政策点不遗漏（含重点任务、改革举措、开放安排、监管导向、时间表/里程碑等）。\n- 6–8分：覆盖主要政策方向，但层级区分不够清晰，或对部分关键领域（如跨境资金、监管框架、产业导向与金融结合点）覆盖不充分。\n- 0–5分：政策点零散或以二手解读为主，缺乏系统梳理；重要政策遗漏明显或层级混淆。\n\n#### （2）政策解读准确性与条款级可追溯引用（10分）\n- 9–10分：对政策表述解读严谨，能做到**“原文要点—作者解释—银行含义”**三段对应；关键结论均有明确出处（文件名称、发布时间/机构、章节或条款要点），不存在扩大解释或断章取义。\n- 6–8分：引用较充分但不够规范（缺条款要点/出处不全），或少量解释带主观推断但整体不影响结论可信度。\n- 0–5分：大量结论无依据；存在明显误读、把“方向性表述”当“已落地制度”的情况，影响报告可信度。\n\n#### （3）目标—任务—抓手的结构化梳理能力（10分）\n- 9–10分：将规划目标、重点任务、制度抓手按逻辑结构化呈现（如“发展目标→重点领域→制度工具→预期效果/指标”）；能提炼对金融相关的可检验要点（如开放领域、监管口径、跨境要素便利化方向等）。\n- 6–8分：结构基本清晰，但多为罗列；金融相关提炼不够聚焦，缺少“可检验要点/指标化表达”。\n- 0–5分：结构松散、堆砌原文或口号化归纳，难以支撑后续制度推演与银行建议。\n\n---\n\n### 2. 封关运作制度安排与影响机制分析（30分）\n\n#### （1）制度要素拆解的全面性（贸易/投资/资金/监管）（12分）\n- 11–12分：围绕封关运作，分别对**贸易便利化与监管、投资准入与负面清单逻辑、跨境资金流动与账户/结算体系、金融监管与风险隔离**等要素进行拆解；能识别“政策方向、制度工具、可能变化点、边界条件”。\n- 7–10分：覆盖大部分要素，但某一关键维度分析偏薄（如只讲贸易不讲资金与监管联动），或对制度边界条件讨论不足。\n- 0–6分：仅停留在概念描述，对“制度安排是什么/可能如何变化”缺乏要素化拆解。\n\n#### （2）对金融体系的传导路径与影响评估（10分）\n- 9–10分：形成清晰因果链条（示例：制度变化→跨境资金/结算/融资需求变化→银行资产负债与中收结构影响→风险暴露与监管要求变化）；能区分短期与中长期影响、直接与间接影响。\n- 6–8分：能描述主要影响方向，但传导链条不够闭环（缺“通过什么机制影响银行/市场”），或缺少分期/分场景讨论。\n- 0–5分：影响判断泛化（如“利好金融发展”），缺乏机制解释或前后矛盾。\n\n#### （3）关键假设、时点与不确定性管理（8分）\n- 7–8分：明确列出关键假设（至少3项，如封关节奏、监管口径、跨境资金管理强度、产业落地进度等），并给出时间维度（如未来6–12个月/规划期内）或触发条件；对不确定性提供情景推演或敏感性讨论。\n- 4–6分：有假设或时间判断，但隐含较多、缺少触发条件；情景分析较弱。\n- 0–3分：无关键假设；把不确定事项写成确定结论，或完全缺少时点与条件约束。\n\n---\n\n### 3. 银行业务机遇、创新方向与挑战评估（25分）\n\n#### （1）业务机遇识别的贴合度与优先级排序（10分）\n- 9–10分：围绕银行核心条线（跨境金融、贸易金融、资金管理/现金管理、投融资与银团、金融市场与结售汇、托管与供应链金融、机构业务等）提出**具体机遇点**，并按市场空间、可获得性、风险成本、竞争态势进行优先级排序。\n- 6–8分：方向基本正确但偏泛（“发展跨境业务、加强贸易金融”），优先级与取舍依据不足。\n- 0–5分：机遇点与封关制度变化关联弱，或明显脱离银行业务实践。\n\n#### （2）产品/方案设计与落地路径（10分）\n- 9–10分：给出可执行的业务方案（至少2–3项），包含目标客群/场景、产品结构或服务流程、关键依赖（系统/账户/清算通道/数据）、协同部门、里程碑与KPI建议；能体现“制度变化如何转化为产品与收入/客户价值”。\n- 6–8分：有方案雏形，但缺少关键落地要素（流程、系统、协同、指标），或可执行性论证不足。\n- 0–5分：停留在口号或原则性建议，无法指导业务动作。\n\n#### （3）挑战识别与能力建设建议（5分）\n- 5分：系统识别制约因素（资本/流动性、合规审查、人行/外汇/反洗钱要求、IT与数据、客户准入、跨境法律税务、同业竞争等），并提出对应能力建设路径（模型/流程/系统/人才/合作伙伴）。\n- 3–4分：能提挑战但缺少应对路径，或建议较笼统。\n- 0–2分：基本不谈挑战与能力约束，或将挑战简单归因为外部因素。\n\n---\n\n### 4. 风险合规意识与研究规范（15分）\n\n#### （1）风险识别与合规边界（8分）\n- 7–8分：覆盖并能解释影响路径的风险清单，至少包含：跨境资金异常与套保/套利风险、反洗钱与反恐融资（AML/CFT）、制裁与声誉风险、操作与模型风险、监管政策变动风险、跨境法律与税务合规；提出可执行的控制建议（准入、尽调、监测、限额、报告机制等）。\n- 4–6分：提及主要风险，但与业务/制度变化的对应关系不够清晰，控制建议偏原则。\n- 0–3分：缺乏风险合规讨论，或把合规当作一句话带过。\n\n#### （2）数据与方法透明度（4分）\n- 4分：清晰列明信息来源范围（政策原文、官方解读、权威统计/监管口径等），并说明分析方法（框架、推演步骤、假设来源）；结论可复核。\n- 2–3分：有来源与方法描述，但不完整（如仅列参考资料不说明如何使用）。\n- 0–1分：基本无来源/方法披露，难以验证。\n\n#### （3）结构化表达与可读性（3分）\n- 3分：报告结构与题目三大任务高度对应；图表/清单/框架图使用恰当；关键结论有“要点化输出”（便于管理层阅读）。\n- 2分：结构基本对应但层次不够清晰，重点不突出。\n- 0–1分：结构混乱、冗长重复，影响使用。"
  },
  {
    "id": "26",
    "question": "请围绕“阿里云 AI Capex (AI 资本开支) 的经济性与可持续性”这一主题，撰写一份深度研究报告，系统评估其在 AI 驱动周期中的增长潜力、盈利能力及长期资本回报，重点分析以下问题：\n\n1. 在中美科技与芯片供给不确定性的背景下，分析阿里巴巴 AI 资本开支的结构特征，并解释算力供给、服务器配置与 GPU 可得性对云业务扩张的影响，同时评估未来国产算力生态演进将如何影响公司 AI 资本开支预期；\n2. 基于 AI 云计算的单位经济学框架，对比 AI 训练与推理业务与传统公有云在收入结构、成本构成及 EBITA/EBITDA 利润率方面的差异，并评估高折旧背景下的利润释放节奏；\n3. 结合中国云计算竞争格局与 AI 应用形态演进，分析阿里云 AI 投资对中期收入增长与 ROIC 的影响，并基于阿里巴巴集团业务结构拆分及盈利预测讨论公司价值重估的合理性；\n4. 在上述分析基础上，识别关键风险，包括芯片供给、AI 定价下行、政策不确定性及其他可能风险，并评估其对长期回报的潜在影响。",
    "classification": "Capital Markets-Company & Equity Research",
    "classification_code": "CAP-COM",
    "report_type": "Economic Analysis of Alibaba Cloud AI Capital Expenditure",
    "report_type_zh": "阿里巴巴云计算与 AI 资本开支经济性分析",
    "language": "zh",
    "expert_evaluation_criteria": "## 阿里云 AI Capex 经济性与可持续性深度研究报告评价标准\n\n本评价体系用于对阿里云 AI Capex 经济性与可持续性深度研究报告进行可执行的量化评估，强调“供给链与Capex结构 → unit economics → 利润/现金流释放 → ROIC与估值 → 风险情景”的闭环论证。体系共设 4 个一级维度、12 个二级维度，总分 100 分，支持人工评审与大模型（LLM）一致性评分。\n\n### 1. Capex 结构与算力供给约束分析（共 30 分）\n\n#### （1）Capex 口径定义与结构拆解（10 分）\n- **9–10 分**：清晰界定 AI Capex 的会计/经营口径（资本化范围、折旧年限假设、是否含数据中心土建/租赁改造、服务器与网络设备、预付款/在建工程等），并对 Capex 进行结构化拆解（至少覆盖 GPU/加速卡、CPU/内存、网络互联、存储、机房与电力配套），能映射到阿里云业务扩张所需的关键资源约束。\n- **6–8 分**：给出主要 Capex 构成与方向判断，但口径边界不够严谨（如 AI Capex 与通用 Capex 混用），或拆解颗粒度不足，难以支持后续经济性测算。\n- **0–5 分**：仅罗列 Capex 增长或泛泛描述“加大投入”，缺少结构、口径与可用于分析的分类。\n\n#### （2）中美科技/芯片不确定性下的供给链分析（10 分）\n- **9–10 分**：系统讨论出口管制/合规约束下的 GPU 可得性（型号/替代方案/交期/采购策略）、国产替代路径及其对性能/能效/软件生态的影响；明确指出供给链不确定性如何影响算力上线节奏、客户交付能力与收入确认（例如“供给→交付→计费→收入”的链条）。\n- **6–8 分**：识别到供给不确定性与国产替代主题，但影响机制较概括，缺少对“交付节奏/客户结构/收入兑现”的具体映射。\n- **0–5 分**：将供给风险停留在新闻层面，缺少与阿里云扩张约束的因果关系。\n\n#### （3）服务器配置/集群架构与扩张约束的技术—经营联动（10 分）\n- **9–10 分**：把服务器配置与集群能力转化为可经营指标（如 GPU 利用率、可售 GPU 小时、网络瓶颈/互联架构、功耗与机柜密度、PUE、电力获取与并网周期），并解释这些指标如何决定供给上限、单位成本与毛利；能区分训练与推理对网络、显存、吞吐的差异化需求。\n- **6–8 分**：能描述基础架构要素与若干指标，但缺乏与收入/成本/扩张节奏的量化联动，相关测算缺乏论证支撑。\n- **0–5 分**：技术描述与经营判断割裂，或仅用笼统措辞（“算力更强”“配置升级”）替代分析。\n\n---\n\n### 2. AI 云 Unit Economics 与盈利能力测算（共 30 分）\n\n#### （1）收入结构与定价/计费单位对比（10 分）\n- **9–10 分**：建立清晰的对比框架：AI 训练、AI 推理、传统公有云（通用计算/存储/网络）的收入来源、计费单位（GPU小时/Token/调用次数/实例规格等）、合同形态（按量/包年包月/预留实例/专属算力）、客户结构差异；能解释业务组合变化对 ARPU、收入可持续性与价格弹性的影响。\n- **6–8 分**：对比维度基本覆盖，但对定价机制与收入弹性解释不够，或训练/推理与传统云边界不清。\n- **0–5 分**：仅做“AI 更贵/更快增长”的定性判断，缺乏可评估的收入拆解。\n\n#### （2）成本构成拆解与 EBITA/EBITDA 桥接（10 分）\n- **9–10 分**：将成本按云业务UE模型进行拆分，包括固定成本的折旧/摊销（D&A）与可变成本的变动，如机柜等硬件设备、软件授权使用费、电力基础设施、通信互联、运维与人力、渠道与营销费用等；构建从毛利→EBITA/EBITDA 的桥接逻辑，并解释训练/推理 vs 传统云的成本驱动差异（利用率、峰谷、网络与存储占比、支持成本等）。\n- **6–8 分**：成本项覆盖主要部分，可以从不同角度解释成本驱动因素，但桥接不完整（如混用 EBITA/EBITDA、未说明口径），或关键成本（折旧/电力/带宽）缺少量化抓手。\n- **0–5 分**：成本分析碎片化，利润率结论缺少支撑或口径混乱。\n\n#### （3）高折旧背景下的利润释放节奏与现金流一致性（10 分）\n- **9–10 分**：明确解释“Capex→在建/投产→折旧→EBITDA/EBITA→经营现金流/自由现金流”的时序关系；给出利润释放节奏的量化判断（例如按利用率爬坡、价格下行、折旧年限、资本化比例等做敏感性），并能讨论资本开支前置导致的短期利润承压与中长期经营杠杆。\n- **6–8 分**：能指出折旧压力与利润滞后，但缺少节奏测算或对现金流口径的统一解释。\n- **0–5 分**：忽略折旧与现金流错配，或仅用结论性语言替代推导。\n\n---\n\n### 3. 竞争格局、增长潜力与 ROIC/估值评估（共 25 分）\n\n#### （1）中国云与 AI 竞争格局的结构化判断（8 分）\n- **7-8 分**：对主要竞品与差异化维度形成结构化框架（价格/产品矩阵/政企与行业渗透/生态与渠道/算力供给与交付能力/自研芯片与软硬协同等），并能解释行业竞争对定价、利用率与投入强度的约束。\n- **4–6 分**：覆盖主要竞争者与趋势，但分析停留在“份额/价格战”层面，缺乏对 AI 应用形态变化的联动。\n- **0–3 分**：竞争分析泛化或与 AI Capex 主题脱节。\n\n#### （2）中期收入增长与 ROIC 驱动因素测算（10 分）\n- **9–10 分**：给出可检验的中期增长框架（例如容量供给×利用率×单位价格×业务组合），并将其与投入资本（Capex、营运资本、资本化软件等）联动，形成 ROIC（或 ROIIC）测算逻辑；能说明推理占比上升、利用率提升、软件栈与平台服务附加值等对 ROIC 的影响路径。\n- **6–8 分**：能讨论增长与资本效率方向，但 收入增长驱动拆分、ROIC 计算口径、关键假设或传导机制不够清晰。\n- **0–5 分**：只谈收入增长、ROIC，不给出测算过程及依据；或 ROIC 结论与前文 Capex/利润分析断裂。\n\n#### （3）估值重估的合理性与对标框架（7 分）\n- **6–7 分**：估值讨论与“ROIC–WACC–增长–再投资率”一致，或提供可比公司/可比业务对标（云厂商、AI 基础设施、国内外同业）并解释差异来源；能指出重估的触发条件（利用率拐点、利润率修复、供给缓解、产品结构升级等），结合阿里巴巴集团业务盈利预测给出估值重估的合理性。\n- **3–5 分**：有对标或估值方法，但与经营假设耦合弱，触发条件不清，重估论证缺乏逻辑支撑。\n- **0–2 分**：估值仅给倍数结论，缺少逻辑链条与条件约束。\n\n---\n\n### 4. 风险识别、情景推演与可持续性（共 15 分）\n\n#### （1）关键风险全面性与影响传导机制（7 分）\n- **6–7 分**：覆盖题干要求的四类风险（芯片供给、AI 定价下行、资本效率下降、政策不确定性），并补充与主题强相关的次级风险（能耗与电力获取、客户集中度、技术路线变化/开源替代、合规与数据安全等）；每项风险说明“发生条件→影响变量（价格/利用率/Capex/折旧/交付）→对 ROIC/FCF 的影响方向”。\n- **3–5 分**：风险覆盖较全，但多为罗列，缺少传导机制或量化抓手。\n- **0–2 分**：风险缺失、与主题不相关或无法用于投资/经营决策。\n\n#### （2）情景分析/敏感性分析的可执行性（4 分）\n- **4 分**：至少给出 2–3 个情景（基准/乐观/悲观）或关键变量敏感性（GPU 价格与交期、利用率爬坡、ASP 下行、折旧年限、资本化比例等），并明确对收入、EBITA/EBITDA、FCF、ROIC 的影响量级或方向边界。\n- **2–3 分**：有情景想法但变量选择不关键、参数不透明或无法复现。\n- **0–1 分**：无情景与敏感性分析。\n\n#### （3）长期回报及可持续性：风险情形与资产定价（4 分）\n- **4 分**：量化评估关键风险在特定风险情形下的长期回报影响，分析相关风险情形下资本纪律（回收期/IRR、阶段性投入与止损机制）、资产更新周期（GPU 迭代导致的经济寿命风险）、以及能耗/政策约束等因素对长期 Capex 强度与资本回报的影响。\n- **2–3 分**：进行长期回报评估但缺乏逻辑推导，未对中长期资产定价影响展开讨论。\n- **0-1 分**：未讨论长期可持续性。"
  },
  {
    "id": "27",
    "question": "在医保支付方式改革（DRG/DIP）持续推进、创新药加速上市、商业健康险增速放缓的背景下，商业健康险如何在支持创新药支付、提升患者医疗可及性方面发挥更大作用，成为行业关注焦点。\n\n请撰写一份题为《商业健康险如何打通创新药“最后一公里”？》的深度研究报告，并重点分析以下问题：\n1. 在我国医保控费与公立医院绩效考核约束下，创新药在“进院—处方—支付”链条中面临的主要制度性堵点；\n2. 商业健康险、医药企业与医疗机构开展协同创新的主要模式及其对创新药可及性与可支付性的影响；\n3. 结合典型实践案例，评估商业健康险在支付机制、产品设计与服务模式上的改进空间；\n4. 从监管协同、数据互通和产品体系建设角度，提出推动商业健康险更有效参与多层次医疗保障体系的政策与业务建议。",
    "classification": "Insurance-Health Insurance",
    "classification_code": "INS-HLT",
    "report_type": "Research Report on Commercial Health Insurance Participation in Innovative Drug Payment and Healthcare Collaboration",
    "report_type_zh": "商业健康险参与创新药支付与医疗协同机制研究报告",
    "language": "zh",
    "expert_evaluation_criteria": "## 商业健康险与创新药协同发展研究报告评价标准\n\n本评价体系适用于评估以 商业健康保险与创新药协同发展模式 为主题的研究报告质量，兼容人工评审与大模型自动评分。共设4个一级维度、12个二级维度，总分100分。评分重点关注：**制度堵点拆解深度、协同模式的机制清晰度与可复制性、案例与数据方法的可验证性、建议的可落地性与合规风控完备性**。\n\n---\n\n## 1. 制度性堵点诊断与链条机制解释（30分）\n\n### （1）政策与制度环境理解准确性（8分）\n- 7–8分：准确把握DRG/DIP“总额/打包付费—控费—行为响应”的机制，能区分医保目录/谈判药、双通道、集采、处方管理、医院绩效考核（如国考/DRG结余留用等）对创新药准入与使用的不同影响；引用政策口径基本无误、边界清晰。\n- 4–6分：能描述主要政策方向，但对DRG/DIP传导机制、医院绩效约束或目录/双通道边界存在概念混用或关键遗漏。\n- 0–3分：政策理解明显错误或停留在口号式表述，无法支撑后续分析。\n\n### （2）“进院—处方—支付”链条拆解完整性与因果逻辑（12分）\n- 10–12分：将链条分解为可检验环节（如：医院准入/药事会→临床路径与科室预算→医生处方激励与合规风险→患者自付与支付能力→商保理赔与直付→结算与风控），并说明每一环的**约束变量、决策主体、利益冲突与结果**；逻辑闭环、少跳跃。\n- 6–9分：链条拆解较完整，但对关键一到两个环节（常见为“处方端激励/合规”或“支付端风控/结算”）解释不足，因果链偏弱。\n- 0–5分：只列堵点清单，不解释机制或前后矛盾。\n\n### （3）利益相关方激励与制度性“堵点”优先级判断（10分）\n- 9–10分：能从医院（成本/考核/预算）、医生（合规/收益/学术）、患者（自付/可得性/信息不对称）、医保（预算约束/公平性）、商保（逆选择/赔付波动/获客成本）、药企（放量/价格体系/证据生成）多方激励出发，给出**堵点优先级**与“为什么是制度性堵点”的判断；可指出不同地区/医院层级的差异。\n- 6–8分：覆盖主要主体，但优先级与权重判断偏主观或缺少约束条件说明。\n- 0–5分：主体视角单一（仅患者或仅保险），缺乏激励分析。\n\n---\n\n## 2. 协同创新模式与对可及性/可支付性的影响评估（30分）\n\n### （1）协同模式覆盖与分类框架清晰度（10分）\n- 9–10分：建立清晰分类框架并覆盖主流路径，例如（不限于）：特药险/院外购药保障、惠民保与创新药责任扩展、直付网络与DTP/双通道衔接、药企风险分担（价格-销量、疗效付费/结果付费、封顶/共付）、保险+健康管理/患者管理、创新药真实世界证据（RWE）共建等；能说明各模式适用疾病/药品阶段与准入条件。\n- 6–8分：列举较多模式，但分类标准混乱或未说明适用边界。\n- 0–5分：模式零散罗列，缺乏结构化框架。\n\n### （2）影响评估：可及性、可支付性与行为响应（12分）\n- 10–12分：对每类模式至少从两类维度评估其效果与副作用：  \n  - 可及性：进院率/处方渗透、院内外可得性、患者启动治疗时间等；  \n  - 可支付性：患者自付下降、年度封顶、支付连续性；  \n  - 行为响应：医院控费行为、处方合规、诱导需求与过度医疗风险；  \n  能明确“改善来自哪里”（结算便利、风控前置、证据换准入、风险共担等）。\n- 6–9分：能讨论方向性影响，但缺少指标化表达或未讨论行为响应/副作用。\n- 0–5分：只谈“促进可及性/减轻负担”，缺乏机制解释或评估维度。\n\n### （3）交易结构与风控可行性（8分）\n- 7–8分：清楚描述协同中的关键“结构件”：责任边界、触发条件（适应症/用药线别/既往症）、审核与授权（PA）、反欺诈、费用控制（共付/免赔/封顶/网络管理）、与药企/医院的数据与对账闭环；能指出落地难点与替代方案。\n- 4–6分：提及风控与结算，但流程不完整或忽视逆选择/道德风险。\n- 0–3分：将协同简单等同于“多赔一点/放开赔付”，缺乏风控设计。\n\n---\n\n## 3. 案例研究质量、数据与方法可验证性（25分）\n\n### （1）案例选择代表性与对比设计（10分）\n- 9–10分：案例具备代表性与可比性（例如不同地区、不同产品形态、不同病种/药品阶段），并采用对比或分层（前后对比/同类对比/机制对比）来支持结论；能说明案例选择标准与局限。\n- 6–8分：有案例但偏“宣传式成功案例”，对比不足或选择理由不清。\n- 0–5分：案例缺失、过旧或与题目（创新药支付/最后一公里）关联弱。\n\n### （2）数据来源可靠性、颗粒度与合规边界（8分）\n- 7–8分：数据来源清晰（公开统计、监管口径、机构披露、理赔/处方/购药等脱敏汇总数据、调研访谈），颗粒度能支撑分析（病种/药品/渠道/地区等）；同时明确数据使用合规边界（脱敏、授权、PIPL等）。\n- 4–6分：数据基本可信，但口径说明不足或颗粒度偏粗，仅能支持宏观判断。\n- 0–3分：数据来源不明、口径混乱或存在明显不合规使用倾向。\n\n### （3）分析方法透明度与可复现性（7分）\n- 6–7分：明确给出分析步骤与关键参数（如预算影响、赔付率/费用率拆分、共付与封顶对患者自付的影响测算、风险分担方案的情景分析），结论可被复核；能区分描述性事实与推断性结论。\n- 3–5分：有方法但披露不完整，关键假设隐含或无法复核。\n- 0–2分：仅观点输出，无方法与假设说明。\n\n---\n\n## 4. 政策与业务建议的可落地性、协同治理与风险合规（15分）\n\n### （1）监管协同与多层次保障定位清晰度（6分）\n- 5–6分：能在“基本医保—大病保险—医疗救助—商保（含惠民保/中高端/团险等）”框架下界定商保职责，提出与医保支付方式改革相容的监管/政策建议（如支付衔接、目录与责任边界、结算规则、试点机制）；建议具有可执行抓手。\n- 3–4分：定位基本清晰，但建议偏原则、缺少执行路径或部门协同机制。\n- 0–2分：定位模糊，建议与现行监管逻辑冲突或不可操作。\n\n### （2）数据互通与治理方案（4分）\n- 4分：提出可实施的数据互通路径（如标准化字段、对接接口、授权与脱敏机制、第三方可信计算/隐私计算等）并说明用途（直付、风控、疗效评估、RWE）；兼顾隐私安全与最小必要原则。\n- 2–3分：提及数据互通但停留在愿景层，缺少治理与安全设计。\n- 0–1分：忽视数据治理与隐私合规。\n\n### （3）产品体系建设与商业可持续（5分）\n- 5分：能从产品分层（普惠/补充/高端）、责任设计（免赔/共付/封顶/网络/特药目录）、定价与精算可行性（逆选择、风险池稳定、续保与长期性）、运营（理赔效率、直付与控费）提出系统方案，并说明在“增速放缓”背景下的可持续增长路径。\n- 3–4分：有产品建议但偏单点优化，缺少可持续性（赔付波动、费用率、获客成本）讨论。\n- 0–2分：建议泛化或明显不可持续（只强调扩大赔付、不谈风险池与控费）。"
  },
  {
    "id": "28",
    "question": "你是一名商业银行信贷领域的资深专家，长期从事行业风险研判与信贷决策支持工作。现需编制一份 《2025 年第四季度 AI 算力硬件产业链风险分析报告》，为商业银行信贷管理与授信决策提供参考依据。研究重点为 AI 大模型算力需求快速增长背景下，相关硬件价格上涨与供需变化所带来的风险影响。报告需满足以下要求：\n\n1.基于权威、可核验的公开数据，从宏观、中观、微观三个层面，系统分析 2025 年第四季度 AI 算力硬件产业链的风险变化情况。\n2.报告应明确列示关键数据指标及其来源，引用内容需来自具有公信力的权威机构或公开披露资料，确保数据真实、可追溯，不得基于未经验证的假设进行推断。\n3.分析结论需服务于商业银行信贷管理实践，重点说明对信贷准入、授信额度管理及风险控制的启示，结论应以事实和数据为依据，避免主观判断。",
    "classification": "Banking-Risk Management",
    "classification_code": "BNK-RSK",
    "report_type": "Industry Credit Risk Analysis Report",
    "report_type_zh": "行业信贷风险分析报告",
    "language": "zh",
    "expert_evaluation_criteria": "## AI算力硬件产业链风险分析报告评价标准\n\n本评分体系用于对AI算力硬件产业链风险分析报告质量进行可执行的量化评估，兼容人工评审与LLM自动评分。共设**4个一级维度、12个二级维度**，总分**100分**。评分重点：**产业链风险识别的专业深度、宏观-中观-微观传导的严谨性、数据可核验性、以及对商业银行授信决策的可落地性**。\n\n### 1. 产业链理解与风险识别质量（30分）\n#### （1）产业链拆解完整性与关键环节抓取（10分）\n- **9–10分**：清晰拆解AI算力硬件产业链（至少覆盖芯片/存储/封装代工/板卡与服务器/关键元件与材料/散热电源与互连/渠道与下游云与IDC等），明确各环节**价值量、瓶颈与议价结构**；能指出Q4时点的关键约束（如产能、交期、良率、先进封装能力、关键材料供给等）并与风险讨论直接对应。\n- **6–8分**：产业链覆盖较全，但关键瓶颈与风险关联不够紧；对部分环节（如封装、存储、光互连、电源散热等）论述偏泛。\n- **0–5分**：仅笼统描述“芯片/服务器/云厂商”，链条断裂或关键环节遗漏，难以支撑后续风险推导。\n\n#### （2）供需变化与价格机制分析（10分）\n- **9–10分**：围绕“需求增长—供给约束/释放—价格—利润—现金流—信用风险”的链条，建立可解释的机制；能区分不同产品的**定价逻辑**（如合约价/现货价、ASP变化、产品迭代导致的结构性涨价等），并解释Q4价格波动对上下游的传导与滞后。\n- **6–8分**：能描述供需与价格上行，但传导机制不够细（如缺少对库存、交期、订单可持续性、替代与降配的讨论）。\n- **0–5分**：将“价格上涨=风险上升/机会增加”简单等同，缺乏机制与可验证链条。\n\n#### （3）风险因子覆盖面与针对性（10分）\n- **9–10分**：风险识别全面且贴近信贷（至少覆盖：价格波动与周期性、产能/交付风险、技术迭代与产品替代、下游CAPEX波动、客户集中与议价、应收与回款、存货与跌价、出口管制/合规与地缘、汇率与融资环境等）；能明确**风险触发条件**与受影响主体。\n- **6–8分**：覆盖主要风险点，但缺少触发条件或对不同主体（上游/中游/下游）区分不足。\n- **0–5分**：风险罗列化、与“AI算力硬件”主题或Q4时点结合弱，缺乏信贷相关性。\n\n---\n\n### 2. 数据合规性、权威性与可核验性（30分）\n#### （1）数据来源权威性与可追溯引用（12分）\n- **11–12分**：关键结论均有**权威公开来源**支撑（如政府/监管与统计口径、上市公司公告与财报、权威行业机构报告、交易所披露、可信的价格与出货跟踪机构等）；引用具备**可核验要素**（机构名、报告/公告名、发布日期、指标口径/单位、原始链接或可定位信息）。\n- **8–10分**：大部分数据来源可靠，但部分关键指标缺少完整引用要素（如缺发布日期/口径），或混用多来源导致可追溯性下降。\n- **0–7分**：数据来源模糊（如“业内人士称/某机构预计”不注明）、无法核验，或出现明显不具备公信力的来源支撑核心结论。\n\n#### （2）关键指标体系与口径一致性（8分）\n- **7–8分**：明确列示并解释核心指标（宏观/中观/微观至少各有一组），指标口径一致、单位统一、同比/环比与区间定义清晰；能说明数据缺口与处理方式（如缺失值、不一致口径的对齐策略）。\n- **4–6分**：给出部分指标，但口径解释不足或指标与结论映射不清。\n- **0–3分**：指标零散、口径混乱，无法支撑可复核的推断。\n\n#### （3）分析方法透明度与可复现性（10分）\n- **9–10分**：明确说明测算/判断方法（如供需平衡表、产能与交期推导、价格敏感性/压力测试、情景分析、信用风险传导路径图等），关键参数与步骤可复核；能对关键假设做**边界声明**（哪些为事实、哪些为条件性推演）。\n- **6–8分**：有方法描述，但关键步骤或参数不完整，复现性一般。\n- **0–5分**：以结论代替方法，或关键结论依赖“未经验证假设”且未披露推导过程。\n\n---\n\n### 3. 宏观—中观—微观三层联动论证（25分）\n#### （1）宏观层面风险变量与传导（8分）\n- **7–8分**：覆盖并解释宏观变量对产业链的传导路径（如利率与融资环境、汇率、全球贸易与出口管制、产业政策与算力基础设施投入、能源与电力约束等），并明确指向Q4时间窗口的变化与不确定性。\n- **4–6分**：提及宏观因素，但与产业链/信贷风险传导结合弱。\n- **0–3分**：宏观部分与主题脱节或停留在口号式叙述。\n\n#### （2）中观层面行业景气、供给约束与结构分化（9分）\n- **8–9分**：对Q4行业景气度、产能/良率/交期、库存与渠道、价格趋势做结构化刻画；能够识别**结构性分化**（如不同芯片/存储代际、不同封装路线、不同下游客户类型的差异）。\n- **5–7分**：能描述行业趋势，但缺少结构分化或关键约束的量化支撑。\n- **0–4分**：中观分析概念化，缺少Q4时点证据。\n\n#### （3）微观层面主体经营与信用风险画像（8分）\n- **7–8分**：能将风险落到可授信主体类型与关键财务/经营变量（如收入与毛利弹性、现金流与资本开支、杠杆水平、应收账款与账期、存货与跌价风险、客户集中、订单可持续性、关联交易与或有负债等），并解释其在价格上行/交期波动下的敏感性差异。\n- **4–6分**：提及财务指标但与产业链变量联动不足，或仅停留在单一指标判断。\n- **0–3分**：缺少微观主体分析，无法支持信贷决策。\n\n---\n\n### 4. 信贷决策可落地性与风控可执行性（15分）\n#### （1）信贷准入与行业/子赛道授信策略（8分）\n- **7–8分**：给出可执行的准入与差异化策略（如子赛道分层准入、名单制、授信集中度与限额逻辑、期限与还款结构、定价与风险溢价、担保/抵质押与增信安排建议），并明确与数据结论的对应关系。\n- **4–6分**：提出方向性建议，但缺少可操作细则（如缺少分层标准、限额依据或触发条件）。\n- **0–3分**：仅给出宏观口号式建议，无法直接用于授信管理。\n\n#### （2）风险控制措施与贷后监测预警指标（5分）\n- **5分**：形成可落地的贷后监测框架（至少包含价格/交期/库存、下游CAPEX、订单与回款、存货与应收、产能与良率、政策与出口管制等指标），并给出**预警阈值/触发事件**或可执行的监测频率建议。\n- **2–4分**：提出监测方向，但缺少阈值/触发条件或指标与风险事件对应不清。\n- **0–1分**：缺少贷后与预警体系设计。\n\n#### （3）结论表达的证据约束与合规边界（2分）\n- **2分**：结论与建议均明确标注证据来源或数据依据；对不可获得数据/不确定性有边界声明；避免将推演当作事实，符合银行风险文化与审慎原则。\n- **1分**：整体较审慎，但个别关键判断证据链不完整或表述偏主观。\n- **0分**：存在明显主观化断言、用未经验证假设支撑核心结论，或引用不可核验信息影响决策。"
  },
  {
    "id": "29",
    "question": "2026年，全球宏观经济进入新的变奏期。美联储货币政策路径的微妙变化、地缘政治格局的持续重塑，以及人工智能与能源转型对上游资源的结构性拉动，共同构成了有色金属行业复杂的定价环境。行业转向“金融属性、商品属性与战略属性”的三维分析范式，贵金属的避险与货币功能、工业金属的周期复苏弹性、小金属的战略属性溢价，呈现出显著的分化特征。\n\n请撰写一份《2026年全球有色金属行业投资策略：分化与重估》深度研究报告。请基于当前宏观与产业背景，自主搭建多维分析框架，系统梳理各细分板块的投资逻辑，并完成以下任务：\n1、宏观驱动逻辑拆解：深入拆解2026年影响有色板块的核心宏观变量（如实际利率走势、美元指数周期、全球制造业PMI趋势等），分析贵金属、工业金属、小金属三类资产对上述变量的敏感性差异，论证在不同宏观情景（如“软着陆”、“再通胀”或“滞胀”）下各类资产的相对表现逻辑。\n2、细分赛道深度研判：\n贵金属：结合央行购金行为、地缘风险溢价及货币信用体系变化，评估黄金、白银的中期配置价值及价格中枢。\n工业金属：聚焦铜、铝等核心品种，分析全球供给约束（矿山品位下降、资本开支不足）与新兴需求（AIDC、电网升级）的缺口演变，判断当前所处周期位置及本轮周期的持续性。\n小金属：筛选与AI算力、人形机器人等前沿技术强相关且具有战略价值的小金属（如锑、钨、稀土等），评估其供需平衡状态及战略资源属性。\n3、投资组合构建与标的推荐（核心任务）：\n基于上述分析，构建一个2026年有色金属行业的推荐组合。要求明确组合中贵金属、工业金属、小金属的配置比例及其背后的策略意图（如防守、进攻或对冲），在A股或港股市场中自主筛选具有代表性的推荐标的（每类至少2只），并就包括资源储量、成本曲线位置、产能释放节奏、估值水平及潜在催化剂等方面推荐理由给出充分详尽的论证。",
    "classification": "Capital Markets-Sector & Thematic Research",
    "classification_code": "CAP-STR",
    "report_type": "Global Nonferrous Metals Industry Investment Strategy Research Report",
    "report_type_zh": "全球有色金属行业投资策略研究报告",
    "language": "zh",
    "expert_evaluation_criteria": "## 《2026年全球有色金属行业投资策略：分化与重估》评价标准\n\n本评价体系用于评估有色金属行业投资策略类深度研究报告质量。评分重点在于：是否准确拆解宏观变量对不同金属板块的传导机制，是否清晰比较贵金属、工业金属、小金属的差异化投资逻辑，是否能结合供需、周期与战略属性完成赛道研判，并形成具备配置价值的组合与个股推荐。体系共设 **4 个一级维度、12 个二级维度，总分 100 分**。\n\n---\n\n### 1. 宏观框架与资产定价逻辑（30 分）\n\n#### （1）宏观变量识别与框架完整性（10 分）\n- **9–10 分**：准确识别实际利率、美元指数、全球制造业 PMI、通胀预期、地缘风险等核心变量，并形成清晰分析框架。\n- **6–8 分**：覆盖主要宏观变量，但框架性不足。\n- **0–5 分**：变量识别零散，缺乏主线。\n\n#### （2）三类金属的敏感性差异分析（10 分）\n- **9–10 分**：能清晰比较贵金属、工业金属、小金属对宏观变量的敏感性差异，基于历史数据给出明确的数量化关系，并解释背后机制。\n- **6–8 分**：有比较分析，但不够系统、数据支撑不严谨或机制展开不足。\n- **0–5 分**：仅做现象描述或粗糙判断，缺乏差异化逻辑。\n\n#### （3）宏观情景推演能力（10 分）\n- **9–10 分**：能够在软着陆、再通胀、滞胀等情景下推演各类资产相对表现，并给出清晰判断，商品价格判断给出严谨依据。\n- **6–8 分**：有情景分析，但结论不够完整或区分度不足，价格判断的严谨性不足。\n- **0–5 分**：缺乏情景推演，仅有大致方向判断或结论缺乏支撑。\n\n---\n\n### 2. 细分赛道深度研判（30 分）\n\n#### （1）贵金属配置逻辑（10 分）\n- **9–10 分**：能结合央行购金、地缘风险、货币信用与实际利率，分析黄金、白银的配置价值与价格中枢，基于商品价格与相关宏观因子关系的定量化测算给出价格判断依据。\n- **6–8 分**：逻辑基本完整，但价格判断或驱动分析较弱。\n- **0–5 分**：仅泛泛看多或看空，缺乏中期配置逻辑。\n\n#### （2）工业金属周期与供需研判（10 分）\n- **9–10 分**：围绕铜、铝等品种，清晰分析供给约束、需求拉动、库存对金属价格的影响，并判断当前周期位置及本轮周期持续性。\n- **6–8 分**：能识别主要供需因素，但周期判断不够深入，价格预测逻辑支撑不够严谨。\n- **0–5 分**：停留于景气描述，缺乏供需和周期框架，预测更偏向于“拍数”。\n\n#### （3）小金属战略属性分析（10 分）\n- **9–10 分**：能筛选与 AI、机器人等前沿需求相关的小金属，并分析其供需平衡、资源约束与战略溢价。\n- **6–8 分**：能指出重点品种，但战略属性与供需分析不够充分。\n- **0–5 分**：仅列举品种，缺乏逻辑论证。\n\n---\n\n### 3. 组合构建与标的推荐（25 分）\n\n#### （1）组合配置思路与比例设计（8 分）\n- **7–8 分**：明确贵金属、工业金属、小金属的配置比例，并清楚解释防守、进攻、对冲等策略意图。\n- **4–6 分**：有配置思路，但比例依据或策略目标不够清晰。\n- **0–3 分**：组合构建随意，缺乏统一逻辑。\n\n#### （2）个股筛选质量（9 分）\n- **8–9 分**：推荐标的具有代表性，覆盖资源储量、成本曲线、产能释放、行业地位等核心维度，并体现相对优势。\n- **5–7 分**：推荐逻辑基本成立，但比较性或深度不足。\n- **0–4 分**：个股推荐流于罗列，缺乏研究支撑。\n\n#### （3）估值与催化剂分析（8 分）\n- **7–8 分**：能结合估值水平、盈利弹性与潜在催化剂说明推荐合理性。\n- **4–6 分**：有估值或催化判断，但量化不足。\n- **0–3 分**：缺乏估值支撑，推荐可信度弱。\n\n---\n\n### 4. 报告规范、结论与风险提示（15 分）\n\n#### （1）结构与专业性（5 分）\n- **5 分**：结构完整，逻辑清晰，符合顶级策略报告规范。\n- **3–4 分**：结构基本完整，但层次或衔接一般。\n- **0–2 分**：结构松散，专业性不足。\n\n#### （2）投资结论可执行性（6 分）\n- **5–6 分**：结论明确，能提出清晰配置方向、重点品种与跟踪指标，具备实操价值。\n- **3–4 分**：结论较明确，但落地性一般。\n- **0–2 分**：结论空泛，缺乏配置意义。\n\n#### （3）风险识别完整性（4 分）\n- **4 分**：系统识别宏观波动、政策变化、需求不及预期、供给释放超预期、地缘风险缓和等主要风险，并说明影响路径。\n- **2–3 分**：列出主要风险，但分析不够深入。\n- **0–1 分**：风险提示缺失或流于形式。"
  },
  {
    "id": "30",
    "question": "随着稳定币在跨境支付、数字资产交易及金融基础设施中的应用加速扩展，全球主要金融市场正加快构建相应的监管框架。2025 年 8 月，中国香港《稳定币条例》正式生效，成为亚太地区法币稳定币监管的重要里程碑。然而，其对银行体系的实际影响仍存在分歧：稳定币究竟是支付与结算的补充工具，还是对传统存款体系构成潜在替代？\n\n请以香港《稳定币条例》落地为制度背景，结合国际稳定币市场的发展趋势及典型模式（如 USDT、USDC），围绕“稳定币监管化对金融体系与市场结构的影响”开展系统研究，重点完成以下分析任务：\n\n1. 存款替代效应评估：在香港稳定币发行规模达到千亿港元量级的情景下，其对商业银行零售存款及企业结算账户的替代效应有多大？不同类型银行中，哪些机构面临的冲击更为直接？\n\n2. 银行参与策略选择：在稳定币纳入监管体系后，香港商业银行应选择“直接申请发行牌照”，“作为托管与服务机构参与”，还是“阶段性观望等待”？不同路径的优劣势及适用条件是什么？\n\n3. 跨境流动性格局演变：稳定币合规化是否可能重塑跨境支付体系，并为离岸人民币使用提供新的通道？这一变化将对银行跨境金融业务带来哪些机遇与潜在冲击？",
    "classification": "Banking-Treasury & ALM",
    "classification_code": "BNK-TSY",
    "report_type": "Assessment of Hong Kong Stablecoin Regulation and Bank Liquidity Impact",
    "report_type_zh": "香港稳定币监管对银行流动性影响评估报告",
    "language": "zh",
    "expert_evaluation_criteria": "## 《稳定币监管化对金融体系与市场结构影响》评价标准\n\n本评价体系用于评估围绕“稳定币监管化”对银行体系与市场结构影响的研究报告质量。重点在于：是否准确理解监管框架、是否量化存款替代效应、是否提出清晰的银行策略路径，并能分析跨境流动性重构逻辑。总分 100 分。\n\n---\n\n### 1. 监管框架与市场机制理解（25 分）\n\n#### （1）香港《稳定币条例》解析（10 分）\n- **9–10 分**：准确拆解牌照、储备资产、托管、合规要求等核心机制，并说明其对市场行为的约束。\n- **6–8 分**：覆盖主要内容，但机制分析不深入。\n- **0–5 分**：仅描述政策，缺乏理解。\n\n#### （2）国际稳定币模式对比（8 分）\n- **7–8 分**：对 USDT、USDC 等模式进行结构化对比（发行机制、储备、合规、应用场景）。\n- **4–6 分**：有对比但不系统。\n- **0–3 分**：缺乏有效对比。\n\n#### （3）运行机制与金融属性（7 分）\n- **6–7 分**：清晰解释稳定币在支付、结算与资产替代中的功能定位。\n- **3–5 分**：有功能描述但逻辑不清。\n- **0–2 分**：理解偏差明显。\n\n---\n\n### 2. 存款替代效应与银行冲击（30 分）\n\n#### （1）替代效应量化分析（12 分）\n- **10–12 分**：基于规模假设（如千亿港元），量化测算对零售存款与企业账户的影响路径与幅度。\n- **7–9 分**：有方向判断但量化不足。\n- **0–6 分**：缺乏实质分析。\n\n#### （2）银行类型差异影响（10 分）\n- **9–10 分**：区分大型银行、中小银行、跨境业务银行的受冲击程度与机制。\n- **6–8 分**：有分类但不深入。\n- **0–5 分**：未区分银行类型。\n\n#### （3）影响机制与边界条件（8 分）\n- **7–8 分**：分析替代发生的条件（收益率、便利性、监管限制等）及其约束。\n- **4–6 分**：有机制但不完整。\n- **0–3 分**：缺乏机制分析。\n\n---\n\n### 3. 银行策略选择与商业模式（25 分）\n\n#### （1）路径对比（发行/托管/观望）（10 分）\n- **9–10 分**：系统比较三种路径的收益、风险、资本占用与监管要求。\n- **6–8 分**：有对比但不系统。\n- **0–5 分**：缺乏清晰路径分析。\n\n#### （2）适用条件与决策逻辑（8 分）\n- **7–8 分**：结合银行禀赋（资本、客户、技术）提出差异化策略。\n- **4–6 分**：有建议但缺乏条件约束。\n- **0–3 分**：策略泛化。\n\n#### （3）商业模式与盈利来源（7 分）\n- **6–7 分**：分析托管、结算、流动性管理等潜在收入来源。\n- **3–5 分**：提及但不深入。\n- **0–2 分**：缺乏商业模式分析。\n\n---\n\n### 4. 跨境流动性与体系重构（20 分）\n\n#### （1）跨境支付体系影响（8 分）\n- **7–8 分**：分析稳定币对传统跨境支付（SWIFT、代理行体系）的冲击路径。\n- **4–6 分**：有判断但不深入。\n- **0–3 分**：缺乏体系分析。\n\n#### （2）离岸人民币与资金流动（6 分）\n- **5–6 分**：评估稳定币对离岸人民币使用与跨境流动的潜在促进作用。\n- **3–4 分**：有提及但逻辑不足。\n- **0–2 分**：缺乏分析。\n\n#### （3）投资结论与风险（6 分）\n- **5–6 分**：结论清晰，识别监管、技术、替代不及预期等关键风险。\n- **3–4 分**：结论一般。\n- **0–2 分**：缺乏有效结论或风险提示。"
  },
  {
    "id": "31",
    "question": "中国新能源汽车行业完成 2025 年关键运行年度后，行业增速、市场结构与竞争格局出现新的变化，政策与出口变量对中期走势的影响更加突出。\n\n请围绕《中国新能源汽车 2025 年复盘与 2026 年趋势展望》这一主题，撰写一份深度研究报告，至少完成以下分析任务：\n\n1. 2025 年行业复盘：基于销量、同比增速与新能源渗透率等核心指标，总结 2025 年行业整体表现，并分析行业增速放缓的主要原因；\n2. 结构性变化：从国内与出口、乘用车与商用车、纯电与插混（含增程）等维度拆解市场结构，分析不同细分赛道的分化趋势，判断该趋势背后的主要驱动因素；\n3. 竞争格局与品牌演化：结合主要品牌销量与市占率变化，分析比亚迪、吉利、新势力品牌，以及华为、小米等跨界参与者在 2025 年的相对表现与竞争逻辑；\n4. 政策、出口与前瞻：评估“双新”补贴、以旧换新等政策效果，分析政策退坡与出口扩张对行业的影响，并结合前述行业总量及结构变化可持续性的分析对 2026 年行业销量、增长动力及主要不确定性作出前瞻判断。",
    "classification": "Capital Markets-Sector & Thematic Research",
    "classification_code": "CAP-STR",
    "report_type": "Annual Review and Outlook Report on China’s New Energy Vehicle Industry",
    "report_type_zh": "中国新能源汽车行业年度复盘与趋势展望研究报告",
    "language": "zh",
    "expert_evaluation_criteria": "## 中国新能源汽车行业年度复盘与趋势展望研究报告评价标准\n\n本评分体系用于量化评估《新能源汽车行业 2025 复盘与 2026 展望》研究报告质量，强调**数据口径一致、因果链条完整、结构拆解到位、竞争逻辑可验证、政策/出口变量可量化、预测可落地**。共 4 个一级维度、12 个二级维度，总分 100 分，适用于人工与 LLM 一致性评分。\n\n### 1. 2025 年行业复盘与增速归因（共 25 分）\n\n#### （1）核心指标覆盖与口径一致性（8 分）\n- 7–8 分：完整覆盖至少“总销量、同比增速、新能源渗透率”且数据准确，并清楚说明统计口径（销量定义、渗透率口径、乘用/商用是否含出口、月度/年度汇总方式等）；跨来源数据能对齐差异并给出解释；图表/表格支撑充分。\n- 4–6 分：指标覆盖基本齐全，但口径说明不充分或跨来源可比性处理不足（如出口、批发/零售、上险口径混用）；仍能支撑大体结论。\n- 0–3 分：关键指标缺失或口径混乱导致结论不可靠；大量“引用数字”无法溯源。\n\n#### （2）增速放缓原因拆解的完整性与定量化（10 分）\n- 9–10 分：对增速放缓给出精确的结构化拆解（至少覆盖需求侧、供给/竞争侧、政策侧、外需/出口侧或基数效应），并尽量定量呈现贡献（如：价格战导致 ASP 下行与需求弹性、渗透率边际放缓、低线/低价位段饱和、PHEV 拉动/BEV 承压等）；明确哪些是短期扰动、哪些是中期趋势。\n- 6–8 分：能提出多个合理原因并形成主次排序，但定量支撑不足或因果链条中存在跳步（“因为竞争激烈所以增速降”缺少机制/证据）。\n- 0–5 分：原因罗列化、常识化，缺少机制解释与证据支撑，或将相关性当因果。\n\n#### （3）关键经营变量联动（价格/利润/补能/产品周期等）（7 分）\n- 6–7 分：能把销量与渗透率变化同“价格带迁移、促销强度、成本变化（电池/原材料）、车企盈利与现金流压力、补能基础设施/充电体验、车型迭代节奏”等变量建立联动解释，并指出对 2026 的延续影响。\n- 3–5 分：提及部分变量，但联动分析浅或缺少证据（只描述不解释“为什么影响销量/结构”）。\n- 0–2 分：几乎不涉及关键经营变量，复盘停留在宏观叙述。\n\n---\n\n### 2. 结构性变化与细分赛道分化（共 25 分）\n\n#### （1）结构拆解维度完备性与粒度（10 分）\n- 9–10 分：按题目要求至少完成三组拆分并准确量化（国内vs出口、乘用vs商用、BEV vs PHEV/EREV），且能进一步下钻到关键子维度（如：出口区域/国家、乘用车价位段与级别、商用车场景如城配/客运/重卡等）；结构变化用份额/增量贡献呈现。\n- 6–8 分：完成基本拆分，但粒度偏粗或缺少“份额变化/增量贡献”的呈现，导致分化结论说服力一般。\n- 0–5 分：拆分不完整或仅定性描述，难以支撑“结构性变化”的判断。\n\n#### （2）分化趋势识别与驱动机制（10 分）\n- 9–10 分：明确指出哪些细分赛道“跑赢/跑输”的趋势（例如 BEV 与插混/增程的相对强弱、出口对总增量的贡献变化、商用电动化拐点特征等），并将驱动机制讲清楚（技术路线、成本/补能、政策、供需错配、产品供给、海外壁垒与合规等），证据链完整。\n- 6–8 分：能识别主要分化趋势并给出解释，但机制不够闭环或缺少关键证据（如只用单一指标支撑复杂结论）。\n- 0–5 分：趋势判断模糊或与数据不一致，驱动解释空泛。\n\n#### （3）结构性变化可持续性的判断及可验证跟踪指标（5 分）\n- 4–5 分：基于上述结构性变化的可持续性判断给出赛道判断（如结构占比、渗透率、出口节奏、BEV/PHEV 变化），并配套可验证的跟踪指标与触发条件（如：价格战强度指标、原材料价格、充电桩利用率、海外关税/认证进度、订单/交付周期等）。\n- 2-3 分：有判断但缺少可验证指标或触发条件，落地性一般。\n- 0–1 分：没有明确可检验的赛道观点，或仅口号式判断。\n\n---\n\n### 3. 竞争格局与品牌演化（共 25 分）\n\n#### （1）品牌销量/市占率变化的定量刻画（10 分）\n- 9–10 分：基于销量与市占率（必要时区分国内/出口、BEV/PHEV、价位段/级别）刻画格局变化；对比至少覆盖：比亚迪、吉利、主流新势力阵营及华为、小米等跨界参与者等，并合理纳入关键对手作为参照（如合资/外资头部、其他自主头部）；使用清晰表格/图形呈现份额迁移。\n- 6–8 分：覆盖主要品牌且有份额对比，但维度较少（只看总盘），或缺少对结构差异的控制（例如把出口与国内混在一起导致误判）。\n- 0–5 分：只做定性评价或数据零散，无法支撑“格局变化”的结论。\n\n#### （2）竞争逻辑解释（成本、技术、产品、渠道、智驾/生态、出海）（10 分）\n- 9–10 分：能把“份额变化”解释到可检验的竞争逻辑：如平台与成本曲线、三电与供应链、自研/采购策略、产品节奏与爆款结构、渠道与交付能力、智驾与座舱能力、品牌心智与定价权、海外布局与合规；并指出各阵营的优势边界与可持续性。\n- 6–8 分：解释方向基本正确，但停留在概念层（例如只说“技术领先/渠道强”），缺少对应证据（成本、毛利、配置、产品周期、销量结构等）。\n- 0–5 分：将竞争结果简单归因于单一因素或主观判断，缺少机制与证据。\n\n#### （3）主要玩家商业模式及行业竞争影响评估（5 分）\n- 5 分：清楚界定主要玩家的商业模式（品牌/渠道/生态赋能、合作模式、产品定位与交付约束等），并评估重点玩家（如华为、小米等跨界参与者）对行业竞争格局影响的冲击路径（价位段挤压、渠道变迁、智驾能力门槛、生态锁定等），结论边界清晰。\n- 3–4 分：有讨论但偏描述，缺少冲击路径或缺少与数据/案例的连接。\n- 0–2 分：未分析或仅停留在舆论层面。\n\n---\n\n### 4. 政策、出口与 2026 前瞻（共 25 分）\n\n#### （1）政策效果评估与退坡影响（10 分）\n- 9–10 分：对“双新补贴/以旧换新”等政策给出效果评估框架（覆盖对象、力度、传导机制、时点与边际变化），并尽可能用数据或可观察指标验证（如订单/上险变化、价位段受益、换购比例、地区差异等）；同时分析退坡/财政约束对 2026 的影响路径。\n- 6–8 分：能说明政策方向与可能影响，但量化不足或缺少“效果验证”的证据。\n- 0–5 分：政策仅做概述或简单转述，未形成影响评估。\n\n#### （2）出口扩张空间与外部约束（8 分）\n- 7–8 分：拆分出口到主要地区/国家或车型结构，分析增长驱动（价格/产品力/渠道/本地化）与约束（关税与反补贴、认证与合规、汇率与物流、海外产能/CKD、政治风险等），并给出对国内供需与竞争的回传影响。\n- 4–6 分：能讨论出口重要性与部分约束，但结构拆分不足或缺少对“回传影响”的分析。\n- 0–3 分：出口仅作为笼统利好/利空描述，缺少约束与路径。\n\n#### （3）2026 销量与增速展望：情景、假设与不确定性（7 分）\n- 6–7 分：给出 2026 年销量/增速/渗透率（至少一个明确数值区间或情景区间），披露关键假设（宏观、价格战强度、政策延续、出口、供给投放等）并做情景或敏感性分析；明确列出 3 项及以上主要不确定性及其影响方向/触发条件。\n- 3–5 分：有预测结论，但缺少关键假设披露或没有情景/敏感性，导致可检验性一般。\n- 0–2 分：预测缺失或仅口号式判断（“持续增长/高景气”），不可验证。"
  },
  {
    "id": "32",
    "question": "在我国房地产市场进入存量主导阶段、住房金融监管政策持续优化的背景下，二手房按揭贷款已成为商业银行零售资产投放与客户获取的重要领域。\n\n请你以工商银行个人金融业务部（二手房按揭方向）负责人的身份，围绕工行在二手房按揭贷款市场中的竞争地位开展系统分析。\n\n在梳理同业主要商业银行二手房贷款业务发展情况的基础上，对工商银行在产品体系、定价能力、渠道与流程效率、风险控制及客户覆盖等方面的竞争优势与短板进行对比评估，并结合房地产市场与住房金融政策环境变化，提出具有针对性和可操作性的业务优化与竞争策略建议。",
    "classification": "Banking-Assets",
    "classification_code": "BNK-AST",
    "report_type": "Competitive Landscape and Operating Strategy Study of Residential Mortgage Finance",
    "report_type_zh": "住房按揭金融业务竞争格局与经营策略研究报告",
    "language": "zh",
    "expert_evaluation_criteria": "## 二手房按揭竞争分析与策略报告评价标准\n\n本评价体系用于对“二手房按揭市场竞争地位分析与策略建议”类报告进行结构化量化评估，强调**同业对标的可比性、诊断结论的可验证性、策略落地的可执行性**以及**风险合规边界**。体系共设 **4 个一级维度、14 个二级维度**，总分 **100 分**，支持人工评审与大模型（LLM）一致性评分。\n\n### 1. 市场与政策环境洞察（共 25 分）\n\n#### （1）存量房市场机制与交易链条理解（8 分）\n- 7–8 分：准确刻画二手房成交驱动（换房链条、改善/刚需占比、挂牌—成交周期、议价率）、交易关键环节（资金监管、网签/过户、抵押登记、评估核验、带押过户等）及其对按揭获客与放款效率的影响；能体现区域分化与城市能级差异。\n- 4–6 分：描述市场进入存量阶段的特征，但对交易链条与对按揭业务的传导影响拆解不够。\n- 0–3 分：仅停留在宏观判断（如“成交下行”“存量为主”），缺少机制解释与业务映射。\n\n#### （2）住房金融监管与政策演变解读（9 分）\n- 8–9 分：覆盖并正确解读利率与首付政策、认房认贷、差别化信贷、资金监管/反洗钱、风险分类与拨备、房地产相关审慎要求等关键政策；明确政策对**准入、定价、额度、流程时效、风险暴露**的约束与机会，并区分全国性规则与地方执行差异。\n- 5–7 分：政策点覆盖较全，但缺少对传导路径（如何影响获客/定价/风控/资本占用）的分析。\n- 0–4 分：政策表述笼统或存在明显误读，无法指导策略设计。\n\n#### （3）市场阶段判断与关键变量/情景（8 分）\n- 7–8 分：给出清晰的阶段判断与时间维度（如未来 6–12 个月），识别关键变量（成交量、房价预期、提前还款、利差、资金成本、区域供需等），并至少提供 2 种情景及对按揭投放/风险的影响。\n- 4–6 分：有趋势判断，但变量与情景不完整或缺少量化指引。\n- 0–3 分：缺乏时间指向与关键变量，结论不可用于决策。\n\n---\n\n### 2. 同业对标与竞争格局分析（共 20 分）\n\n#### （1）同业样本覆盖与分层对标（7 分）\n- 6–7 分：覆盖国有大行、股份行、城农商/区域强行等主要类型，并说明选取逻辑；能按城市能级/区域、渠道结构（中介/直营网点/线上）进行分层对标，避免“用全国均值对比单一城市”的口径偏差。\n- 3–5 分：覆盖主要同业，但分层不足或对标对象单一。\n- 0–2 分：仅零散举例或以主观印象代替对标。\n\n#### （2）对标指标体系的可比性与完整性（7 分）\n- 6–7 分：建立清晰指标口径并覆盖关键维度，例如：产品要素（期限、组合贷/公积金衔接、提前还款规则等）、定价（加点策略、分客群利率、费用）、效率（审批/放款 TAT、材料清单、线上化率）、渠道（中介合作深度、平台导流、网点转化）、风险（逾期/NPL、欺诈拦截、LTV/DTI）、客户（覆盖人群、交叉销售）。\n- 3–5 分：指标覆盖部分关键维度，但口径不清或缺少可比性说明。\n- 0–2 分：没有成体系指标，难以支撑结论。\n\n#### （3）竞争格局结论的清晰度与证据支撑（6 分）\n- 5–6 分：能明确回答“谁在赢、赢在哪里、在哪些城市/客群/渠道赢”，并给出证据（数据、案例、流程对比、政策差异、可核验材料）；结论不自相矛盾。\n- 3–4 分：结论基本成立，但证据不足或与对标指标衔接不紧。\n- 0–2 分：结论泛化（如“同业都在发力”），缺少证据链。\n\n---\n\n### 3. 工行竞争力诊断（产品/定价/渠道流程/风控/客群）（共 35 分）\n\n#### （1）产品体系与客户方案竞争力（7 分）\n- 6–7 分：从二手房业务痛点出发评估工行产品要素（组合贷衔接、带押过户适配、资金监管方案、还款方式、提前还款规则、增值服务等），并与同业形成可对比差异；能提出针对性改造方向。\n- 3–5 分：描述产品较全面，但缺少“痛点—要素—差异—影响”的链条。\n- 0–2 分：仅罗列产品名称或宣传口径。\n\n#### （2）定价能力与授信政策灵活度（7 分）\n- 6–7 分：分析资金成本、FTP/资本占用、风险定价、客群分层定价与区域差异定价能力；识别工行在加点权限、审批授权、白名单、让利策略上的优势与约束，并评估对份额与利润的影响。\n- 3–5 分：提及利率与加点，但缺少机制解释（为何能/不能做、影响多大）。\n- 0–2 分：定价讨论停留在“降息抢客”等口号层面。\n\n#### （3）渠道生态与获客能力（7 分）\n- 6–7 分：对网点、客户经理、房产中介/平台合作、开发商/交易中心合作、公积金中心协同等渠道进行拆解；能量化或半量化评估获客效率（转化率、获客成本、客群质量），并指出工行渠道结构的优势短板与优化抓手。\n- 3–5 分：渠道覆盖较全，但缺乏效率与质量评估。\n- 0–2 分：渠道分析泛泛而谈，缺少二手房获客“关键入口”意识。\n\n#### （4）流程与运营效率/数字化能力（7 分）\n- 6–7 分：对准入、面签、核验、评估、审批、抵押、放款、贷后等全流程识别瓶颈；以可衡量指标描述效率（TAT、补件率、线上化率、一次通过率），并能对比同业“快批快放/一站式”能力差距与系统原因。\n- 3–5 分：提及流程优化，但缺少端到端拆解与量化指标。\n- 0–2 分：仅提出“提升效率、线上化”等笼统表述。\n\n#### （5）风险控制与资产质量管理（7 分）\n- 6–7 分：覆盖信用风险（LTV/DTI、区域房价波动、提前还款）、操作与欺诈风险（假流水、虚假交易、过桥资金、黑中介）、合规风险（资金用途、反洗钱、消费者保护）与集中度管理；能提出可落地的风控策略（准入规则、反欺诈模型/名单、贷后预警、合作方管理）并权衡业务增长。\n- 3–5 分：风险点覆盖主要方面，但缺少机制、工具或落地方法。\n- 0–2 分：风险讨论缺失或仅停留在“加强风控”。\n\n---\n\n### 4. 策略建议可操作性与风险合规（共 20 分）\n\n#### （1）策略组合的针对性、完整性与优先级（8 分）\n- 7–8 分：策略覆盖“产品—定价—渠道—流程—风控—客户经营”并与诊断短板逐一对应；明确优先级与取舍（例如份额 vs 利差、速度 vs 风险），给出阶段性路线图（短/中期）。\n- 4–6 分：有多项建议，但与问题诊断的映射关系不强或缺少优先级。\n- 0–3 分：建议碎片化、通用化，难以执行。\n\n#### （2）落地路径：组织机制、系统改造与合作管理（6 分）\n- 5–6 分：明确牵头部门与分工（总分支协同、审批中心、风控、IT、渠道合作）、授权机制、合作方准入与评价、流程改造需求；能识别关键资源与约束（人员、系统、预算、合规审批周期）。\n- 3–4 分：提出落地想法，但组织与资源安排较粗。\n- 0–2 分：缺乏落地设计，停留在理念层面。\n\n#### （3）量化目标、KPI、测算与风险收益权衡（6 分）\n- 5–6 分：为核心策略设定可衡量目标与KPI（投放规模、份额、TAT、不良/逾期、获客成本、交叉销售、净息差/资本回报等），并给出基础测算或敏感性分析；明确风险边界与止损机制。\n- 3–4 分：有KPI但缺少测算与风险收益权衡。\n- 0–2 分：无量化目标，无法评估效果。"
  },
  {
    "id": "33",
    "question": "国内保险市场竞争持续加剧，监管要求趋严，叠加银行端资本约束、考核机制与风险偏好变化，银保渠道作为银保系保险公司的核心业务来源，其盈利能力与可持续经营模式正面临系统性重塑。\n\n请围绕国内银保系保险公司的经营实践，在以国有大型商业银行为核心合作对象的银保渠道场景下，撰写一份《银保渠道经营盈利能力规划报告》。报告应结合银保渠道客户结构特征、银行端经营约束以及当前市场环境，对渠道盈利能力现状进行系统分析，重点梳理收入结构、成本构成及影响盈利水平的关键因素。\n\n同时，围绕长期保障类产品与年金类产品构成的核心业务结构，提出提升银保渠道盈利质量与可持续经营能力的中长期规划思路，明确在产品结构优化、渠道协同与激励机制设计、经营效率提升以及风险控制等方面的重点举措与实施路径，为银保系保险公司银保渠道的经营管理提供具备可落地性的决策参考。",
    "classification": "Insurance-Life Insurance",
    "classification_code": "INS-LIF",
    "report_type": "Bancassurance Channel Profitability Planning Report",
    "report_type_zh": "银保渠道盈利能力规划报告",
    "language": "zh",
    "expert_evaluation_criteria": "## 银保系保险公司银保渠道经营盈利能力规划报告评价标准\n\n本评价体系适用于评估国内银保系寿险/人身险公司的银保渠道盈利能力规划类研究与经营方案报告。体系共设4个一级维度、13个二级维度，总分100分；支持人工评审与LLM一致性评分。\n\n---\n\n## 1. 经营现状诊断与关键洞察（30分）\n\n### （1）银保渠道客户结构与经营场景刻画（10分）\n- 9–10分：清晰刻画银保客户画像与分层（如高净值/大众富裕/养老客群、代际、风险偏好、资金属性、期限偏好、保障缺口），并与国有大行客群与网点经营特征相匹配；能落到可执行标签/客群包（例如“养老金客群-年金转化”“存款搬家客群-保障补位”），区分新单/存量、趸交/期交、线上/线下与分行差异。\n- 6–8分：画像与分层基本合理，能覆盖主要客群与场景，但标签化、分行差异或与银行端经营机制的映射不够具体。\n- 0–5分：客群描述泛化，缺乏与银保渠道真实触点（网点、理财经理/客户经理、私行）的对应关系，难以支持经营动作设计。\n\n### （2）收入结构拆解与利源归因（10分）\n- 9–10分：对银保渠道收入进行结构化拆解并可量化归因（至少覆盖：新单价值/规模、续期与存量贡献、手续费/佣金结构、投资/利差相关影响、产品定价与费用假设对价值的作用）；能解释“保障 vs 年金”“趸交 vs 期交”“不同合作行/分行/客群”对VNB/利润的差异，并识别主要驱动因子与可控/不可控边界。\n- 6–8分：能拆解主要收入项并指出驱动因素，但定量深度不足，或未把利源与产品结构/银行端约束联动起来。\n- 0–5分：仅停留在规模或保费口径，缺少利源逻辑（价值、费用、风险）与归因分析。\n\n### （3）成本构成、资本/风险成本与盈利瓶颈定位（10分）\n- 9–10分：系统拆解成本（渠道佣金及其节奏、运营与服务成本、获客与培训成本、系统与合规成本等），并纳入资本占用/偿付能力约束、风险成本（退保、保证利率、久期错配等）对利润与ROE的影响；能明确指出当前“盈利瓶颈=结构性问题”还是“效率/激励问题”，并给出可验证证据。\n- 6–8分：覆盖主要成本项与部分风险，但对资本占用、风险成本或费用节奏（如前置佣金）分析不充分。\n- 0–5分：成本分析碎片化或仅罗列，未定位关键矛盾与可控抓手。\n\n---\n\n## 2. 盈利测算框架与方法可验证性（25分）\n\n### （1）指标体系与口径一致性（8分）\n- 7–8分：建立适用于银保渠道的指标体系并定义口径清晰（如FYP/APE、VNB/NBV与价值率、综合费用率/新单成本、13/25/37月继续率、退保率、保单边际贡献、资本占用与ROE等）；明确分产品、分合作行、分分行/网点、分期限与缴费方式的切片口径。\n- 4–6分：指标较完整但关键口径未统一（如利润/VNB口径混用、分摊规则不清），或缺少分层切片。\n- 0–3分：指标选择随意或仅停留在保费规模，无法支撑盈利质量评价。\n\n### （2）测算模型、数据需求与可复现性（9分）\n- 8–9分：披露测算框架（利润表/价值模型/单位经济模型）与关键参数来源，说明数据需求与可获取性（公司内部、银行侧可提供、第三方/行业口径）；给出计算步骤或示例，使第三方可复核主要结论。\n- 4–7分：有测算思路但关键步骤、分摊规则或参数来源描述不足，可复现性一般。\n- 0–3分：以结论替代测算，缺少模型与参数说明，无法验证。\n\n### （3）关键假设、敏感性与情景分析（8分）\n- 7–8分：明确列出并量化关键假设（至少覆盖利率/投资收益、退保与继续率、佣金/手续费政策、产品定价要素、银行端考核与转化效率等），并做敏感性或情景分析（如利率下行、监管趋严、银行资本更紧、同业费率战加剧）说明盈利与偿付能力影响路径。\n- 4–6分：有假设但量化不足，或仅做单一情景，未覆盖银行端约束与监管变量。\n- 0–3分：假设隐含且不可检验，无敏感性分析。\n\n---\n\n## 3. 中长期规划与可落地举措设计（30分）\n\n### （1）核心业务结构：长期保障类与年金类的产品策略与组合优化（10分）\n- 9–10分：围绕“保障+年金”给出清晰的组合目标与迁移路径（例如从高趸交、低价值向期交与保障型提升；或年金与保障的交叉销售策略），并说明对VNB、资本占用、持续率与风险的影响；能提出产品层面的可执行动作（期限/缴费期设计、利益演示与适当性、附加服务、定价与费用结构优化、停售/迭代机制）。\n- 6–8分：提出方向性组合优化，但缺少对价值与风险的量化影响评估，或缺少产品落地动作。\n- 0–5分：产品策略空泛，未回应“盈利质量与可持续”的核心矛盾（如只讲规模增长）。\n\n### （2）渠道协同与激励机制设计（银行端约束匹配）（10分）\n- 9–10分：能把国有大行的组织与考核现实纳入方案（总分行条线、网点KPI、理财/个金/私行协同、合规与声誉约束、资本与风险偏好），提出可落地的协同机制与激励方案（佣金与费用节奏、分润与资源投入、名单制/客群包、联合经营与培训、过程KPI与结果KPI组合），并说明对转化率、继续率、投诉与合规的影响。\n- 6–8分：提出协同与激励思路，但缺少对银行端考核/流程的嵌入式设计，或缺少可执行的KPI与责任分工。\n- 0–5分：忽视银行端约束，把银保当作单边推动，缺乏可执行机制。\n\n### （3）经营效率提升与运营能力建设（5分）\n- 5分：提出可执行的效率提升抓手（投保/核保/回访/保全/理赔/续期经营、队伍与网点赋能、数字化工具与流程再造），并量化目标（如人均产能、件均成本、时效、线上化率、续期触达率）。\n- 2–4分：覆盖主要流程，但缺少量化目标或工具/流程与组织能力匹配不足。\n- 0–1分：仅口号式描述“提升效率/数字化”，无具体动作。\n\n### （4）路线图、资源与KPI闭环（5分）\n- 5分：给出中长期分阶段路线图（如0–6月/6–18月/18–36月），明确里程碑、资源投入（系统、人员、预算、培训）、治理机制与KPI看板（领先指标+滞后指标），并说明复盘与迭代机制。\n- 2–4分：有时间规划但里程碑、资源或KPI闭环不完整。\n- 0–1分：无明确落地节奏与管理闭环。\n\n---\n\n## 4. 风险合规与可持续经营（15分）\n\n### （1）监管与销售合规适配（6分）\n- 6分：明确识别并嵌入银保渠道关键合规要求（适当性管理、销售误导与双录/回访、费用与手续费合规、反洗钱、信息披露、消费者保护等），并说明其对产品设计、激励机制、过程管理与数据使用的约束与改造点。\n- 3–5分：提及主要合规点，但未落到流程与制度设计，或缺少对激励/费用结构的约束分析。\n- 0–2分：合规风险缺失或流于形式。\n\n### （2）资产负债管理与利率/久期风险（5分）\n- 5分：能把利率环境、保证利率/负债成本、久期错配、再投资风险、退保行为等纳入产品结构与渠道节奏规划，并给出控制抓手（产品条款与定价、资产配置约束、负债结构优化、压力测试触发阈值）。\n- 2–4分：识别部分风险但缺少量化或缺少与产品/渠道动作的联动。\n- 0–1分：未考虑ALM与利率风险对银保年金/长期产品的系统性影响。\n\n### （3）声誉、操作与合作方风险管理（4分）\n- 4分：覆盖投诉与声誉风险、合作银行操作风险、数据与系统安全、第三方/外包管理等，并给出可执行的监控指标与应对预案（如投诉率阈值、回访质检、合规抽检、分行白名单/黑名单机制）。\n- 2–3分：有风险提示但缺少监控指标或预案。\n- 0–1分：缺少对银保渠道特有的声誉与合作风险管理。"
  },
  {
    "id": "34",
    "question": "近期宏观数据显示核心CPI增速出现持续回升的迹象，引发市场对通胀预期和债市走向的关注。假设你是一位头部券商固定收益部的宏观分析师，请你撰写一份固收点评报告，深入探讨这一现象背后的经济逻辑。\n\n报告需包含以下关键内容：\n\n1. 数据回顾与结构拆解： 回顾近期CPI与PPI数据的最新变化特征，系统梳理其同比与环比走势，分析价格变动的表层含义。在此基础上，进一步拆解推动核心CPI回升的主要驱动力，重点比较服务型消费（如旅游、住宿等）与实物消费在价格表现上的差异，并结合季节性因素判断本轮回升的持续性；\n2. 宏观经济信号解读： 探讨核心CPI回升是否意味着国内有效需求出现实质性改善，分析其与PPI（工业生产者出厂价格指数）走势之间的联动关系，并重点评估“CPI-PPI剪刀差”变化对企业盈利修复及产业链传导的潜在影响；\n3. 货币政策与债市策略： 在通胀温和回升的基准情景下，评估其对央行货币政策宽松空间（尤其是降息节奏）的约束程度。在此基础上，对债券市场短期走势进行判断，并提出具有可操作性的投资策略建议（如长端利率债的配置价值与期限选择）。",
    "classification": "Capital Markets-Macro & Strategy Research",
    "classification_code": "CAP-MAC",
    "report_type": "Fixed Income Commentary / Macroeconomic Analysis Report",
    "report_type_zh": "固定收益点评/宏观经济分析报告",
    "language": "zh",
    "expert_evaluation_criteria": "## 固收宏观点评报告评价标准\n\n本评价体系用于对“核心CPI回升”主题的券商固收宏观点评报告进行量化评估，强调**数据归因的可复核性、宏观逻辑的严谨性、政策推演的贴近央行框架、债市策略的可交易性**。体系共设 4 个一级维度、12 个二级维度，总分 100 分，适用于人工评审与LLM自动评分的一致性应用。\n\n---\n\n## 1. 数据归因与结构拆解质量（30 分）\n### （1）核心CPI口径与数据处理规范（6 分）\n- 6 分：清晰说明核心CPI口径（剔除食品、能源等）、同比/环比（必要时区分季调/非季调）、基数效应；对“回升迹象”给出明确时间段与数值变化，并指出统计噪声来源（节假日错位、价格管制/补贴、样本替换等）。\n- 3–5 分：能交代主要口径与同比回升事实，但对环比、季调或基数效应说明不足。\n- 0–2 分：口径含混、仅泛谈“回升”，缺少可核对的数值与期间界定。\n\n### （2）分项归因与贡献测算：服务 vs 实物（12 分）\n- 11–12 分：对核心CPI做结构拆解，至少覆盖“服务型消费（旅游/住宿/交通服务等）”与“实物消费（耐用品/日用品等）”的价格表现差异；给出**贡献度/拉动**（可用近似贡献、权重×涨幅、或分项加总解释），识别 2–3 个关键分项并解释驱动（需求、供给、成本、竞争格局、监管定价等）。\n- 7–10 分：能比较服务与实物方向差异，指出主要分项，但贡献测算或驱动解释偏定性、证据不足。\n- 0–6 分：仅罗列分项涨跌或套用结论，未回答“谁在拉动核心CPI回升”。\n\n### （3）季节性、节假日与可持续性判断（12 分）\n- 11–12 分：显式处理季节性（旅游旺季、春节/五一/暑期等）与节假日错位影响，结合环比/季调或历史同期对比判断；对可持续性给出**条件化结论**（例如“若服务需求韧性+工资/收入改善，则回升更可持续；若仅节假日/供给扰动，则回落概率高”），并提出可跟踪指标（出行高频、餐饮/酒店价格、租金、教育服务等）。\n- 7–10 分：能提到季节性并做初步判断，但缺少量化对照或跟踪指标，结论偏一刀切。\n- 0–6 分：忽略季节性/基数效应，或直接下“通胀拐点已至”等不可验证结论。\n\n---\n\n## 2. 宏观经济信号解读与论证严谨性（25 分）\n### （1）有效需求改善的判定框架（10 分）\n- 9–10 分：不把核心CPI回升简单等同需求修复，而是用可解释框架（产出缺口、就业与工资、居民收入预期、服务消费恢复、信用扩张/财政脉冲等）进行交叉验证；明确区分“价格回升来自需求”还是“来自供给/成本/结构性因素”。\n- 6–8 分：能讨论需求与价格关系，但验证链条不完整，证据更多停留在经验判断。\n- 0–5 分：用单一指标或口号式表述替代论证（如“核心CPI回升=需求好转”）。\n\n### （2）与PPI联动、剪刀差与利润修复机制（10 分）\n- 9–10 分：清晰讨论CPI与PPI可能分化/收敛的原因（上游成本、产能利用率、定价权、地产链拖累、出口与输入性因素等）；解释“CPI-PPI剪刀差”变化对企业利润的方向性影响与传导路径（收入端/成本端/库存周期），并指出受益/受损行业特征。\n- 6–8 分：能提到剪刀差与利润关系，但机制解释较粗，缺少行业/链条层面的展开。\n- 0–5 分：仅机械描述“剪刀差扩大/收敛”，未解释盈利含义与条件。\n\n### （3）关键假设、情景推演与反证意识（5 分）\n- 5 分：至少列出 2–3 个关键假设（例如服务价格韧性、油价与输入性通胀、汇率、需求修复斜率、政策力度），并给出上/下行情景或敏感性判断；能提供反证路径（哪些数据走弱会推翻结论）。\n- 3–4 分：有假设但不成体系，或缺少反证与触发条件。\n- 0–2 分：无假设、无情景，结论不可检验。\n\n---\n\n## 3. 货币政策含义与债市策略（30 分）\n### （1）央行反应函数与降息节奏约束评估（12 分）\n- 11–12 分：围绕“通胀温和回升”评估对宽松空间的约束，能把通胀与**稳增长、汇率/资本流动、金融稳定、地产与信用修复**等约束放在同一反应函数框架下；对降息节奏给出条件化判断（“何种通胀/汇率/信用组合下更可能降息或降准”）。\n- 7–10 分：能讨论通胀对宽松的影响，但对央行多目标权衡（汇率、稳增长、金融稳定）展开不足。\n- 0–6 分：把政策判断简化为“通胀上行=不能降息/必然收紧”等线性结论。\n\n### （2）债市定价逻辑：利率、曲线与期限结构推演（10 分）\n- 9–10 分：将通胀变化映射到债市定价（实际利率、期限溢价、通胀预期、资金面预期），给出短期走势判断并说明作用渠道；能讨论曲线形态（陡峭/平坦）、长短端不同驱动与可能分化。\n- 6–8 分：有方向判断，但缺少定价变量拆解或曲线层面推演。\n- 0–5 分：只给“看多/看空”结论，缺少机制与变量对应。\n\n### （3）投资策略可执行性：品种选择与风控（8 分）\n- 7–8 分：给出具体策略建议（如长端利率债配置性价比、久期建议、骑乘/杠杆/曲线交易、与同业存单/信用债的相对价值比较），并包含至少 1–2 个可执行要素（入场条件、止损/止盈、对冲工具如IRS、关键数据/会议触发点）。\n- 4–6 分：策略方向明确但偏“观点型”，缺少仓位/触发/风控细节。\n- 0–3 分：策略泛化（如“关注机会、谨慎配置”），不可交易。\n\n---\n\n## 4. 研究方法规范、风险提示与表达（15 分）\n### （1）数据来源权威性、时效性与引用规范（5 分）\n- 5 分：数据来源清晰（统计局、央行、行业高频/第三方等），时间戳明确；图表/引用规范，关键结论可追溯到对应数据。\n- 3–4 分：来源基本可信但引用不够规范或缺少时间戳。\n- 0–2 分：来源不明、引用混乱，影响可信度。\n\n### （2）风险提示的针对性与覆盖度（5 分）\n- 5 分：风险与正文观点形成制衡，至少覆盖：输入性通胀（油价/汇率）、政策超预期（宽松/收紧）、需求再度走弱、供给扰动、地产与信用链条变化等，并说明对利率与曲线的影响方向。\n- 3–4 分：列出主要风险但缺少影响机制或触发条件。\n- 0–2 分：风险提示形式化或缺失。\n\n### （3）报告结构与专业表达（5 分）\n- 5 分：结构符合卖方点评习惯（摘要/核心结论/数据拆解/逻辑推导/策略与风险），语言简洁、结论前置；关键变量、图表与小标题帮助读者快速定位。\n- 3–4 分：可读性尚可，但结构松散或结论不够前置。\n- 0–2 分：表达冗长含混，难以提炼交易含义。"
  },
  {
    "id": "35",
    "question": "请参照中金公司等头部机构的宏观研究方法，撰写一份系统性的《美国通胀跟踪与预测框架》研究报告，重点涵盖以下内容：\n\n1. 通胀指标体系构建：拆解美国 CPI 与 PCE 在权重结构上的主要差异，并说明在构建预测模型时，应如何对通胀篮子进行分类（如能源、食品、核心商品、核心服务），总结不同类型商品价格的特点（季节性、周期、波动、趋势等）；\n2. 关键分项的领先指标挖掘：围绕核心商品（如汽车）与核心服务（尤其是住房租金 / OER），识别最具预测意义的高频领先指标（如 Manheim 指数、Zillow 租金指数等），并解释其向 CPI / PCE 传导的时间滞后关系；\n3. 综合预测逻辑：在上述分项分析基础上，总结一套“自下而上”的通胀预测方法论，说明如何通过持续跟踪细分项高频数据，前瞻性识别整体通胀的趋势变化与潜在拐点。",
    "classification": "Capital Markets-Macro & Strategy Research",
    "classification_code": "CAP-MAC",
    "report_type": "Thematic Research Report on US Inflation Tracking System and Forecasting Framework",
    "report_type_zh": "美国通胀跟踪体系与预测框架专题研究报告",
    "language": "zh",
    "expert_evaluation_criteria": "## 美国通胀跟踪与预测研究报告评价标准\n\n本评价体系适用于评估以美国通胀跟踪与预测框架为主题的研究报告质量，采用结构化、可执行、可量化、可复核的评价方式，兼容人工评审与大模型（LLM）自动评分的一致性应用。评分重点关注指标体系拆解的准确性、关键分项领先指标的有效性、分项预测到总体通胀的聚合方法可复用性，以及回测验证与实务可操作性。体系共设4个一级维度、12个二级维度，总分100分。\n\n---\n\n### 1. 指标体系构建与口径拆解（共30分）\n\n#### (1) CPI 与 PCE 统计口径差异拆解（10分）\n- 9–10分：准确拆解 CPI 与 PCE 在权重来源、覆盖范围、住房权重与修订机制上的核心差异，并说明其对预测建模与传导强度的影响。\n- 6–8分：说明主要差异，但对预测影响或模型处理解释不足。\n- 0–5分：差异表述笼统或存在关键事实错误。\n\n#### (2) 通胀篮子分类方案的合理性与可映射性（10分）\n- 9–10分：构建可落地的分项分类体系（能源、食品、核心商品、核心服务），并明确官方分项到模型分组的映射规则，解释分类目的与驱动差异。\n- 6–8分：分类基本合理，但映射规则不清或颗粒度不足。\n- 0–5分：分类随意或无法支撑预测框架。\n\n#### (3) 数据处理与可比性设计（10分）\n- 9–10分：明确频率与指标变换方式（MoM、YoY、年化指标），说明季调、基数效应与数据修订对预测的影响，并区分实时数据与修订数据口径。\n- 6–8分：具备基础处理说明，但对修订或异常处理考虑不足。\n- 0–5分：缺乏数据处理说明，结论不可比。\n\n---\n\n### 2. 关键分项领先指标与传导滞后机制（共30分）\n\n#### (1) 核心商品领先指标选择与解释（10分）\n- 9–10分：构建机制清晰的领先指标体系，并明确其对应通胀分项及结构变化对指标有效性的影响。\n- 6–8分：列出指标但缺少分项映射或机制解释。\n- 0–5分：指标选择与核心商品关系弱或缺乏解释。\n\n#### (2) 核心服务与住房通胀领先指标（10分）\n- 9–10分：清晰区分新签租金与存量租约传导路径，合理使用市场租金指标，并扩展至非住房服务的成本或工资驱动变量。\n- 6–8分：能说明指标关系，但对传导机制或口径映射不足。\n- 0–5分：领先指标或传导解释明显不正确。\n\n#### (3) 滞后结构与传导模型清晰度（10分）\n- 9–10分：对领先指标与通胀分项之间的时间滞后给出可检验设定，并解释不同分项滞后差异及预测应用。\n- 6–8分：具备滞后判断，但缺少可操作建模表达。\n- 0–5分：仅定性描述滞后关系，无法落地。\n\n---\n\n### 3. 自下而上预测框架与可验证性（共25分）\n\n#### (1) 分项预测模型设计与总体聚合（10分）\n- 9–10分：形成完整分项预测与权重聚合流程，明确模型与数据输入对应关系，并可输出持续更新的预测框架。\n- 6–8分：具备自下而上思路，但聚合与权重处理不严密。\n- 0–5分：停留在概念层，缺少可执行路径。\n\n#### (2) 持续跟踪机制与拐点识别（8分）\n- 7–8分：建立清晰更新节奏与信号判断规则，能够区分短期噪声与趋势变化。\n- 4–6分：提及高频跟踪，但缺少明确机制。\n- 0–3分：缺乏持续跟踪设计。\n\n#### (3) 回测验证、误差归因与稳健性（7分）\n- 6–7分：提供可复现验证方法并进行误差归因，讨论结构性变化下的稳健性处理。\n- 3–5分：存在验证但缺少归因或样本外检验。\n- 0–2分：无验证或可靠性评估。\n\n---\n\n### 4. 研究表达、实用价值与风险情景（共15分）\n\n#### (1) 结论前瞻性与机构化表达（6分）\n- 5–6分：结构清晰，结论可直接用于交易或政策研判，明确通胀驱动来源与未来主导变量。\n- 3–4分：结构完整但结论偏描述性。\n- 0–2分：结构混乱或结论不清。\n\n#### (2) 风险情景与政策含义（5分）\n- 4–5分：给出关键风险情景并说明对通胀与政策路径的影响。\n- 2–3分：风险提示存在但缺少机制分析。\n- 0–1分：缺乏有效风险情景。\n\n#### (3) 可复用交付物与引用规范（4分）\n- 4分：提供可复用指标清单、数据来源与更新规则，引用规范且可追溯。\n- 2–3分：部分可复用内容，但复现性一般。\n- 0–1分：缺少附录与来源说明。"
  },
  {
    "id": "36",
    "question": "随着科技创新型企业融资需求不断上升，以核心技术与专利资产为基础的专利质押融资模式，正逐步成为支持高新技术企业发展的重要金融工具。然而，高技术专利在估值方法、变现能力以及风险可控性方面仍存在显著不确定性，对金融机构的风险管理能力与授信结构设计提出了更高要求。\n\n请以高新技术企业“技术专利质押融资”模式为研究主题，选取寒武纪、中微公司等具有代表性的科技企业作为案例，围绕其核心技术与专利资产开展系统分析，重点完成以下研究任务：\n\n1. 分析高技术专利在不同发展阶段下的估值方法与模型适用性，包括收益法、市场法及期权化估值等，并比较其在实际授信场景中的优劣；\n2. 评估专利质押率、专利资产变现能力及其不确定性来源，分析技术替代风险、法律保护强度及市场流动性等因素对融资安全性的影响；\n3. 系统梳理专利质押融资面临的主要风险类型，并评估常见风险缓释机制（如保证金安排、政府或第三方增信、分阶段放款等）的有效性；\n4. 结合高新技术企业研发投入强度与技术迭代节奏，探索差异化授信方案设计路径，例如基于研发投入或专利质量的动态授信额度安排，或通过与知识产权交易所、专业服务机构合作，构建专利质押物的登记、流转与处置通道，以提升该类融资模式的可持续性与风险可控性。",
    "classification": "Banking-Assets",
    "classification_code": "BNK-AST",
    "report_type": "Patent-Backed Financing Assessment and Credit Structuring Report for High-Tech Enterprises",
    "report_type_zh": "高新技术企业专利质押融资评估与授信方案研究报告",
    "language": "zh",
    "expert_evaluation_criteria": "## 技术专利质押融资研究报告评价标准\n\n本评价体系用于对“高新技术企业技术专利质押融资”主题的深度研究报告进行可执行的量化评估，兼容人工评审与大模型（LLM）自动评分。评分重点关注：**专利资产与技术竞争力洞察、估值/质押定价的模型适配与可验证性、风险识别与缓释机制的有效性、授信结构设计的可落地性与合规性**。体系共设 4 个一级维度、15 个二级维度，总分 100 分。\n\n### 1. 案例与专利资产洞察质量（共 25 分）\n\n#### （1）案例选择代表性与对比设计（7 分）\n- 6–7 分：选择寒武纪、中微公司等**具备明显技术路线差异/商业化阶段差异**的代表性企业；对比维度清晰（技术壁垒、收入结构、研发强度、专利组合质量、产业链位置、客户集中度等），能支撑“阶段—估值—授信结构”的讨论。\n- 3–5 分：案例合理但对比维度单一，或仅做公司简介式描述，与质押融资问题结合不紧。\n- 0–2 分：案例随意或与专利质押融资适配性弱，缺乏对比逻辑。\n\n#### （2）核心技术与专利组合拆解深度（10 分）\n- 9–10 分：识别企业**核心技术模块/关键工艺/关键 IP**，将专利按技术族/产品线/应用场景拆分；覆盖专利数量与结构（发明/实用新型、国内/海外、同族、剩余期限、权利要求范围）、质量指标（被引、同族规模、维持年限、无效/诉讼记录、标准必要性潜力等）；明确“哪些专利可质押、为何可质押”。\n- 6–8 分：能描述核心技术与主要专利，但专利质量、法律状态或可质押边界分析不够细。\n- 0–5 分：停留在“有很多专利/技术领先”的口号层，缺乏可核查的专利资产拆解。\n\n#### （3）技术竞争格局与替代路径分析（8 分）\n- 7–8 分：从技术路线、产品迭代节奏、竞品与替代技术（国产/海外、不同架构/工艺路线）出发，解释对专利现金流/许可空间/处置价值的影响；能把“替代风险”量化为情景假设或风险折价依据。\n- 4–6 分：有竞争格局描述，但与专利价值和授信安全性传导不充分。\n- 0–3 分：几乎不讨论技术替代与竞争格局，或仅做主观判断。\n\n---\n\n### 2. 估值方法与授信定价适配性（共 30 分）\n\n#### （1）发展阶段划分与估值方法匹配（8 分）\n- 7–8 分：清晰划分专利/技术的生命周期阶段（研发期、导入期、放量期、成熟期/衰退期等），并解释各阶段现金流可得性、可比交易可得性、波动性来源，从而匹配收益法/市场法/期权化方法的适用区间与局限。\n- 4–6 分：提及阶段差异，但方法匹配理由不够严格或缺少边界条件。\n- 0–3 分：不区分阶段，或方法选择随意、套模板。\n\n#### （2）收益法建模质量（现金流/特许权使用费/超额收益）（10 分）\n- 9–10 分：给出可复核的建模路径（如 RFR/Relief-from-Royalty、超额收益法、增量利润法），明确关键输入（可许可收入基数、合理费率区间、渗透率、毛利/费用、折现率、经济寿命、税率等）及来源；进行敏感性/情景分析并说明对授信折扣与质押率的影响。\n- 6–8 分：模型框架基本正确，但关键参数依据薄弱、敏感性不足或与授信指标联动不清。\n- 0–5 分：仅概念描述，无可计算过程或关键假设缺失。\n\n#### （3）市场法与可比交易选择合理性（6 分）\n- 5–6 分：能找到并筛选可比（专利许可、专利/技术转让、并购交易中的 IP 定价、同类技术授权费率等），说明可比性调整逻辑（地域、剩余期限、权利范围、排他性、诉讼风险、产业周期等），避免“硬套倍数”。\n- 3–4 分：引用可比但可比性解释不足，或缺少调整过程。\n- 0–2 分：无可比依据或可比明显不相关。\n\n#### （4）期权化/实物期权估值的使用边界与落地性（6 分）\n- 5–6 分：说明为何高技术专利具有期权特征（不确定性、可延后投资、分阶段决策），并给出可落地的期权化路径（如二叉树/蒙特卡洛/决策树，参数含义与取值来源、里程碑触发）；强调该方法在授信中的作用更偏向“风险折价与情景权重”而非单一点估值。\n- 3–4 分：提及期权估值但停留在概念层，参数与授信应用缺失。\n- 0–2 分：不理解期权化估值含义，或生搬硬套公式。\n\n#### （5）估值结果向授信定价/质押率的映射逻辑（LTV/折扣/保证金）（必选）（6 分）\n\n---\n\n### 3. 质押物变现能力、风险识别与缓释机制（共 30 分）\n\n#### （1）变现路径与市场流动性评估（8 分）\n- 7–8 分：系统梳理处置路径（协议转让/拍卖、许可变现、与产业方/PE 的定向交易、知识产权交易所挂牌等），评估流动性约束（买方数量、定价透明度、交易周期、信息不对称、处置费用）；给出可操作的处置预案与时间预期。\n- 4–6 分：描述处置方式但缺少对交易可达性与周期的量化/证据。\n- 0–3 分：默认“可卖出”但无路径、无约束分析。\n\n#### （2）不确定性来源拆解与对融资安全性的传导（8 分）\n- 7–8 分：将不确定性拆为可操作因子：技术替代、专利稳定性（无效/侵权诉讼风险）、权属与共有/质押重复、地域保护差异、剩余期限、标准/生态变化、客户集中导致的现金流脆弱性等，并说明各因子如何影响折扣率、回收率、处置时间与契约条款。\n- 4–6 分：识别部分因素但传导机制不清，偏“罗列风险点”。\n- 0–3 分：风险点泛化或与专利质押无关。\n\n#### （3）主要风险类型全景覆盖与分层（7 分）\n- 6–7 分：至少覆盖并区分：估值风险、法律与权属风险、技术与商业化风险、操作与道德风险（信息披露/专利维护费/重复质押）、集中度与关联交易风险、政策与合规风险、处置执行风险；形成“发生概率—影响程度—可监控指标”的分层。\n- 3–5 分：覆盖面尚可，但分层与指标化不足。\n- 0–2 分：风险分类缺失或明显遗漏关键风险。\n\n#### （4）风险缓释机制有效性评估（保证金/增信/分阶段放款等）（7 分）\n- 6–7 分：不止列举措施，能评价机制有效性与适用边界：保证金/风险准备金如何覆盖损失；政府/担保/保险/第三方增信的代偿条件与道德风险；分阶段放款与研发/商业化里程碑如何设定；契约条款（信息披露、专利维持、禁止再质押、交叉违约、现金流归集）如何降低损失率与处置难度。\n- 3–5 分：提到常见工具，但缺少触发条件、约束条款或有效性比较。\n- 0–2 分：缓释措施空泛或与高技术专利风险不匹配。\n\n---\n\n### 4. 授信方案设计、运营落地与合规（共 15 分）\n\n#### （1）差异化授信结构与动态额度设计（7 分）\n- 6–7 分：结合研发投入强度、技术迭代节奏与专利质量指标，提出动态授信方案（额度随里程碑/专利质量/收入验证调整；期限与宽限期匹配研发周期；分层定价与风险加点逻辑）；明确关键触发指标与监控频率。\n- 3–5 分：提出思路但缺少可执行条款（指标、阈值、触发动作）。\n- 0–2 分：授信方案泛化为“提高额度/降低利率”，缺乏结构设计。\n\n#### （2）与知识产权交易所/专业机构协同的处置与运营机制（5 分）\n- 4–5 分：给出可落地的协同机制：质押登记、评估复核、持续监测（专利法律状态/年费/诉讼/被引）、挂牌与撮合、处置执行；明确银行—交易所—评估/律所—产业方的分工与流程节点。\n- 2–3 分：提及合作但流程与角色不清。\n- 0–1 分：不讨论落地通道，或将其视为“外包即可解决”。\n\n#### （3）合规与信息披露质量（3 分）\n- 3 分：明确遵循专利质押登记、公示、权属核查、关联交易与授信合规要求；对数据来源、估值假设、利益冲突与局限性有清晰披露，避免夸大确定性。\n- 1–2 分：有合规意识但披露不完整。\n- 0 分：缺乏合规边界或存在明显误导性表述。"
  },
  {
    "id": "37",
    "question": "你是一位券商宏观研究团队的负责人，需要撰写一份美国宏观经济年度展望报告，对 2026 年美国经济增长、通胀、政策环境及潜在风险进行系统研判，为全球资产配置与投资决策提供依据。\n\n请围绕“美国经济不均衡的再加速”这一主线，系统分析以下问题：\n\n1. 回顾 2025 年美国经济运行特征，重点分析AI 相关投资与传统经济部门分化的成因，及其所形成的“双速经济”格局对增长和金融条件的影响；\n2. 对 2026 年美国经济增长、通胀与内需结构进行前瞻分析，重点评估 AI 投资、居民消费、企业投资与地产投资对经济的支撑力度及可持续性；\n3. 结合特朗普政府政策取向与美联储政策框架变化，分析 2026 年财政政策、货币政策及流动性环境的主要变量，并讨论其对利率、美元及金融市场的潜在影响；\n4. 从中长期视角，识别 2026 年美国经济面临的主要不确定性与风险因素，包括通胀黏性、财政可持续性、收入分配不均衡及其对宏观稳定与资产价格的影响。\n\n请确保报告逻辑清晰、论证严谨，并体现对宏观经济、政策传导机制与金融市场联动关系的深入理解。",
    "classification": "Capital Markets-Macro & Strategy Research",
    "classification_code": "CAP-MAC",
    "report_type": "U.S. Macroeconomic Outlook and Policy Scenario Analysis Report",
    "report_type_zh": "美国宏观经济年度展望与政策情景分析报告",
    "language": "zh",
    "expert_evaluation_criteria": "## 美国宏观经济年度展望报告评价标准\n\n本评价体系用于对《2026 年美国宏观经济年度展望报告》（主线：美国经济不均衡的再加速）进行结构化、可执行的量化评估。体系共设 4 个一级维度、13 个二级维度，总分 100 分，兼容人工评审与大模型（LLM）自动评分。评分重点覆盖：主线洞察与框架、2026 预测的假设与拆解、政策与金融市场传导、风险与资产配置可操作性。\n\n---\n\n### 1. 主线框架与2025复盘质量（共 25 分）\n\n#### (1) 主线一致性与“双速经济”刻画（10 分）\n- 9-10 分：围绕“非均衡再加速”建立明确框架，清晰刻画 AI 投资部门与传统部门在增长、利润、融资条件、就业/工资、生产率等方面的分化；能解释分化为何导致“总量再加速/表观韧性”与“结构性脆弱”并存；主线贯穿全文并驱动预测与资产含义。\n- 6-8 分：识别到双速格局与分化现象，但传导机制不完整（仅停留在描述），或主线与后续预测/市场结论衔接一般。\n- 0-5 分：主线不清或频繁切换；双速经济仅作概念性提及，缺少可验证刻画。\n\n#### (2) 2025年宏观运行特征复盘与机制归因（8 分）\n- 7-8 分：对 2025 的增长动能、通胀演变、金融条件、劳动力市场与信用环境进行结构化复盘；能将“政策不确定性上升”映射到企业资本开支、风险溢价、期限溢价、美元与金融条件，并指出与 AI 周期的相互作用。\n- 4-6 分：复盘覆盖主要事实，但机制归因偏弱，或对政策不确定性的金融条件渠道讨论不足。\n- 0-3 分：复盘碎片化、缺乏关键变量，或存在明显事实/逻辑错误。\n\n#### (3) 认知增量与可检验命题（7 分）\n- 6-7 分：提出至少 2 个可检验的“增量命题”（例如：AI 投资对生产率/通胀的方向与时滞、金融条件对传统部门的非线性约束、收入分配导致的消费结构差异等），并给出验证指标或数据证据。\n- 3-5 分：有一定新意，但多为定性判断，缺少指标化或对主流观点的清晰对照。\n- 0-2 分：主要复述市场共识与公开叙事，几乎无增量。\n\n---\n\n### 2. 2026前瞻预测：假设、拆解与可持续性（共 35 分）\n\n#### (1) 政策与外生变量假设的明确性与一致性（10 分）\n- 9-10 分：明确列示并量化关键假设（至少包含：财政立场/赤字路径、关税/移民/监管取向、油价或大宗、金融条件或信用利差、美联储反应函数/中性利率判断等），并确保假设与后文增长/通胀/市场结论一致；提供基准+至少 1 个替代情景（上行/下行）。\n- 6-8 分：给出核心假设，但量化不足或情景不完整；部分结论与假设之间存在“隐含跳步”。\n- 0-5 分：假设缺失、含混或前后矛盾，导致预测不可评估。\n\n#### (2) 2026年增长预测与内需结构拆解（10 分）\n- 9-10 分：对 GDP 增长进行分项拆解（消费、非住投资/AI 相关、住投、政府、净出口至少覆盖主要项），并解释各分项的驱动变量（收入与就业、财富效应、融资成本、利润/产能利用率、库存周期等）；能把“双速经济”映射到分项贡献与边际变化。\n- 6-8 分：有分项讨论但不够闭环（缺少驱动变量或缺少量化贡献/方向力度对比）。\n- 0-5 分：仅给总量预测或泛泛而谈，缺乏结构拆解。\n\n#### (3) 通胀前瞻：黏性、再加速与供需约束（8 分）\n- 7-8 分：区分核心/非核心、服务/商品、住房相关通胀的驱动；讨论工资-生产率、房租指标传导、关税/供给冲击、需求韧性对通胀路径的影响；给出通胀回落/反复的条件与指标。\n- 4-6 分：能给出通胀方向判断，但结构与机制拆解不足，或缺少对“黏性”的具体来源讨论。\n- 0-3 分：通胀判断与逻辑支撑薄弱，忽略关键约束（如劳动力、住房、政策冲击）。\n\n#### (4) AI投资、消费、企业投资、地产投资的“支撑力度+可持续性”评估（7 分）\n- 6-7 分：分别评估四类动能的边际变化、融资条件敏感度与约束（如资本开支回报、供给瓶颈、电力/算力/半导体链条、居民资产负债表、信用卡与学生贷压力、商业地产与银行信贷等），并给出可持续性的判据（阈值/领先指标）。\n- 3-5 分：覆盖四项但比较表面，缺少“可持续性条件/领先指标”或缺少对传统部门掣肘的讨论。\n- 0-2 分：只强调某单一动能（如 AI）而忽视约束与均衡条件。\n\n---\n\n### 3. 政策框架与金融市场联动（共 25 分）\n\n#### (1) 特朗普政府政策取向的变量拆解与宏观影响路径（8 分）\n- 7-8 分：将财政扩张/减税、关税与贸易、移民与劳动力供给、监管与产业政策等拆解为可追踪变量，讨论其对增长、通胀、期限溢价与风险偏好的传导路径与可能时序。\n- 4-6 分：能讨论主要政策方向，但变量化不足，或对通胀/增长的双向影响缺少权衡。\n- 0-3 分：政策讨论停留口号化，缺少传导机制。\n\n#### (2) 美联储政策框架变化、反应函数与流动性环境（9 分）\n- 8-9 分：明确讨论美联储目标函数与约束（通胀目标理解、就业权重、金融稳定考量）、政策路径（降息/维持/再收紧条件）、资产负债表/QT-QE 与准备金/逆回购等流动性指标；能将宏观情景映射到利率路径分布与金融条件。\n- 5-7 分：讨论利率方向与部分工具，但对“框架变化/反应函数”或流动性机制不够深入。\n- 0-4 分：仅给出点状预测（如“降息两次”），缺少框架与条件触发。\n\n#### (3) 对利率、美元与金融市场的可交易含义（8 分）\n- 7-8 分：给出对收益率曲线（期限结构/期限溢价）、信用利差、美元（与利差/风险偏好/经常账户逻辑）、美股风格（AI vs 传统）及跨资产联动的清晰判断；明确“兑现条件/催化剂/时间窗口”。\n- 4-6 分：能给出市场方向判断，但缺少触发条件与联动解释，或资产含义与宏观情景对应不清。\n- 0-3 分：市场结论零散或与宏观判断脱节。\n\n---\n\n### 4. 风险不确定性、情景管理与报告可用性（共 15 分）\n\n#### (1) 风险识别的全面性与针对性（7 分）\n- 6-7 分：覆盖并深入讨论至少四类核心风险：通胀黏性/二次上行、财政可持续性与债务上限/拍卖压力、收入分配不均衡与社会/政策反馈、金融条件突变或资产泡沫回撤；每项风险都说明影响路径与首先反映的市场/宏观指标。\n- 3-5 分：列出主要风险，但影响机制与指标不充分，或缺少对“非均衡结构”下风险放大的讨论。\n- 0-2 分：风险提示流于形式或与正文观点重复。\n\n#### (2) 情景推演、敏感性与风险对冲/配置建议（5 分）\n- 4-5 分：至少给出基准+1 个替代情景，并对关键参数做敏感性（如油价、关税强度、金融条件、生产率）或概率权重；给出与情景匹配的资产配置或对冲思路（方向、工具类别、适用条件）。\n- 2-3 分：有情景但缺少敏感性或缺少配置落地；建议偏口号化。\n- 0-1 分：无情景管理，无法用于决策。\n\n#### (3) 表达与可验证性（3 分）\n- 3 分：结构清晰、图表/数据引用规范；关键结论可回溯到数据、假设与机制；对不确定性表述克制（区分事实、判断、情景）。\n- 1-2 分：基本清晰，但存在引用不规范、关键结论缺少证据或表述边界不清。\n- 0 分：结构混乱、结论难以核验或存在明显误导性表述。"
  },
  {
    "id": "38",
    "question": "你是一家中资大型再保险公司的战略与风险管理负责人。经历近几年全球再保险市场高盈利周期后，在国内监管趋严、再保险资本实力持续增强以及自然灾害与责任风险并存的背景下，中国再保险市场正面临承保环境变化与竞争格局调整的关键阶段。\n\n请你撰写一份题为《国内再保险新周期下的承保纪律与长期竞争力》的深度研究报告，重点分析以下问题：\n\n1. 市场周期与供需格局变化：结合国内直保公司分出行为变化、再保险供给能力提升及价格竞争加剧等因素，分析当前中国再保险市场是否正由偏紧状态向相对宽松阶段演进，并评估其对承保纪律与盈利稳定性的影响；\n2. 风险结构与资本约束因素：围绕自然灾害风险上升、责任险与长尾风险不确定性、再保险集中度变化以及监管对资本与偿付能力的要求，分析中资再保险公司在风险转移、资本配置与业务结构调整方面面临的主要挑战；\n3. 战略应对与长期发展路径：在承保环境趋于竞争的背景下，探讨中资再保险公司如何在保持风险定价理性与规模增长之间取得平衡，并从周期管理、科技能力建设与专业化人才布局等维度，总结构建长期竞争优势的关键路径。",
    "classification": "Insurance-Reinsurance",
    "classification_code": "INS-RIN",
    "report_type": "Reinsurance Market Cycle and Strategic Response Report",
    "report_type_zh": "再保险行业周期与战略应对研究报告",
    "language": "zh",
    "expert_evaluation_criteria": "## 国内再保险新周期下的承保纪律与长期竞争力报告评价标准\n\n本评价体系用于对国内再保险市场与公司战略/风险管理类深度研究报告进行结构化量化评估。强调“结论可验证、机制可执行、指标可考核、情景可推演”。体系共设 4 个一级维度、14 个二级维度，总分 100 分，支持人工评审与大模型（LLM）自动评分的一致性应用。\n\n### 1. 市场周期与供需格局洞察（共30分）\n#### （1）周期阶段判断的清晰度与可检验性（10分）\n- 9–10分：明确给出“由偏紧转向相对宽松/仍偏紧/结构性宽松”等判断，并给出时间区间与可检验指标（如费率变动、条款宽松度、再保佣金/滑点、承保利润率、分出比例、再保容量、竞价频次等）；能区分“名义价格”与“条款/结构”层面的软化。\n- 6–8分：判断相对明确，但时间刻度或验证指标不完整；更多定性描述，量化证据不足。\n- 0–5分：仅复述“竞争加剧/市场变化”，缺少明确周期结论或可验证标准。\n\n#### （2）供需拆解的专业度（直保分出、再保供给、竞争行为）（10分）\n- 9–10分：同时拆解需求端（直保公司分出动机变化、资本与净自留策略、业务结构变化）与供给端（中资再保资本实力/偿付能力、同业竞争、可能的追求规模行为），并讨论“价格竞争如何传导到条款、限额、免赔、比例/非比例结构、续转率”；能识别结构性分化（不同险种/地区/层级的不同供需）。\n- 6–8分：覆盖供需两端，但对传导机制或结构性分化讨论较浅。\n- 0–5分：只谈单侧因素或泛泛而谈，缺乏再保业务特有的供需机制分析。\n\n#### （3）对承保纪律与盈利稳定性的影响评估（10分）\n- 9–10分：明确指出在竞争趋强下最易受损的承保纪律环节（定价、条款、风险选择、累积控制、再保结构/追溯、准备金审慎性等），并用风险调整口径衡量影响（如承保利润/综合成本率、风险调整收益、波动率、尾部损失）；能提出“领先指标”（leading indicators）监测纪律松动。\n- 6–8分：能讨论纪律与盈利影响，但多为方向性判断，缺少指标体系或领先指标。\n- 0–5分：未将市场变化与承保纪律/盈利稳定性建立清晰联系。\n\n---\n\n### 2. 风险结构与资本约束分析（共25分）\n#### （1）灾害风险与累积暴露管理分析（9分）\n- 8–9分：识别国内主要灾害风险（如台风、洪涝、地震等）的变化趋势与脆弱点，讨论模型风险与数据不足；能落到累积暴露/峰值区（peak zones）、PML/TVaR/尾部指标、分层承保与限额策略、再保/逆向再保/分散化的控制手段。\n- 5–7分：认识到灾害风险上升，但对累积、模型、分层策略等再保核心抓手展开不够。\n- 0–4分：仅笼统描述“灾害增加”，缺少可操作的风险计量与管理框架。\n\n#### （2）责任险与长尾风险不确定性（8分）\n- 7–8分：能解释长尾风险的关键不确定性来源（通胀/司法环境/赔付周期/再保险条款触发、批量赔案、社会风险等），并讨论准备金、IBNR、赔付发展三角/情景、条款清晰度与追溯风险；能指出对定价与资本占用的影响路径。\n- 4–6分：能识别长尾不确定性，但缺少方法或指标（如发展因子、情景、准备金边际）。\n- 0–3分：将责任/长尾仅作为“风险更大”概念化表述。\n\n#### （3）集中度变化、分散化与对手方/信用风险（4分）\n- 4分：讨论业务集中度（行业/地区/直保对手/险种层级）变化及其对波动与尾部风险的影响，并覆盖对手方信用风险、追偿与结算风险、合同措辞与争议风险。\n- 2–3分：提到集中度或对手方风险，但缺少传导机制或管理措施。\n- 0–1分：忽略集中度与信用/合同风险。\n\n#### （4）监管资本与偿付能力约束下的资本配置与结构优化（4分）\n- 4分：能结合国内偿付能力监管框架讨论资本占用与业务结构调整（如业务线资本效率、风险分散效应、再保购买对资本的影响、盈利与资本的约束联动），并提出资本预算/限额/风险偏好表述方式。\n- 2–3分：笼统提监管趋严与资本压力，缺少对资本配置机制的展开。\n- 0–1分：不涉及或明显不符合国内监管语境与可行边界。\n\n---\n\n### 3. 战略应对与长期竞争力路径（共30分）\n#### （1）“理性定价 vs 规模增长”的权衡框架（10分）\n- 9–10分：提出可执行的权衡机制（如风险偏好声明、利润/资本双约束、RAC/RORAC/风险调整收益门槛、业务线准入与退出标准），并说明在竞争加剧时如何“守纪律但不丢核心客户”（差异化条款、服务、结构化方案、组合定价）。\n- 6–8分：提出平衡思路，但机制与指标不够可执行，更多停留在原则。\n- 0–5分：以口号式表述（“既要规模又要效益”），缺少决策框架。\n\n#### （2）周期管理与组合管理（8分）\n- 7–8分：明确周期工具箱（如价格/条款底线、分保结构调整、动态限额、分出策略与复保、组合再平衡、承保授权与例外审批），并能结合“新周期”给出不同情景下动作清单（软化加深/灾害反转/责任风险恶化等）。\n- 4–6分：提到周期管理，但动作不成体系或缺少情景触发条件。\n- 0–3分：不具备周期管理思维或仅泛谈“顺周期/逆周期”。\n\n#### （3）科技能力建设与数据/模型体系（6分）\n- 5–6分：将科技落到再保关键流程（风险选择、定价、暴露与累积管理、文本条款解析、理赔与追偿、组合监控），明确数据治理、模型验证与模型风险管理（MRM）思路；能说明“短期可落地工具 + 中长期平台化”路径。\n- 3–4分：提到科技/AI，但与再保具体场景结合一般，缺少落地路径与治理。\n- 0–2分：科技表述空泛或偏概念堆砌。\n\n#### （4）专业化人才与组织机制（6分）\n- 5–6分：提出与再保匹配的人才版图（精算/灾害模型/责任险法务与条款/资本管理/投资与ALM/IT数据/承保专家），并说明组织协同与激励约束（承保授权、例外审批、绩效KPI与风险调整口径、前中后台分工）。\n- 3–4分：有人才与组织观点，但缺少机制设计或考核口径。\n- 0–2分：忽略“人和机制”对承保纪律的决定性作用。\n\n---\n\n### 4. 数据方法、可验证性与合规风险（共15分）\n#### （1）数据来源可靠性与口径一致性（6分）\n- 5–6分：清晰标注数据来源与口径（监管/行业统计/公司经营数据/灾害与赔付数据/公开年报等），说明关键指标计算方式与局限；对“价格、费率、条款软化、分出比例、承保利润”等口径做一致性处理。\n- 3–4分：引用数据较多但口径说明不足，或来源权威性/时效性一般。\n- 0–2分：数据来源不明、口径混乱或明显不可信。\n\n#### （2）研究方法严谨性与可复现性（5分）\n- 4–5分：至少具备一种可复现的分析方法（如供需指标框架、情景/敏感性分析、资本效率测算、组合分解、损失分布与尾部指标、准备金情景），并披露关键假设；结论能被方法链条支撑。\n- 2–3分：有方法雏形，但关键步骤/假设披露不足。\n- 0–1分：以观点堆砌为主，缺少方法论与可验证链条。\n\n#### （3）合规、声誉与模型/数据风险提示（4分）\n- 4分：识别监管合规边界、数据使用与隐私、模型偏误与过拟合、条款争议与诉讼、重大灾害下声誉风险等，并说明这些风险如何约束战略与承保政策。\n- 2–3分：有风险提示但偏概括，未体现对战略选择的约束。\n- 0–1分：缺少合规/声誉/模型风险意识。"
  },
  {
    "id": "39",
    "question": "近期，债券市场收益率快速上行，国债期货明显调整，市场情绪趋于谨慎，部分投资者开始担忧债券市场是否正从牛市阶段转向“债熊”格局。在此背景下，请你作为一名券商固收研究团队的分析师，撰写一份题为《债熊要来了吗？》的深度点评报告，并完成以下分析任务：\n\n1. 市场复盘与调整归因：系统回顾本轮国债期货（如T主力合约）与现券收益率（如10年期国债）调整的幅度与节奏，重点剖析驱动本轮调整的核心因素，例如经济数据好转、市场风险偏好的变化（如股市赚钱效应增强）、利率债供给压力、海外利率环境的影响、宽信用政策的推进、资金面边际变化以及通胀预期的阶段性抬升等；分析各久期、品种调整幅度和节奏的差异，以及各利差如期限利差、品种利差和信用利差等的表现，说明其驱动因素。\n2. 基本面与政策面博弈：结合近期宏观经济高频指标（如PMI、社融、工业生产等）以及央行公开市场操作与政策表态，分析当前基本面修复的强度是否足以支撑一轮趋势性的债券熊市，并讨论货币政策在“稳增长”与“防止资金空转”之间的权衡及其对流动性的影响；\n3. 趋势研判与交易策略：在上述分析基础上，结合历史上债市牛熊转换时的基本面与货币政策等背景和债市表现特征（如2016年10月-2017年11月、2020年5月-2021年11月），给出明确判断：当前债市调整更可能是“牛熊切换”的起点，还是牛市过程中的阶段性回撤？并据此分别为配置型机构（如商业银行）和交易型机构（如券商、基金）提出久期管理、杠杆运用、品种选择及仓位调整方面的策略建议。",
    "classification": "Capital Markets-Fixed Income & Rates Research",
    "classification_code": "CAP-FIX",
    "report_type": "Fixed Income Market In-depth Commentary and Strategy Report",
    "report_type_zh": "固定收益市场深度点评与策略报告",
    "language": "zh",
    "expert_evaluation_criteria": "## 《债熊要来了吗？》固收深度点评报告评价标准\n\n本评价体系用于对“债市调整—牛熊判断—交易策略”类固收点评报告进行结构化、可执行评分，兼容人工评审与大模型自动评分。体系共设 **4个一级维度、15个二级维度**，总分 **100分**。\n\n---\n\n### 1. 市场复盘与调整归因（25分）\n#### （1）期货与现券复盘的准确性与完整性（8分）\n- **7–8分**：明确给出样本区间（起止日期/关键事件点），量化呈现T主力（或活跃合约）与10Y国债收益率的**变动幅度、日内/日间节奏、关键拐点**；包含必要的对照（如2Y/5Y/30Y、国开与国债、IRS等）以验证“不是单点数据误判”。\n- **4–6分**：能描述主要方向与大致幅度，但区间设定/关键节点不清，或只写期货不写现券（或反之），缺少曲线/品种对照。\n- **0–3分**：复盘笼统、缺少定量信息或事实错误（合约、收益率方向、幅度等明显不符）。\n\n#### （2）调整“节奏与结构”的拆解能力（6分）\n- **5–6分**：区分“快速上行/缓慢抬升”“平坦化/陡峭化”“名义利率 vs 实际利率/通胀预期”等结构变化；解释期现差、基差、CTD/久期变化或移仓因素对期货表现的影响（无需过度技术化，但逻辑要对）。\n- **3–4分**：提到节奏或曲线形态，但缺少结构化拆解或对期货技术因素解释不足。\n- **0–2分**：仅叙述“跌了/涨了”，无法说明调整的结构特征。\n\n#### （3）核心驱动因素归因：主次排序与证据链（8分）\n- **7–8分**：围绕题干因素（宽信用推进、资金面边际变化、通胀预期抬升等）建立“事件—机制—市场变量”的证据链；能给出**主因/次因**及相互作用（如信用扩张预期→风险偏好→期限溢价；资金面收敛→杠杆降速→曲线短端/中端传导）。\n- **4–6分**：归因覆盖面较全，但主次不清，或更多为经验判断，证据（数据/政策/市场指标）引用不足。\n- **0–3分**：归因单一或与市场变量对应关系弱，出现“用结论替代论证”。\n\n#### （4）历史可比阶段/跨周期对照（3分）\n- **3分**：选择1–2段可比时期（如“稳增长发力+防空转”组合、或资金面阶段性偏紧）做对照，指出本轮相同点/不同点及为何影响牛熊判断。\n- **1–2分**：仅泛泛提及“历史上也发生过”，缺少可比性说明。\n- **0分**：无对照。\n\n---\n\n### 2. 基本面与政策面博弈（30分）\n#### （1）高频与宏观指标选取、质量与解读（10分）\n- **9–10分**：选取与债市相关且时效匹配的指标（PMI、社融结构、工业生产、地产链、出口、物价、就业/工资、库存周期等），并能解释“数据改善=短期脉冲”还是“趋势修复”；区分**同比/环比、季调、基数效应**等口径影响。\n- **6–8分**：覆盖主要指标，但对口径/噪音处理不足，或结论停留在“好/坏”描述。\n- **0–5分**：指标零散或与论点关联弱，用个别数据点替代整体判断。\n\n#### （2）“修复强度是否足以支撑趋势熊市”的判别框架（7分）\n- **6–7分**：给出清晰判别框架（如：潜在增速与产出缺口→通胀与政策函数→实际利率/期限溢价；或信用周期斜率→融资需求→利率中枢），并明确“趋势熊市”需要满足的条件（至少列出2–3条）。\n- **3–5分**：有框架雏形，但条件不够可检验，或只给方向性判断。\n- **0–2分**：直接下结论，缺少判别标准。\n\n#### （3）资金面与央行操作：从工具到传导（8分）\n- **7–8分**：结合公开市场操作（OMO、MLF、降准预期等）、政策表态与资金价格/量（DR007、回购成交量、同业存单利率等）解释流动性边际变化；能说明“稳增长”与“防空转”如何体现在**价格端（利率）**与**数量端（投放/回笼）**，并映射到杠杆行为与利率曲线。\n- **4–6分**：提及央行工具与资金面现象，但传导链条不完整或缺少市场化指标验证。\n- **0–3分**：对政策工具理解错误或与资金面变化对应不清。\n\n#### （4）宽信用推进的结构性分析（5分）\n- **5分**：不仅看社融总量，还拆分**政府债、企业中长贷、居民按揭、票据/短贷**等结构，讨论其对利率的影响路径（供给冲击、融资需求、风险偏好、期限溢价）。\n- **2–4分**：能提到社融结构但未形成机制解释。\n- **0–1分**：仅用“宽信用=利空债券”简单推导。\n\n---\n\n### 3. 趋势研判与交易策略（30分）\n#### （1）牛熊切换判断的清晰度与可证伪性（10分）\n- **9–10分**：明确回答“更可能是牛熊切换起点/还是阶段性回撤”，并给出**可证伪条件**（例如：若资金利率中枢持续抬升至某区间且信用斜率改善持续X月，则熊市概率上升；反之若通胀与融资脉冲回落则回撤性质更大）；论证能回扣前两部分的证据链。\n- **6–8分**：结论明确但证伪条件不够具体，或对关键变量的持续性判断偏主观。\n- **0–5分**：结论含糊（“可能”“或许”）或前后自相矛盾。\n\n#### （2）情景分析/敏感性：关键变量与催化剂（6分）\n- **5–6分**：至少给出2种情景（如“强修复+偏紧流动性”“弱修复+稳流动性”），明确关键变量（社融脉冲、通胀、资金利率中枢、政府债供给节奏、汇率/外部利率等）与可能催化剂/兑现窗口。\n- **3–4分**：有情景描述但缺少变量联动或时间/触发条件。\n- **0–2分**：无情景推演。\n\n#### （3）配置型机构（银行等）策略：久期、杠杆、仓位与负债约束（7分）\n- **6–7分**：建议能体现银行约束（负债稳定性、久期缺口、OCI/AC配置、资本占用、同业存单成本等），给出可执行动作（如久期从X调整到Y的方向性区间、分批建仓/止盈止损原则、曲线段选择、信用/利率品种搭配）。\n- **3–5分**：给出方向性建议但缺少“在银行约束下如何落地”的细节。\n- **0–2分**：泛化建议（如“逢高减仓/逢低买入”）缺少可执行性。\n\n#### （4）交易型机构（券商/基金）策略：交易结构与风控（7分）\n- **6–7分**：给出交易框架（方向/曲线/利差/期现/套保），明确使用工具（国债期货、IRS、现券、回购杠杆等）及风险控制（止损条件、仓位上限、基差/保证金压力、流动性冲击预案）。\n- **3–5分**：有交易观点但缺少结构化策略与风控条款。\n- **0–2分**：仅给观点、不具备交易可用性或忽略杠杆与流动性风险。\n\n---\n\n### 4. 数据方法、风险与表达合规（15分）\n#### （1）数据来源、口径披露与可复现性（6分）\n- **5–6分**：关键数据注明来源与口径（交易所/中债/同花顺/Wind/央行/统计局等均可描述为“权威公开数据”），重要计算（变动bp、分位数、期限利差、资金利率中枢等）有清晰定义，使读者可复核。\n- **3–4分**：数据大体可信但口径/来源披露不完整。\n- **0–2分**：数据来源不明、口径混乱或难以复核。\n\n#### （2）风险提示的针对性与“对冲/应对”闭环（5分）\n- **5分**：风险不仅罗列，还说明影响路径与应对方式（如政策超预期转向、资金面突然收紧、通胀反弹、政府债供给放量、外部利率上行等），并与策略中的止损/对冲对应。\n- **2–4分**：提到主要风险但缺乏机制或应对措施。\n- **0–1分**：无风险提示或流于形式。\n\n#### （3）研究表达专业性与合规边界（4分）\n- **4分**：结构清晰（摘要/结论前置、分任务作答）、术语使用规范（bp、久期、期限利差等），避免“确定性收益”表述；如涉及建议，具备必要的风险揭示与合规措辞（研究观点属性清楚）。\n- **2–3分**：表达基本清楚但结构松散或有少量不严谨表述。\n- **0–1分**：逻辑混乱、表述夸大或明显越界。"
  },
  {
    "id": "40",
    "question": "在财富管理竞争持续加剧、客户分层不断深化的背景下，越来越多大型商业银行开始系统性引入 DT（数据技术）与 AI，以提升客户洞察能力并构建数字化、可规模化的运营机制。然而，相关技术投入在多大程度上能够转化为可量化、可持续的经营价值，仍有待系统评估。\n\n请围绕“DT 和 AI 赋能财富管理客户洞察与数字化运营机制建设”这一主题，完成一份结构化研究报告，重点从以下三个层面展开分析：\n\n1. 客户与产品基础能力构建  \n如何基于 DT 与 AI 构建可运营的客户分层与画像体系？不同客群在需求结构、渠道偏好与服务模式上有何差异？  \n在此基础上，如何构建覆盖财富与私银业务的动态产品体系，并形成统一的产品与服务视图？\n\n2. 精细化运营与触客机制  \n如何利用机器学习、大模型等技术，实现客户与产品的精准匹配，提升投顾配置与营销转化效率？  \n如何基于客户偏好开展内容陪伴与投教服务？  \n如何设计客户旅程与触客节奏，在降低打扰的同时提升转化效率？  \n如何构建以 AI 为核心的数字化运营机制，并实现策略评估与持续优化？\n\n3. 人机协作与经营成效评估  \nDT 与 AI 的引入如何重塑财富顾问与投顾团队的分工？如何构建“AI 赋能 + 人工决策”的协作模式？  \n结合实际案例，分析其在客户体验、AUM 增长与运营效率方面的影响，并评估其可持续性。",
    "classification": "Banking-Customer & Marketing Management",
    "classification_code": "BNK-CMM",
    "report_type": "Wealth Management Customer Insight and Digital Operations Research Report",
    "report_type_zh": "财富管理客户洞察与数字化运营机制研究报告",
    "language": "zh",
    "expert_evaluation_criteria": "## 客户画像与数字化运营研究报告评价体系\n\n本评价体系适用于评估商业银行财富管理与私人银行条线中，围绕 **客户画像构建与数字化客户运营体系建设** 所形成的研究报告质量。体系重点关注客户分层洞察的专业深度、数字化运营逻辑的可落地性、数据与方法的银行适配性，以及对客户经营与 **AUM 增长与客户生命周期价值提升** 的实际指导意义。体系共设 **4 个一级维度、11 个二级维度，总分 100 分**，支持人工专家评审与大模型（LLM）自动评分的一致性应用。\n\n---\n\n## 1. 客户画像与需求—产品逻辑洞察质量（30 分）\n\n### （1）客户分层与画像精细度（10 分）\n\n* **9–10 分**：清晰区分大众富裕、准高净值、高净值及超高净值客群；画像维度覆盖资产规模、财富来源、风险偏好、生命周期阶段、行为特征与服务偏好等多个层面，并能够直接映射至银行客户标签体系或经营分层策略。\n* **6–8 分**：客户分层总体合理，画像维度较完整，但部分内容停留在描述层，未充分转化为可运营标签。\n* **0–5 分**：客户划分粗糙或概念化，画像泛化，难以支撑差异化客户经营。\n\n### （2）客户核心需求与痛点识别能力（10 分）\n\n* **9–10 分**：能够准确识别不同客群在资产配置、收益稳定性、流动性管理、税务与财富传承、服务体验等方面的核心需求，并区分显性与隐性需求；结合主流私人银行产品与服务体系，形成需求与产品匹配逻辑，痛点分析贴合真实业务场景。\n* **6–8 分**：覆盖主要需求方向，但痛点拆解深度有限，对具体产品与服务关联分析不足。\n* **0–5 分**：需求分析笼统，与实际私人银行业务场景关联度较低。\n\n### （3）洞察增量与差异化认知（10 分）\n\n* **9–10 分**：在客户画像或需求分析中提出具有认知增量的洞察（如财富来源结构变化、代际迁移趋势、数字化行为差异等），并对传统私人银行经营模式形成修正或补充。\n* **6–8 分**：具备一定细分视角，但整体仍以行业共识为主。\n* **0–5 分**：主要重复公开材料或行业常识，缺乏新增洞察价值。\n\n---\n\n## 2. 数字化运营逻辑与体系设计（30 分）\n\n### （1）数字化运营框架完整性（10 分）\n\n* **9–10 分**：构建清晰的“数据 → 客户洞察 → 触达与服务 → 转化与留存 → 客户价值提升”运营闭环，并结合财富管理业务实际展开说明，体现长期客户经营逻辑。\n* **6–8 分**：运营框架基本完整，但在投后陪伴或长期留存等关键环节论述不足。\n* **0–5 分**：运营逻辑零散，缺乏系统性设计。\n\n### （2）数字化手段与业务场景适配性（10 分）\n\n* **9–10 分**：对数据中台、客户标签体系、智能投顾、AI 推荐、CRM 与客户旅程管理等工具的应用路径与银行实际高度匹配，并区分零售财富与私人银行客群的运营差异。\n* **6–8 分**：数字化工具选择合理，但对私行业务特性与差异化运营考虑不足。\n* **0–5 分**：偏互联网或泛科技视角，缺乏银行业务落地可行性。\n\n### （3）组织架构与运营机制协同（10 分）\n\n* **9–10 分**：系统讨论总行—分行协同机制、数字化工具嵌入方式及客户经理/投顾协作模式，明确角色分工、流程变化及绩效影响，体现财富管理业务对人机协同的依赖。\n* **6–8 分**：提及组织协同，但分析较为概括。\n* **0–5 分**：将数字化简单等同自动化或忽视一线人员作用。\n\n---\n\n## 3. 数据可靠性与研究方法可借鉴性（25 分）\n\n### （1）数据来源与客户数据合理性（10 分）\n\n* **9–10 分**：清晰区分银行内部数据、行业统计数据、调研数据及第三方研究来源；说明数据时效性与可获得性，并符合银行数据合规与现实使用边界。\n* **6–8 分**：数据来源总体可靠，但颗粒度或披露程度不足。\n* **0–5 分**：数据来源模糊或明显脱离银行可获得数据范围。\n\n### （2）分析方法与可验证性（8 分）\n\n* **7–8 分**：明确描述客户画像构建、分群分析及运营策略形成路径，方法具备可复现性或清晰执行流程。\n* **4–6 分**：方法逻辑基本成立，但关键步骤说明不足。\n* **0–3 分**：方法论缺失或研究结论不可验证。\n\n### （3）方法与业务目标匹配度（7 分）\n\n* **6–7 分**：方法选择能够直接支撑客户转化率提升、客户留存、AUM 增长及客户生命周期价值（CLV）提升等核心经营目标。\n* **3–5 分**：方法具备一定可用性，但与经营目标衔接不紧密。\n* **0–2 分**：研究方法与银行经营目标明显脱节。\n\n---\n\n## 4. 实际业务价值与风险合规意识（15 分）\n\n### （1）对私人银行经营实践的指导价值（10 分）\n\n* **9–10 分**：明确说明研究成果在客户运营中的应用方式，如提升转化率、增强客户黏性、优化服务模式或推动 AUM 增长，具备直接业务指导意义。\n* **6–8 分**：具有一定应用价值，但适用场景有限。\n* **0–5 分**：停留在理论层面，难以指导实际经营。\n\n### （2）合规、数据安全与声誉风险意识（5 分）\n\n* **4–5 分**：系统识别数据隐私保护、算法偏见、客户信任及监管约束等风险，并说明其对数字化运营设计与实施的影响。\n* **2–3 分**：提及合规风险，但缺乏深入分析。\n* **0–1 分**：未充分考虑银行合规与声誉风险。"
  }
]