AI 底层逻辑 / 经典论文
AI Foundations 论文精读索引
先读 ../AI_SYSTEMATIC_LEARNING_ROADMAP_2026.md。本目录是论文和架构深度笔记库，不是要求从 1 到 174 平均用力读完的每日课程。
425 行ai-foundations/README.md
AI Foundations
AI 底层逻辑、经典论文、系统架构、评估证据和金融零售落地模式的长期学习入口。 这个目录是新的技术底座扩展，不替代
docs/ai/、docs/llm/、docs/AI_ENGINEERING_METHODOLOGY.md或 ABPA 业务资产。
How To Use
先读 ../AI_SYSTEMATIC_LEARNING_ROADMAP_2026.md。本目录是论文和架构深度笔记库，不是要求从 1 到 174 平均用力读完的每日课程。
推荐三遍阅读：
| 轮次 | 读法 | 目标 |
|---|---|---|
| 第一遍 | 只读核心导读、概念图、架构模型、金融场景、指标/控制/证据 | 建立系统理解 |
| 第二遍 | 回看 ADR、failure modes、metrics、templates | 做 case、架构取舍和上线门禁 |
| 第三遍 | 集中整理 decision narrative、evidence pack、长 checklist | 形成可复盘的设计和治理材料 |
单篇第一遍先跳过：
- 角色化表达、长问答和求职材料
- 练习计划
- 过长 checklist
- 详细 ADR 模板
- Source anchors 的逐条扩展阅读
每读完一个模块，只产出一个小工件即可，例如一张 C4 图、一个 eval matrix、一个 action authority matrix 或一个 control-to-evidence graph。不要要求每篇都同时产出图、ADR、eval、金融映射和长篇表达材料；第二、三遍再把这些证据串成完整方案。
Course Module Map
本表为
../AI_SYSTEMATIC_LEARNING_ROADMAP_2026.md第 3 节模块表（单一事实源）的视图，如有出入以路线图为准。第一遍必读子集见路线图第 4 节各周清单；模块外 24 篇选读池见路线图 3.3；每模块配套 playbook 见路线图 3.2。
| 模块 | 先读论文 | 学习重点 |
|---|---|---|
| 1 | 01, 04, 05, 10, 11, 22 | LLM 底座、对齐、推理和模型边界 |
| 2 | 02, 13, 14, 23, 31, 93 | RAG、GraphRAG、embedding、知识治理和语义层 |
| 3 | 03, 12, 15, 16, 21, 27, 42, 99, 100 | Agent、工具调用、权限、状态、工作流和授权 |
| 4 | 08, 17, 18, 24, 26, 44, 60, 63, 88, 90 | EvalOps、release gate、traceability 和持续治理 |
| 5 | 06, 07, 09, 28, 30, 37, 41, 45, 57, 59, 168 | 模型适配、成本、性能、数据、模型组合和技术债 |
| 6 | 34-36, 47-56（37 为跨模块复用，归属模块 5） | 推荐、排序、特征、预测、异常、因果、优化和传统 ML 决策组件 |
| 7 | 128-147（第一遍必读: 128, 131, 139, 143, 144, 145） | 金融零售场景：支付、信贷、开户、AML、投诉、监管报送 |
| 8 | 58, 78, 82, 95-126, 170-174 | 治理、模型风险、证据、监管、三道防线和审计 |
| 9 | 64, 67-77, 79-81, 83-94, 150-162 | 产品架构、企业架构、平台化、组织和运营模型 |
Paper Reading Track
下面是完整索引，主要用于查找和第二遍深读。第一遍请按上面的模块地图读。
| Order | Paper note | Core capability |
|---|---|---|
| 1 | papers/01-attention-is-all-you-need.md | Transformer、Q/K/V、attention、decoder-only LLM、企业架构映射 |
| 2 | papers/02-retrieval-augmented-generation.md | RAG、外部知识、检索链路、引用、权限、企业知识治理 |
| 3 | papers/03-react-toolformer-agent-foundations.md | ReAct、Toolformer、工具调用、Agent 状态、HITL、审计 |
| 4 | papers/04-instructgpt-rlhf-alignment.md | SFT、Reward Model、RLHF、Constitutional AI、对齐与产品治理 |
| 5 | papers/05-chain-of-thought-self-consistency.md | CoT、Self-Consistency、test-time compute、解释与证据边界、推理型产品设计 |
| 6 | papers/06-lora-peft-adaptation.md | LoRA、PEFT、QLoRA、模型适配、RAG/微调边界、adapter 治理 |
| 7 | papers/07-inference-optimization-kv-cache-flashattention-speculative.md | KV Cache、FlashAttention、Speculative Decoding、成本/延迟/SLO 架构 |
| 8 | papers/08-llm-as-judge-evaluation.md | LLM-as-Judge、G-Eval、MT-Bench、rubric、golden set、release gate |
| 9 | papers/09-mixture-of-experts-sparse-scaling.md | MoE、sparse activation、router/gating、expert capacity、load balancing、成本/SLO/多租户架构 |
| 10 | papers/10-scaling-laws-pretraining-bert-gpt-t5.md | Scaling laws、pretraining objective、BERT/GPT/T5、RAG/微调/自训 foundation model 取舍 |
| 11 | papers/11-dpo-constitutional-ai-preference-optimization.md | DPO、RLAIF、Constitutional AI、preference data、reward hacking、alignment governance |
| 12 | papers/12-tool-use-security-prompt-injection.md | Tool use security、indirect prompt injection、data exfiltration、confused deputy、tool gateway |
| 13 | papers/13-rag-evaluation-ragas-retrieval-metrics.md | RAG evaluation、RAGAS、retrieval metrics、faithfulness、citation support、release gate |
| 14 | papers/14-graphrag-knowledge-graph-rag.md | GraphRAG、knowledge graph RAG、entity/relation/path/community、multi-hop reasoning、graph eval |
| 15 | papers/15-generative-agents-memory-reflection-planning.md | Generative Agents、memory stream、reflection、planning、enterprise memory governance |
| 16 | papers/16-autogen-multi-agent-orchestration.md | AutoGen、multi-agent conversation、role/handoff/shared state、HITL、orchestration eval |
| 17 | papers/17-helm-holistic-evaluation-models.md | HELM、holistic evaluation、scenario/metric/adapter、透明评测、模型选择和 release gate |
| 18 | papers/18-model-cards-datasheets-ai-documentation.md | Model Cards、Datasheets、AI 文档化、dataset/model/system evidence、治理证据链 |
| 19 | papers/19-tree-of-thoughts-planning-search.md | Tree of Thoughts、planning search、多候选方案、搜索预算、人工选择点和复杂任务推理 |
| 20 | papers/20-self-rag-crag-agentic-retrieval.md | Self-RAG、CRAG、retrieval need、context quality gate、纠错检索、拒答/升级 |
| 21 | papers/21-agentbench-taubench-agent-evaluation.md | AgentBench、τ-bench、tool-agent-user interaction、policy oracle、tool sandbox、agent release gate |
| 22 | papers/22-mechanistic-interpretability-transformer-circuits-sae.md | Mechanistic interpretability、Transformer circuits、sparse autoencoders、解释性证据和模型风险边界 |
| 23 | papers/23-long-context-lost-in-the-middle-ruler.md | Long context、Lost in the Middle、RULER、position robustness、长上下文/RAG 混合架构 |
| 24 | papers/24-dspy-opro-automatic-prompt-optimization.md | DSPy、OPRO、APE、LM program、prompt registry、eval-driven prompt optimization |
| 25 | papers/25-reflexion-self-refine-agent-feedback-loops.md | Reflexion、Self-Refine、feedback object、refinement policy、reflection memory、agent 改进循环 |
| 26 | papers/26-process-supervision-step-by-step-verification.md | Process supervision、step-level verification、outcome vs process supervision、critical step gate |
| 27 | papers/27-structured-output-constrained-decoding-lmql-guidance.md | Structured output、constrained decoding、LMQL、Guidance、schema contract、tool payload |
| 28 | papers/28-model-routing-semantic-cache-frugal-ai.md | Model routing、semantic cache、FrugalGPT、RouteLLM、成本/质量/SLO 路由 |
| 29 | papers/29-swe-bench-webarena-agent-benchmarks.md | SWE-bench、WebArena、OSWorld、GAIA、真实环境 Agent benchmark、release gate |
| 30 | papers/30-mamba-state-space-models-efficient-sequence.md | Mamba、S4、state space models、long sequence、成本/延迟/模型架构取舍 |
| 31 | papers/31-embeddings-ann-vector-search-faiss-hnsw.md | Embeddings、ANN、FAISS、HNSW、Sentence-BERT、RAG 检索底座 |
| 32 | papers/32-clip-multimodal-embeddings-product-architecture.md | CLIP、多模态 embeddings、图文对齐、zero-shot、文搜图/图搜图 |
| 33 | papers/33-diffusion-latent-diffusion-generative-media.md | Diffusion、Latent Diffusion、生成式媒体、品牌/版权/安全治理 |
| 34 | papers/34-graph-neural-networks-gnn-fraud-risk-architecture.md | GCN、GraphSAGE、GAT、图学习、欺诈/AML/风险架构 |
| 35 | papers/35-recommender-systems-youtube-wide-deep-two-tower.md | 推荐系统、YouTube DNN、Wide & Deep、Two-Tower、next-best-action |
| 36 | papers/36-learning-to-rank-lambdamart-neural-ranking.md | Learning to Rank、RankNet、LambdaMART、NDCG、搜索/推荐/告警排序 |
| 37 | papers/37-feature-stores-real-time-ml-feast-michelangelo.md | Feature Store、Feast、Michelangelo、实时特征、point-in-time correctness |
| 38 | papers/38-zanzibar-cedar-opa-authorization-policy-architecture.md | Zanzibar、Cedar、OPA、ReBAC、ABAC、PDP/PEP、AI Agent 权限架构 |
| 39 | papers/39-federated-learning-fedavg-cross-silo-ai.md | Federated Learning、FedAvg、cross-silo AI、secure aggregation、跨机构协作 |
| 40 | papers/40-differential-privacy-dpsgd-ai-data-protection.md | Differential Privacy、DP-SGD、epsilon、privacy budget、AI 数据保护 |
| 41 | papers/41-knowledge-distillation-small-models-quantization.md | Knowledge Distillation、小模型、量化、teacher-student eval、模型组合策略 |
| 42 | papers/42-durable-execution-agent-workflow-state-machines.md | Durable Execution、Agent 状态机、Saga、幂等、workflow replay |
| 43 | papers/43-digital-twin-agent-based-simulation-ai-decisioning.md | Digital Twin、agent-based simulation、decision twin、scenario simulation、AI 决策仿真 |
| 44 | papers/44-online-experimentation-cuped-release-science-ai-products.md | Online experimentation、CUPED、release science、guardrail metrics、AI 发布科学 |
| 45 | papers/45-data-lineage-contracts-openlineage-ai-data-quality.md | Data lineage、data contracts、OpenLineage、DataHub、AI 数据质量 |
| 46 | papers/46-ai-security-operations-mitre-atlas-owasp-csf.md | AI Security Operations、MITRE ATLAS、OWASP LLM Top 10、NIST CSF、AI SOC |
| 47 | papers/47-time-series-forecasting-tft-deepar-foundation-models.md | Time-series forecasting、DeepAR、TFT、TimesFM、概率预测、forecast-to-decision |
| 48 | papers/48-anomaly-detection-isolation-forest-autoencoder-risk-monitoring.md | Anomaly detection、Isolation Forest、autoencoder、streaming detection、risk monitoring |
| 49 | papers/49-causal-discovery-dowhy-econml-structural-causal-models.md | Causal discovery、SCM、DAG、DoWhy、EconML、NOTEARS、结构化产品决策 |
| 50 | papers/50-optimization-operations-research-or-tools-ai-decisioning.md | Optimization、operations research、OR-Tools、CP-SAT、solver decision service |
| 51 | papers/51-contextual-bandits-linucb-thompson-online-learning.md | Contextual bandits、LinUCB、Thompson Sampling、propensity logging、adaptive experimentation |
| 52 | papers/52-reinforcement-learning-offline-rl-cql-policy-decisioning.md | Reinforcement learning、MDP、offline RL、CQL、reward design、policy guardrails |
| 53 | papers/53-bayesian-optimization-botorch-optuna-experiment-design.md | Bayesian optimization、BoTorch、Optuna、surrogate、acquisition、实验预算 |
| 54 | papers/54-calibration-conformal-prediction-uncertainty-governance.md | Calibration、conformal prediction、ECE、coverage、abstention、uncertainty governance |
| 55 | papers/55-data-centric-ai-snorkel-programmatic-labeling.md | Data-centric AI、Snorkel、programmatic labeling、weak supervision、label quality governance |
| 56 | papers/56-active-learning-human-in-the-loop-labeling.md | Active learning、HITL labeling、query strategy、SME review、feedback operations |
| 57 | papers/57-dataset-shift-monitoring-model-performance.md | Dataset shift、drift monitoring、training-serving skew、outcome lag、model performance operations |
| 58 | papers/58-ai-management-system-iso42001-operating-model.md | AI Management System、ISO 42001、NIST AI RMF、AI inventory、operating model |
| 59 | papers/59-hidden-technical-debt-ml-systems-ai-architecture.md | ML technical debt、CACE、entanglement、hidden feedback loops、AI architecture debt |
| 60 | papers/60-cd4ml-mlops-continuous-delivery-ai-release.md | CD4ML、MLOps、release bundle、shadow/canary/ramp、AI release engineering |
| 61 | papers/61-human-ai-interaction-guidelines-product-design.md | Human-AI interaction、calibrated trust、recoverability、automation bias、AI product design |
| 62 | papers/62-ai-architecture-decision-records-governance.md | AI ADR、decision governance、risk tier、evidence、reversal trigger、architecture knowledge |
| 63 | papers/63-ai-requirements-engineering-gqm-eval-contracts.md | AI requirements engineering、GQM、eval contract、release gate、monitoring gate |
| 64 | papers/64-ai-quality-attributes-atam-architecture-tradeoff.md | AI quality attributes、ATAM、utility tree、tradeoff point、architecture review |
| 65 | papers/65-ai-safety-engineering-stpa-control-structure.md | AI safety engineering、STPA、control structure、unsafe control action、safety constraints |
| 66 | papers/66-sociotechnical-ai-resilience-work-as-done.md | Sociotechnical AI、work-as-done、human-AI collaboration、resilience、operating model |
| 67 | papers/67-ai-capability-based-planning-business-architecture.md | AI capability-based planning、business architecture、capability map、value stream、portfolio roadmap |
| 68 | papers/68-wardley-mapping-ai-product-platform-strategy.md | Wardley Mapping、AI product strategy、value chain、evolution axis、build-buy-platformize |
| 69 | papers/69-conway-team-topologies-ai-platform-operating-model.md | Conway's Law、Team Topologies、AI platform operating model、cognitive load、team API |
| 70 | papers/70-dora-space-ai-sdlc-engineering-productivity.md | DORA、SPACE、AI SDLC、code agent governance、engineering productivity、release gate |
| 71 | papers/71-continuous-discovery-opportunity-solution-tree-ai-products.md | Continuous discovery、opportunity solution tree、AI opportunity taxonomy、assumption map、pilot learning |
| 72 | papers/72-jtbd-outcome-driven-innovation-ai-use-case-selection.md | Jobs-to-be-Done、Outcome-Driven Innovation、job map、outcome statements、automation boundary |
| 73 | papers/73-north-star-ai-product-metrics-value-measurement.md | North Star metrics、AI value measurement、guardrail metrics、risk-adjusted value、benefits realization |
| 74 | papers/74-ai-product-operating-model-empowered-teams.md | AI product operating model、empowered teams、product trio、decision rights、governance cadence |
| 75 | papers/75-ai-portfolio-management-funding-governance.md | AI portfolio management、funding governance、portfolio kanban、risk-adjusted value、scale/stop decision |
| 76 | papers/76-service-blueprint-ai-customer-journey-trust.md | AI service blueprint、customer journey、trust calibration、frontstage/backstage、handoff and appeal |
| 77 | papers/77-ai-business-process-reengineering-bpmn-dmn.md | AI BPR、BPMN、DMN、decision automation、eval traceability、control evidence |
| 78 | papers/78-ai-risk-appetite-policy-product-management.md | AI risk appetite、policy product management、risk-tiered guardrails、runtime controls、stop rules |
| 79 | papers/79-enterprise-ai-reference-architecture-control-plane.md | Enterprise AI reference architecture、control plane、model/tool gateway、eval/observability、evidence |
| 80 | papers/80-ai-product-line-engineering-reuse-platform-assets.md | AI product line engineering、core assets、variation points、platform reuse、asset governance |
| 81 | papers/81-ai-maturity-model-roadmap-capability-assessment.md | AI maturity model、capability domains、evidence standard、roadmap sequencing、capability assessment |
| 82 | papers/82-ai-control-library-assurance-evidence-graph.md | AI control library、assurance case、claim-risk-control-evidence graph、release assurance |
| 83 | papers/83-ai-domain-driven-design-ubiquitous-language.md | AI DDD、bounded context、ubiquitous language、RAG boundary、eval vocabulary |
| 84 | papers/84-event-storming-agent-workflow-design.md | AI EventStorming、domain event、command、policy、agent workflow、compensation |
| 85 | papers/85-ai-knowledge-work-redesign-role-task-architecture.md | AI knowledge work redesign、role-task architecture、human-AI teaming、accountability |
| 86 | papers/86-ai-platform-service-catalog-golden-paths.md | AI platform service catalog、golden paths、self-service with guardrails、platform metrics |
| 87 | papers/87-ai-architecture-views-c4-arc42-42010.md | AI architecture views、C4、arc42、ISO 42010、stakeholder concern、viewpoint governance |
| 88 | papers/88-ai-architecture-fitness-functions-continuous-governance.md | architecture fitness functions、continuous governance、eval gate、runtime telemetry、evidence |
| 89 | papers/89-contract-first-ai-tool-api-design-openapi-asyncapi.md | contract-first AI tool/API、OpenAPI、AsyncAPI、JSON Schema、tool/event governance |
| 90 | papers/90-ai-traceability-requirements-eval-control-graph.md | AI traceability graph、requirements、eval、control、ADR、telemetry、audit evidence |
| 91 | papers/91-ai-enterprise-architecture-togaf-archimate-adm.md | AI enterprise architecture、TOGAF ADM、ArchiMate、architecture repository、governance |
| 92 | papers/92-banking-ai-reference-models-bian-fibo-iso20022.md | banking AI reference models、BIAN、FIBO、ISO 20022、financial semantic boundaries |
| 93 | papers/93-ai-semantic-interoperability-rdf-owl-shacl.md | semantic interoperability、RDF、OWL、SHACL、semantic contracts、GraphRAG governance |
| 94 | papers/94-ai-value-stream-management-flow-metrics.md | AI value stream management、flow metrics、portfolio-to-platform trace、benefits realization |
| 95 | papers/95-ai-regulatory-architecture-eu-ai-act-nist-iso42001.md | AI regulatory architecture、EU AI Act、NIST AI RMF、ISO 42001、obligations-to-controls |
| 96 | papers/96-ai-model-risk-validation-independent-challenge.md | AI model validation、independent challenge、SR 11-7、GenAI system validation、evidence |
| 97 | papers/97-ai-third-party-vendor-contract-exit-architecture.md | AI vendor contract、third-party risk、exit architecture、audit rights、model update notice |
| 98 | papers/98-ai-data-lifecycle-governance-provenance-retention.md | AI data lifecycle governance、provenance、retention、deletion、privacy、lineage |
| 99 | papers/99-ai-agent-autonomy-delegation-architecture.md | agent autonomy、delegated authority、human escalation、kill switch、tool authority |
| 100 | papers/100-ai-agent-identity-delegated-authorization.md | agent identity、delegated authorization、OAuth token exchange、scope、audit claims |
| 101 | papers/101-ai-runtime-evidence-observability-architecture.md | runtime evidence、OpenTelemetry、OpenLineage、CloudEvents、audit trail、SLO/KRI |
| 102 | papers/102-ai-portfolio-systemic-risk-dependency-architecture.md | portfolio systemic risk、dependency graph、concentration risk、blast radius、fallback |
| 103 | papers/103-ai-customer-harm-redress-recovery-architecture.md | customer harm、redress、recourse、remediation、recovery、complaints |
| 104 | papers/104-ai-fairness-fair-lending-bias-control-architecture.md | fairness、fair lending、bias control、proxy risk、segment eval、adverse impact |
| 105 | papers/105-ai-explainability-contestability-adverse-action-architecture.md | explainability、contestability、adverse action、reason code、appeal、audit evidence |
| 106 | papers/106-ai-change-impact-release-governance.md | change impact、release governance、model/prompt/RAG/tool changes、regression gate |
| 107 | papers/107-ai-continuous-control-monitoring-assurance-architecture.md | continuous control monitoring、assurance、control test、exception、KRI、effectiveness |
| 108 | papers/108-ai-operational-resilience-bcp-degraded-mode-architecture.md | operational resilience、BCP、degraded mode、critical operations、fallback、recovery exercise |
| 109 | papers/109-ai-management-information-board-reporting-architecture.md | management information、board reporting、metric contract、risk appetite、MI lineage、action log |
| 110 | papers/110-ai-closed-loop-learning-corrective-action-architecture.md | closed-loop learning、CAPA、root cause、corrective action、effectiveness verification |
| 111 | papers/111-ai-regulatory-horizon-obligation-intelligence-architecture.md | regulatory horizon、obligation intelligence、applicability triage、obligation-to-control graph |
| 112 | papers/112-ai-exception-risk-acceptance-waiver-architecture.md | exception management、risk acceptance、waiver、compensating control、expiry、hard stop |
| 113 | papers/113-ai-supply-chain-ai-bom-provenance-architecture.md | AI BOM、provenance、component inventory、model/data/tool/prompt supply chain、vulnerability response |
| 114 | papers/114-ai-human-review-operations-capacity-architecture.md | human review operations、capacity、skill routing、calibration、reviewer quality、surge mode |
| 115 | papers/115-ai-segregation-of-duties-dual-control-architecture.md | segregation of duties、dual control、maker-checker、four-eyes、independent challenge |
| 116 | papers/116-ai-consent-preference-purpose-bound-data-architecture.md | consent、preference、purpose-bound data、withdrawal、re-consent、runtime enforcement |
| 117 | papers/117-ai-shadow-ai-citizen-development-governance-architecture.md | shadow AI、citizen development、approved pathways、risk tiering、platform migration |
| 118 | papers/118-ai-conduct-risk-suitability-sales-guardrails-architecture.md | conduct risk、suitability、Reg BI、approved claims、sales guardrails、surveillance |
| 119 | papers/119-ai-records-retention-legal-hold-ediscovery-architecture.md | records management、retention、legal hold、eDiscovery、ESI、regulator production |
| 120 | papers/120-ai-data-residency-cross-border-sovereign-architecture.md | data residency、cross-border transfer、sovereign AI、region controls、key residency |
| 121 | papers/121-ai-customer-communications-regulated-content-lifecycle.md | customer communications、regulated content lifecycle、approved claims、disclosure、surveillance |
| 122 | papers/122-ai-financial-crime-typology-scenario-coverage-architecture.md | financial crime typology、scenario library、coverage matrix、SAR evidence、red flags |
| 123 | papers/123-ai-intellectual-property-content-rights-provenance-architecture.md | intellectual property、content rights、copyrightability、C2PA、provenance、rights clearance |
| 124 | papers/124-ai-deepfake-synthetic-identity-authentication-fraud-architecture.md | deepfake、synthetic identity、liveness、presentation attack、step-up authentication、fraud ops |
| 125 | papers/125-ai-workforce-hr-decision-employee-monitoring-governance-architecture.md | workforce AI、HR decision、employee monitoring、adverse impact、worker data minimization |
| 126 | papers/126-ai-incident-disclosure-liability-risk-transfer-architecture.md | AI incident、materiality triage、liability boundary、insurance mapping、risk transfer |
| 127 | papers/127-ai-post-quantum-cryptographic-agility-ai-architecture.md | post-quantum cryptography、cryptographic agility、crypto inventory、key rotation、long-lived evidence |
| 128 | papers/128-ai-authorized-push-payment-scam-intervention-architecture.md | authorized push payment scam、social engineering、beneficiary risk、step-up friction、customer intent |
| 129 | papers/129-ai-agent-marketplace-tool-certification-governance-architecture.md | agent marketplace、tool certification、capability card、risk tier、signed package、runtime permission |
| 130 | papers/130-ai-customer-vulnerability-accessibility-inclusive-ai-architecture.md | vulnerable customer、accessibility、inclusive AI、WCAG、plain language、safe escalation |
| 131 | papers/131-ai-payment-dispute-chargeback-claims-evidence-architecture.md | payment dispute、chargeback、claims evidence、provisional credit、case routing、Reg E/Reg Z |
| 132 | papers/132-ai-collections-hardship-delinquency-treatment-architecture.md | collections、hardship、delinquency、fair treatment、contact strategy、Reg F、complaints |
| 133 | papers/133-ai-voice-ai-contact-center-agent-assist-governance-architecture.md | voice AI、contact center、agent assist、call summary、QA automation、FCC/TSR、complaint linkage |
| 134 | papers/134-ai-digital-identity-wallet-verifiable-credentials-trust-architecture.md | digital identity wallet、verifiable credentials、DID、WebAuthn、selective disclosure、trust framework |
| 135 | papers/135-ai-open-banking-open-finance-consented-data-sharing-architecture.md | open banking、open finance、consented data sharing、CFPB 1033、API ecosystem、revocation |
| 136 | papers/136-ai-personalized-pricing-offer-decisioning-governance-architecture.md | personalized pricing、offer decisioning、surveillance pricing、feature boundary、fairness、complaints |
| 137 | papers/137-ai-document-intelligence-unstructured-data-evidence-quality-architecture.md | document intelligence、OCR、layout extraction、evidence quality、records、human review |
| 138 | papers/138-ai-privacy-clean-room-data-collaboration-measurement-architecture.md | privacy clean room、data collaboration、measurement、PEC、de-identification、partner governance |
| 139 | papers/139-ai-credit-lifecycle-underwriting-line-management-governance-architecture.md | credit lifecycle、underwriting、line management、adverse action、portfolio monitoring、model risk |
| 140 | papers/140-ai-wealth-advice-robo-advisor-best-interest-boundary-architecture.md | wealth advice、robo-advisor、best interest、advice boundary、risk profile、human escalation |
| 141 | papers/141-ai-treasury-liquidity-alm-forecasting-stress-evidence-architecture.md | treasury、liquidity forecasting、ALM、stress testing、deposit runoff、board MI |
| 142 | papers/142-ai-complaint-intelligence-root-cause-regulatory-response-architecture.md | complaint intelligence、root cause、regulatory response、CAPA、remediation、board reporting |
| 143 | papers/143-ai-account-opening-kyc-onboarding-decision-architecture.md | account opening、KYC、CIP/CDD、identity proofing、risk-tiering、exception queue |
| 144 | papers/144-ai-aml-alert-triage-investigation-workbench-architecture.md | AML alert triage、investigation workbench、case assembly、analyst copilot、QA、audit trail |
| 145 | papers/145-ai-payment-operations-reconciliation-settlement-exception-architecture.md | payment operations、reconciliation、settlement exception、repair queue、dual control、SLA |
| 146 | papers/146-ai-regulatory-reporting-risk-data-aggregation-attestation-architecture.md | regulatory reporting、risk data aggregation、BCBS 239、Call Report、FR Y-14、attestation |
| 147 | papers/147-ai-core-banking-modernization-strangler-evidence-architecture.md | core banking modernization、legacy strangler、ACL、dual-run、migration evidence、cutover |
| 148 | papers/148-ai-requirements-mining-process-knowledge-extraction-architecture.md | requirements mining、process knowledge extraction、traceability graph、SME validation、eval contract |
| 149 | papers/149-ai-uat-regression-certification-business-acceptance-architecture.md | UAT、regression certification、business acceptance、golden journeys、release evidence、rollback |
| 150 | papers/150-ai-product-portfolio-rationalization-capability-decommission-architecture.md | product portfolio rationalization、capability decommission、sunset、dependency graph、migration cohorts |
| 151 | papers/151-ai-adoption-analytics-behavior-change-value-realization-architecture.md | adoption analytics、behavior change、value realization、work-as-done baseline、cohort evidence |
| 152 | papers/152-ai-executive-investment-narrative-business-case-board-decision-architecture.md | executive investment narrative、business case、board decision、option architecture、stop/scale/pivot |
| 153 | papers/153-ai-stakeholder-decision-coalition-influence-risk-architecture.md | stakeholder coalition、decision rights、influence risk、concern-to-viewpoint mapping、forum architecture |
| 154 | papers/154-ai-workforce-capability-academy-role-skill-transformation-architecture.md | workforce capability academy、role taxonomy、skill ontology、evidence-of-competence、skills debt |
| 155 | papers/155-ai-delivery-assurance-control-tower-release-readiness-architecture.md | delivery assurance、control tower、release readiness、evidence contract、scale readiness |
| 156 | papers/156-ai-product-operations-operating-cadence-outcome-review-architecture.md | product operations、operating cadence、outcome review、metric contract、incident-to-roadmap loop |
| 157 | papers/157-ai-reference-implementation-pattern-library-reuse-assurance-architecture.md | reference implementation、pattern library、reuse assurance、approved deviations、reusable evidence |
| 158 | papers/158-ai-technology-radar-scenario-planning-strategic-foresight-architecture.md | technology radar、scenario planning、strategic foresight、assumption ledger、option portfolio |
| 159 | papers/159-ai-product-responsibility-accountability-boundary-architecture.md | product responsibility、accountability boundary、decision ownership、product promise、recourse |
| 160 | papers/160-ai-context-supply-chain-provenance-poisoning-defense-architecture.md | context supply chain、provenance、poisoning defense、context trust tier、runtime trace |
| 161 | papers/161-ai-agentic-process-audit-workflow-replay-assurance-architecture.md | agentic process audit、workflow replay、event schema、assurance evidence、incident replay |
| 162 | papers/162-ai-organizational-memory-knowledge-asset-governance-architecture.md | organizational memory、knowledge assets、decision memory、retention/forgetting、knowledge debt |
| 163 | papers/163-ai-eval-dataset-lifecycle-golden-set-test-data-factory-architecture.md | eval dataset lifecycle、golden set、test data factory、coverage drift、evidence |
| 164 | papers/164-ai-synthetic-user-simulation-persona-scenario-lab-architecture.md | synthetic user、persona scenario lab、behavior simulation、journey testbed、scenario governance |
| 165 | papers/165-ai-human-factors-operations-cognitive-load-automation-bias-architecture.md | human factors operations、cognitive load、automation bias、calibrated trust、operator burden |
| 166 | papers/166-ai-procurement-intake-vendor-evaluation-sandbox-build-buy-architecture.md | procurement intake、vendor sandbox、build-buy decision、due diligence evidence、exit constraints |
| 167 | papers/167-ai-reasoning-budget-test-time-compute-verifier-cascade-architecture.md | reasoning budget、test-time compute、verifier cascade、deliberation policy、cost-latency-risk |
| 168 | papers/168-ai-model-portfolio-benchmarking-capability-scorecard-selection-governance-architecture.md | model portfolio benchmarking、capability scorecard、selection governance、champion/challenger、retirement trigger |
| 169 | papers/169-ai-uncertainty-ux-abstention-confidence-escalation-architecture.md | uncertainty UX、abstention、confidence language、escalation policy、customer harm prevention |
| 170 | papers/170-ai-risk-quantification-scenario-loss-control-roi-architecture.md | AI risk quantification、scenario loss、control ROI、residual risk、management action |
| 171 | papers/171-ai-evidence-integrity-tamper-evident-audit-ledger-attestation-architecture.md | evidence integrity、tamper-evident audit ledger、attestation chain、hash chain、audit query |
| 172 | papers/172-ai-shadow-mode-counterfactual-evaluation-silent-launch-architecture.md | shadow mode、counterfactual evaluation、silent launch、outcome delay、rollout gate |
| 173 | papers/173-ai-synthetic-data-governance-privacy-utility-fidelity-architecture.md | synthetic data governance、privacy-utility-fidelity、privacy attack testing、allowed use、provenance |
| 174 | papers/174-ai-three-lines-governance-decision-rights-assurance-operating-model-architecture.md | three lines governance、decision rights、assurance forum、issue management、lifecycle gate |
Related Existing Assets
| Asset | Why it matters |
|---|---|
../AI_ENGINEERING_METHODOLOGY.md | 生产级 AI 系统工程方法论，适合把论文概念落到工程闭环 |
../AGENTIC_RAG_2026.md | 2026 Agentic RAG 实战与本项目改造清单 |
../AI_NEW_DEMANDS_2026_EXPANSION.md | 企业 AI 新需求总纲，解释为什么底层能力要服务 operating capability |
../AGENTIC_ENTERPRISE_ARCHITECTURE_90_PLAN.md | Agentic Enterprise Architecture 90 天架构训练 |
../AI_GOVERNANCE_EVALOPS_RISK_90_PLAN.md | AI Governance / EvalOps / RiskOps 训练线 |
../AI_BA_PM_PRACTICE_LAB.md | 把底层概念转成产品架构练习和金融场景 case drill |
../AI_ARCHITECTURE_DIAGRAM_PLAYBOOK.md | 把论文机制转成 Capability、C4、RAG、Agent、Eval、Risk 和 Cost/Latency 图谱 |
../AI_EXPANSION_MASTER_INDEX.md | 把论文精读、case drill、架构、治理、平台和能力证据串成长期地图 |
../AI_CASE_DRILL_WORKBOOK_30_DAYS.md | 把论文机制放进 30 个金融零售 AI case drill 中训练 |
../AI_PLATFORM_PM_PLAYBOOK.md | 把模型、RAG、Agent、EvalOps、成本和治理抽象成企业 AI 平台能力 |
../AI_REGULATORY_RESPONSE_PLAYBOOK.md | 把 AI 法规、监管预期和行业框架转成 inventory、control、evidence、incident response |
../AI_DATA_PRODUCT_MANAGEMENT_PLAYBOOK.md | 把 RAG、eval、feedback、labels 和 governance 需要的数据能力产品化 |
../AI_BOARD_AUDIT_COMMITTEE_GOVERNANCE_PACK.md | 把 AI 风险、控制有效性、剩余风险和投资决策翻译成董事会/审计委员会语言 |
../AI_CAPABILITY_ASSESSMENT_RUBRIC.md | 用评分体系把论文、case、架构和治理资产转成可复盘的职业能力证据 |
../AI_RETRIEVAL_EVAL_GRAPH_RAG_PLAYBOOK.md | 把 RAG/GraphRAG 论文机制转成 retrieval eval、release gate、GraphRAG ADR 和金融零售落地证据 |
../AI_PLATFORM_SECURITY_GATEWAY_LAB.md | 把 prompt injection、tool gateway、权限、审计、kill switch 转成平台安全网关实验 |
../AI_REGULATOR_EXAM_SIMULATION_PACK.md | 把监管、内审、模型风险问询转成 evidence pack、exam Q&A 和 remediation plan |
../AI_ADVANCED_CASE_DRILL_WORKBOOK_60_DAYS.md | 把 30 天 case drill 升级成 60 天复杂案例、架构评审、治理和证据化训练 |
../AI_MEMORY_CONTEXT_STATE_PLAYBOOK.md | 把 memory、context、workflow state、retention、deletion、privacy eval 转成生产级能力 |
../AI_MULTI_AGENT_ORCHESTRATION_PLAYBOOK.md | 把 multi-agent role、handoff、shared state、policy supervisor 和 HITL 转成可评审工作流 |
../AI_OBSERVABILITY_COST_SLO_PLAYBOOK.md | 把 GenAI trace、latency、quality、cost、SLO、incident 和 ROI 变成运营控制面 |
../AI_AGENT_PROTOCOLS_MCP_A2A_PLAYBOOK.md | 把 MCP、A2A、tool contract、capability discovery、auth、audit 和 vendor integration 变成平台集成能力 |
../AI_ASSURANCE_SAFETY_CASE_PLAYBOOK.md | 把 AI assurance、safety case、claim-argument-evidence 和 release confidence 转成上线证据 |
../AI_MODEL_RISK_MANAGEMENT_PLAYBOOK.md | 把 SR 11-7 / 模型风险管理迁移到 GenAI system risk、validation、change control 和 monitoring |
../AI_SYNTHETIC_EVAL_DATA_PLAYBOOK.md | 把合成评测数据、质量标准、覆盖矩阵和金融零售场景库变成 EvalOps 数据资产 |
../AI_TRANSFORMATION_VALUE_OFFICE_PLAYBOOK.md | 把 AI use case portfolio、funding gate、benefits realization 和 scale/stop 机制变成价值管理能力 |
../AI_HUMAN_OVERSIGHT_HITL_PLAYBOOK.md | 把 human oversight、HITL、handoff、override、kill switch 和 AI literacy 变成可设计工作流 |
../AI_THREAT_MODELING_RED_TEAM_PLAYBOOK.md | 把 LLM/RAG/Agent threat modeling、OWASP、MITRE ATLAS/SAFE-AI 和 red-team eval 变成上线安全证据 |
../AI_AUDIT_EVIDENCE_BINDER_PLAYBOOK.md | 把 control evidence、audit binder、model/system cards、dataset cards 和 regulator-ready documentation 组织成证据库 |
../AI_ADOPTION_CHANGE_MANAGEMENT_PLAYBOOK.md | 把 AI literacy、角色重设计、培训、反馈、采用指标和收益实现连接成组织落地能力 |
../AI_PRIVACY_DATA_PROTECTION_PLAYBOOK.md | 把 privacy-by-design、PII、retention/deletion、DPIA/PIA、prompt/RAG/memory/log 隐私控制变成 AI 架构要求 |
../AI_THIRD_PARTY_VENDOR_RISK_PLAYBOOK.md | 把 AI vendor due diligence、contract clauses、model update notice、audit rights、exit plan 和 concentration risk 变成第三方风险能力 |
../AI_PROCESS_MINING_WORKFLOW_INTELLIGENCE_PLAYBOOK.md | 把 process mining、task mining、event log、variant/bottleneck analysis 和 ROI baseline 变成 AI BA 机会发现能力 |
../AI_CUSTOMER_FACING_REGULATED_PRODUCT_PLAYBOOK.md | 把 customer-facing AI 的 disclosure、advice boundary、complaints、escalation、fair lending 和 monitoring 变成产品设计门禁 |
../AI_KNOWLEDGE_GOVERNANCE_ONTOLOGY_PLAYBOOK.md | 把 taxonomy、ontology、knowledge graph、source authority、freshness、permission 和 evidence lineage 变成 AI 知识治理能力 |
../AI_SEMANTIC_LAYER_METRICS_ARCHITECTURE_PLAYBOOK.md | 把 semantic layer、metric contract、lineage、LLM-to-SQL guardrails 和 AI value metrics 变成指标架构能力 |
../AI_INCIDENT_POSTMORTEM_RELIABILITY_PLAYBOOK.md | 把 AI incident、severity、containment、rollback、postmortem、corrective action 和 reliability review 变成生产运营能力 |
../AI_PRODUCT_ARCHITECTURE_STRATEGY_PLAYBOOK.md | 把 AI 产品战略、能力地图、平台/点方案、architecture runway、funding gate 和 scale/stop 决策变成高级架构叙事 |
../AI_EVALOPS_PLATFORM_ARCHITECTURE_PLAYBOOK.md | 把 eval dataset、judge、human review、release gate、production monitoring 和 evidence binder 变成平台能力 |
../AI_DECISION_INTELLIGENCE_CAUSAL_PRODUCT_PLAYBOOK.md | 把因果推断、uplift、实验、准实验、ROI attribution 和 funding gate 变成 AI 产品决策能力 |
../AI_ENTERPRISE_INTEGRATION_EVENT_DRIVEN_AGENT_PLAYBOOK.md | 把 API、事件、workflow、CloudEvents、AsyncAPI、tool contract、idempotency 和 HITL 队列变成 Agent 集成架构 |
../AI_TRUST_EXPERIENCE_PRODUCT_GOVERNANCE_PLAYBOOK.md | 把 trust calibration、透明度、拒答、升级、解释、投诉/申诉和过度依赖防控变成 AI 产品体验治理 |
../AI_POLICY_AS_CODE_DECISION_AUTOMATION_PLAYBOOK.md | 把 DMN、OPA、Cedar、Zanzibar、PDP/PEP、决策自动化和审计证据变成 AI runtime 控制能力 |
../AI_PERSONALIZATION_RECOMMENDER_PRODUCT_ARCHITECTURE_PLAYBOOK.md | 把推荐系统、个性化、next-best-action、适用性、指标和反馈闭环变成 AI 产品架构能力 |
../AI_REAL_TIME_FEATURE_STORE_DECISIONING_PLAYBOOK.md | 把 feature store、实时特征、freshness SLO、训练-服务一致性和实时决策变成平台能力 |
../AI_ENGINEERING_PRODUCTIVITY_CODE_AGENT_OPERATING_SYSTEM_PLAYBOOK.md | 把 code agent、DORA/SPACE、AI SDLC、coding eval、权限和上线门禁变成工程效率产品能力 |
../AI_PRIVACY_ENHANCING_TECH_CONFIDENTIAL_AI_PLAYBOOK.md | 把差分隐私、联邦学习、TEE、FHE、secure aggregation、clean room 和 privacy budget 变成 Confidential AI 架构能力 |
../AI_DURABLE_AGENT_WORKFLOW_STATE_MACHINE_PLAYBOOK.md | 把 durable execution、状态机、Saga、幂等、HITL、replay 和审计转成 Agent 工作流架构能力 |
../AI_FINOPS_UNIT_ECONOMICS_CAPACITY_PLAYBOOK.md | 把 token/GPU/case 成本、capacity、routing/cache、showback/chargeback 和 budget guardrail 变成 AI FinOps 能力 |
../AI_FRONTIER_MODEL_STRATEGY_DISTILLATION_SMALL_MODELS_PLAYBOOK.md | 把 frontier model、小模型、蒸馏、量化、路由和模型组合治理变成 AI 平台策略能力 |
../AI_DIGITAL_TWIN_SIMULATION_PRODUCT_ARCHITECTURE_PLAYBOOK.md | 把 digital twin、仿真、scenario library、calibration、validation 和 policy simulation 变成 AI 决策实验能力 |
../AI_EXPERIMENTATION_PLATFORM_RELEASE_SCIENCE_PLAYBOOK.md | 把 A/B、CUPED、shadow launch、ramp、guardrail 和 post-experiment decision 变成 AI 发布科学能力 |
../AI_DATA_CONTRACTS_LINEAGE_QUALITY_PLAYBOOK.md | 把 data contract、lineage、metadata、quality SLO、eval/RAG/training 数据血缘变成 AI 数据产品能力 |
../AI_SECURITY_OPERATIONS_SOC_PLAYBOOK.md | 把 MITRE ATLAS、OWASP、AI telemetry、SIEM/SOAR、incident runbook 和 purple team 变成 AI SecOps 能力 |
../AI_FORECASTING_DEMAND_PLANNING_PRODUCT_ARCHITECTURE_PLAYBOOK.md | 把 time-series forecasting、demand planning、capacity、现金流、预测区间和 forecast governance 变成产品架构能力 |
../AI_ANOMALY_DETECTION_RISK_MONITORING_PLAYBOOK.md | 把 anomaly detection、Isolation Forest、autoencoder、告警治理、阈值校准和风险监控变成运营控制面 |
../AI_CAUSAL_DISCOVERY_STRUCTURAL_DECISION_PLAYBOOK.md | 把 causal discovery、DAG、SCM、DoWhy/EconML、识别假设和干预设计变成结构化产品决策能力 |
../AI_OPTIMIZATION_OPERATIONS_RESEARCH_DECISION_PLAYBOOK.md | 把 OR、LP/MIP、CP-SAT、solver、约束、目标函数和例外工作流变成 AI 决策服务架构 |
../AI_CONTEXTUAL_BANDITS_ADAPTIVE_EXPERIMENTATION_PLAYBOOK.md | 把 contextual bandits、adaptive experimentation、propensity logging、OPE 和 next-best-action 变成在线学习决策能力 |
../AI_REINFORCEMENT_LEARNING_POLICY_DECISION_PLAYBOOK.md | 把 RL、MDP、offline RL、reward design、safe exploration 和 policy guardrails 变成序列决策能力 |
../AI_BAYESIAN_OPTIMIZATION_EXPERIMENT_DESIGN_PLAYBOOK.md | 把 Bayesian optimization、BoTorch、Optuna、实验预算、多目标/约束优化变成 AI 平台调优能力 |
../AI_UNCERTAINTY_CALIBRATION_CONFORMAL_PREDICTION_PLAYBOOK.md | 把 calibration、conformal prediction、coverage、abstention 和 confidence UX 变成不确定性治理能力 |
../AI_PROGRAMMATIC_LABELING_DATA_CENTRIC_AI_PLAYBOOK.md | 把 Snorkel、labeling functions、SME workflow、label quality 和弱监督变成数据中心 AI 平台能力 |
../AI_ACTIVE_LEARNING_HUMAN_FEEDBACK_OPERATIONS_PLAYBOOK.md | 把 active learning、query strategy、HITL 队列、reviewer calibration 和反馈运营变成持续学习能力 |
../AI_DATASET_SHIFT_MONITORING_MODEL_PERFORMANCE_PLAYBOOK.md | 把 dataset shift、score/feature/embedding drift、outcome lag 和 alert runbook 变成模型性能运营能力 |
../AI_MANAGEMENT_SYSTEM_ISO42001_OPERATING_MODEL_PLAYBOOK.md | 把 ISO 42001、NIST AI RMF、AI inventory、release gate、control library 和 management review 变成 AI 管理体系 |
../AI_ML_TECHNICAL_DEBT_ARCHITECTURE_PLAYBOOK.md | 把 CACE、entanglement、data/config debt、hidden feedback loop 和 consumer registry 变成 AI 架构债务治理能力 |
../AI_MLOPS_CONTINUOUS_DELIVERY_RELEASE_PLAYBOOK.md | 把 CD4ML、release bundle、CI/CD/CT、shadow/canary/ramp、rollback 和 evidence binder 变成 AI 发布工程能力 |
../AI_HUMAN_AI_INTERACTION_PRODUCT_DESIGN_PLAYBOOK.md | 把 Human-AI Interaction、calibrated trust、automation bias、recoverability、feedback 和 escalation 变成 AI 产品设计能力 |
../AI_ARCHITECTURE_DECISION_RECORDS_GOVERNANCE_PLAYBOOK.md | 把 AI ADR、risk tier、evidence link、reversal trigger、review workflow 和 architecture knowledge 变成决策治理能力 |
../AI_REQUIREMENTS_ENGINEERING_GQM_EVAL_CONTRACTS_PLAYBOOK.md | 把 GQM、AI requirement taxonomy、eval contract、release gate 和 monitoring gate 变成高级需求工程能力 |
../AI_QUALITY_ATTRIBUTES_ATAM_TRADEOFF_PLAYBOOK.md | 把 ATAM、quality attribute scenario、utility tree、tradeoff/sensitivity/risk 变成 AI 架构评审能力 |
../AI_SAFETY_ENGINEERING_STPA_PLAYBOOK.md | 把 STPA、control structure、unsafe control action、safety constraint 和 circuit breaker 变成 AI 安全工程能力 |
../AI_SOCIO_TECHNICAL_RESILIENCE_OPERATING_MODEL_PLAYBOOK.md | 把 work-as-done、human-AI collaboration、handoff、load、resilience 和 operating model 变成社会技术系统设计能力 |
../AI_CAPABILITY_BASED_PLANNING_BUSINESS_ARCHITECTURE_PLAYBOOK.md | 把 capability-based planning、value stream、capability maturity、portfolio prioritization 和 architecture runway 变成企业 AI 业务架构能力 |
../AI_WARDLEY_MAPPING_PRODUCT_STRATEGY_PLAYBOOK.md | 把 Wardley Mapping、AI value chain、evolution axis、build-buy-partner 和 platform boundary 变成 AI 产品战略能力 |
../AI_TEAM_TOPOLOGIES_CONWAY_PLATFORM_OPERATING_MODEL_PLAYBOOK.md | 把 Conway's Law、Team Topologies、cognitive load、team API 和 interaction mode 变成 AI 平台组织设计能力 |
../AI_DORA_SPACE_ENGINEERING_PRODUCTIVITY_SDLC_PLAYBOOK.md | 把 DORA、SPACE、AI code agent governance、PR/eval/release gate 和 DevEx 变成 AI SDLC 生产力治理能力 |
../AI_CONTINUOUS_DISCOVERY_OPPORTUNITY_SOLUTION_TREE_PLAYBOOK.md | 把 Continuous Discovery、Opportunity Solution Tree、assumption map、eval/pilot learning 变成 AI 产品发现能力 |
../AI_JTBD_OUTCOME_DRIVEN_INNOVATION_PLAYBOOK.md | 把 JTBD、ODI、outcome statement、underserved outcome 和 automation boundary 变成 AI use case 选择能力 |
../AI_PRODUCT_METRICS_NORTH_STAR_VALUE_MEASUREMENT_PLAYBOOK.md | 把 North Star、input metrics、guardrail metrics、risk-adjusted value 和因果证据变成 AI 价值度量能力 |
../AI_PRODUCT_OPERATING_MODEL_EMPOWERED_TEAMS_PLAYBOOK.md | 把 empowered teams、product trio+、decision rights 和 discovery-delivery-governance cadence 变成 AI 产品运营模型能力 |
../AI_PORTFOLIO_MANAGEMENT_FUNDING_GOVERNANCE_PLAYBOOK.md | 把 AI portfolio kanban、funding guardrails、capacity allocation、benefits realization 和 scale/stop 决策变成组合治理能力 |
../AI_SERVICE_BLUEPRINT_CUSTOMER_JOURNEY_TRUST_PLAYBOOK.md | 把 service blueprint、customer journey、trust moments、handoff、appeal 和运营证据变成 AI 服务体验治理能力 |
../AI_BUSINESS_PROCESS_REENGINEERING_BPMN_DMN_PLAYBOOK.md | 把 AI BPR、BPMN、DMN、decision service、eval traceability 和 control evidence 变成流程重构能力 |
../AI_RISK_APPETITE_POLICY_PRODUCT_MANAGEMENT_PLAYBOOK.md | 把 risk appetite、policy lifecycle、risk-tiered guardrails、runtime control 和 stop rule 变成 AI 策略产品能力 |
../AI_ENTERPRISE_REFERENCE_ARCHITECTURE_CONTROL_PLANE_PLAYBOOK.md | 把 enterprise AI reference architecture、control plane、model/tool gateway、eval/observability 和 evidence plane 变成企业架构能力 |
../AI_PRODUCT_LINE_ENGINEERING_REUSABLE_PLATFORM_ASSETS_PLAYBOOK.md | 把 product line engineering、core assets、variation points、asset governance 和 reuse ROI 变成 AI 平台复用能力 |
../AI_MATURITY_MODEL_ROADMAP_CAPABILITY_ASSESSMENT_PLAYBOOK.md | 把 AI maturity model、capability heatmap、roadmap dependency 和个人/组织证据标准变成转型路线能力 |
../AI_CONTROL_LIBRARY_ASSURANCE_EVIDENCE_GRAPH_PLAYBOOK.md | 把 control library、assurance case、claim-risk-control-evidence graph 和监管问询证据变成 AI 保证能力 |
../AI_DOMAIN_DRIVEN_DESIGN_UBIQUITOUS_LANGUAGE_PLAYBOOK.md | 把 DDD、bounded context、ubiquitous language、RAG boundary 和 eval vocabulary 变成 AI 语义边界能力 |
../AI_EVENT_STORMING_AGENT_WORKFLOW_DISCOVERY_PLAYBOOK.md | 把 EventStorming、domain events、commands、policies、AI insertion point 和 compensation 转成 Agent workflow discovery 能力 |
../AI_KNOWLEDGE_WORK_REDESIGN_ROLE_TASK_ARCHITECTURE_PLAYBOOK.md | 把知识工作拆成 role、task、decision、artifact、handoff、control 和 metric, 形成 AI 工作重设计能力 |
../AI_PLATFORM_SERVICE_CATALOG_GOLDEN_PATHS_PLAYBOOK.md | 把 AI 平台 service catalog、golden paths、self-service guardrails 和平台指标变成平台产品能力 |
../AI_ARCHITECTURE_VIEWS_C4_ARC42_42010_PLAYBOOK.md | 把 C4、arc42、ISO 42010 转成 AI 架构多视图、stakeholder concern 和评审证据 |
../AI_ARCHITECTURE_FITNESS_FUNCTIONS_CONTINUOUS_GOVERNANCE_PLAYBOOK.md | 把质量属性、风险控制、eval、SLO 和证据转成持续可检查的 architecture fitness functions |
../AI_CONTRACT_FIRST_TOOL_API_DESIGN_OPENAPI_ASYNCAPI_PLAYBOOK.md | 把 OpenAPI、AsyncAPI、JSON Schema、tool contract、event contract 和审计证据转成 Agent 集成治理 |
../AI_TRACEABILITY_REQUIREMENTS_EVAL_CONTROL_GRAPH_PLAYBOOK.md | 把需求、评测、控制、ADR、运行 trace 和 evidence binder 连接成 AI traceability graph |
../AI_ENTERPRISE_ARCHITECTURE_TOGAF_ARCHIMATE_ADM_PLAYBOOK.md | 把 TOGAF ADM、ArchiMate、architecture repository 和 AI governance 转成企业 AI 架构方法 |
../AI_BANKING_REFERENCE_MODELS_BIAN_FIBO_ISO20022_PLAYBOOK.md | 把 BIAN、FIBO、ISO 20022 转成银行 AI 能力边界、金融语义和集成证据 |
../AI_SEMANTIC_INTEROPERABILITY_RDF_OWL_SHACL_PLAYBOOK.md | 把 RDF、OWL、SHACL、JSON Schema 和 FIBO 转成 AI 语义互操作与语义约束能力 |
../AI_VALUE_STREAM_MANAGEMENT_FLOW_METRICS_PLAYBOOK.md | 把 AI use case 从 idea 到 safe release、adoption 和 value realization 的 flow 管起来 |
../AI_REGULATORY_ARCHITECTURE_EU_AI_ACT_NIST_ISO42001_PLAYBOOK.md | 把法规、框架和管理体系要求转成 AI inventory、risk tier、control、gate 和 evidence architecture |
../AI_MODEL_VALIDATION_INDEPENDENT_CHALLENGE_PLAYBOOK.md | 把传统模型验证升级成 GenAI system validation、independent challenge 和 revalidation evidence |
../AI_THIRD_PARTY_VENDOR_CONTRACT_EXIT_ARCHITECTURE_PLAYBOOK.md | 把 AI vendor 采购、合同条款、变更通知、审计权和退出计划转成产品/架构控制 |
../AI_DATA_LIFECYCLE_GOVERNANCE_PROVENANCE_RETENTION_PLAYBOOK.md | 把训练、RAG、eval、prompt、memory 和日志数据的来源、保留、删除、血缘转成治理能力 |
../AI_AGENT_AUTONOMY_DELEGATION_ARCHITECTURE_PLAYBOOK.md | 把 agent 自主性拆成委派权力、任务边界、工具权限、升级路径、kill switch 和证据链 |
../AI_AGENT_IDENTITY_DELEGATED_AUTHORIZATION_PLAYBOOK.md | 把 agent 身份、OAuth token exchange、scope、consent、step-up approval 和审计 claims 设计成授权架构 |
../AI_RUNTIME_EVIDENCE_OBSERVABILITY_ARCHITECTURE_PLAYBOOK.md | 把 prompt、RAG、tool、approval、policy、cost、quality 和 incident 串成 runtime evidence architecture |
../AI_PORTFOLIO_SYSTEMIC_RISK_DEPENDENCY_ARCHITECTURE_PLAYBOOK.md | 把 AI portfolio 的模型、供应商、知识源、工具、身份、人工队列和证据栈依赖做成系统性风险管理 |
../AI_CUSTOMER_HARM_RECOURSE_REMEDIATION_PLAYBOOK.md | 把客户伤害、投诉/申诉、救济、纠正、补偿、恢复和防复发转成 AI 产品控制 |
../AI_FAIRNESS_FAIR_LENDING_BIAS_CONTROL_PLAYBOOK.md | 把公平性/公平信贷、proxy 风险、segment eval、human review calibration 和 evidence binder 转成架构控制 |
../AI_EXPLAINABILITY_CONTESTABILITY_ADVERSE_ACTION_PLAYBOOK.md | 把用户可见解释、reason code、adverse action、contestability 和申诉证据转成治理接口 |
../AI_CHANGE_IMPACT_RELEASE_GOVERNANCE_PLAYBOOK.md | 把 model/prompt/RAG/tool/policy/eval/vendor/workflow 变更转成 impact graph、regression gate 和 release evidence |
../AI_CONTINUOUS_CONTROL_MONITORING_ASSURANCE_PLAYBOOK.md | 把 control library、eval、telemetry、exceptions、sampling、KRI 和 management action 转成持续控制有效性监控 |
../AI_OPERATIONAL_RESILIENCE_BCP_DEGRADED_MODE_PLAYBOOK.md | 把模型/RAG/工具/身份/策略/人工队列/供应商故障转成 degraded mode、BCP、RTO/RPO 和演练证据 |
../AI_MANAGEMENT_INFORMATION_BOARD_REPORTING_PLAYBOOK.md | 把 AI 价值、风险、控制、事故、客户伤害、集中度和采用转成有 lineage 的管理信息和董事会报告 |
../AI_CLOSED_LOOP_LEARNING_CORRECTIVE_ACTION_PLAYBOOK.md | 把反馈、投诉、人工覆盖、eval 失败、漂移、事件和审计发现转成 CAPA 式纠正行动闭环 |
../AI_REGULATORY_HORIZON_OBLIGATION_INTELLIGENCE_PLAYBOOK.md | 把法律、监管指引、监督重点、标准和行业信号转成 obligation-to-control/eval/change 情报系统 |
../AI_EXCEPTION_RISK_ACCEPTANCE_WAIVER_PLAYBOOK.md | 把策略例外、临时豁免、残余风险接受、补偿控制、到期续期和 hard stop 做成治理能力 |
../AI_SUPPLY_CHAIN_AI_BOM_PROVENANCE_PLAYBOOK.md | 把模型、供应商、数据集、RAG 语料、prompt、tool、MCP、eval 和人工服务做成 AI BOM 和 provenance graph |
../AI_HUMAN_REVIEW_OPERATIONS_CAPACITY_PLAYBOOK.md | 把人工审核设计成队列、技能路由、容量、校准、质量、升级、surge mode 和证据运营系统 |
../AI_SEGREGATION_OF_DUTIES_DUAL_CONTROL_PLAYBOOK.md | 把 maker-checker、four-eyes、dual control、incompatible duties、approval-before-action 和审计证据做成 AI 工作流控制 |
../AI_CONSENT_PREFERENCE_PURPOSE_BOUND_DATA_PLAYBOOK.md | 把客户同意、偏好、目的限制、撤回、重新同意和 runtime enforcement 做成 AI 数据使用控制 |
../AI_SHADOW_AI_CITIZEN_DEVELOPMENT_GOVERNANCE_PLAYBOOK.md | 把未授权 AI 使用和公民开发转成发现、分级、批准路径、平台迁移和安全采用机制 |
../AI_CONDUCT_RISK_SUITABILITY_SALES_GUARDRAILS_PLAYBOOK.md | 把金融销售行为、适当性、Reg BI、approved claims、offer guardrails、投诉和整改转成 AI 产品护栏 |
../AI_RECORDS_RETENTION_LEGAL_HOLD_EDISCOVERY_PLAYBOOK.md | 把 prompt、RAG、tool、approval、output、eval、incident 记录转成 retention、legal hold 和 eDiscovery 能力 |
../AI_DATA_RESIDENCY_CROSS_BORDER_SOVEREIGN_AI_PLAYBOOK.md | 把地域、司法辖区、供应商、模型路由、日志、密钥和 transfer review 做成跨境 AI 数据架构 |
../AI_CUSTOMER_COMMUNICATIONS_REGULATED_CONTENT_LIFECYCLE_PLAYBOOK.md | 把 approved claims、forbidden claims、pre-use review、post-use surveillance、disclosure 和投诉链接做成内容生命周期 |
../AI_FINANCIAL_CRIME_TYPOLOGY_SCENARIO_COVERAGE_PLAYBOOK.md | 把 AML typology、red flag、scenario coverage、SAR evidence bundle 和 alert-to-SAR traceability 做成可评测能力 |
../AI_INTELLECTUAL_PROPERTY_CONTENT_RIGHTS_PROVENANCE_PLAYBOOK.md | 把输入权利、RAG 语料许可、生成内容、C2PA provenance、rights clearance 和 takedown 做成内容权利架构 |
../AI_DEEPFAKE_SYNTHETIC_IDENTITY_AUTHENTICATION_FRAUD_PLAYBOOK.md | 把 deepfake、synthetic identity、liveness/PAD、step-up authentication、fraud evidence 和客户摩擦做成认证欺诈防线 |
../AI_WORKFORCE_HR_DECISION_EMPLOYEE_MONITORING_GOVERNANCE_PLAYBOOK.md | 把招聘、排班、绩效、QA、员工监控、adverse impact 和 worker data minimization 做成 workforce AI 治理 |
../AI_INCIDENT_DISCLOSURE_LIABILITY_RISK_TRANSFER_PLAYBOOK.md | 把 AI incident、materiality triage、通知、责任边界、保险映射、供应商赔偿和损失量化做成高管决策包 |
../AI_POST_QUANTUM_CRYPTOGRAPHIC_AGILITY_PLAYBOOK.md | 把模型、RAG、工具、日志、签名、密钥和长期证据转成 crypto inventory、PQC migration 和 crypto-agility roadmap |
../AI_AUTHORIZED_PUSH_PAYMENT_SCAM_INTERVENTION_PLAYBOOK.md | 把实时支付诈骗、客户意图、收款人风险、社工信号、step-up friction、投诉和救济转成 AI 干预架构 |
../AI_AGENT_MARKETPLACE_TOOL_CERTIFICATION_GOVERNANCE_PLAYBOOK.md | 把内部 agent/tool marketplace、capability card、risk tier、认证、签名包、权限和运行监控做成治理系统 |
../AI_CUSTOMER_VULNERABILITY_ACCESSIBILITY_INCLUSIVE_AI_PLAYBOOK.md | 把弱势客户、可访问性、plain language、渠道适配、安全升级和包容性评测转成 AI 产品护栏 |
../AI_PAYMENT_DISPUTE_CHARGEBACK_CLAIMS_EVIDENCE_PLAYBOOK.md | 把支付争议、拒付、EFT error claim、billing error、证据、SLA、临时贷记和客户沟通做成 AI claims evidence 架构 |
../AI_COLLECTIONS_HARDSHIP_DELINQUENCY_TREATMENT_PLAYBOOK.md | 把逾期预测、困难客户处理、联系策略、可访问渠道、投诉链接和 fair treatment 转成催收治理架构 |
../AI_VOICE_AI_CONTACT_CENTER_AGENT_ASSIST_GOVERNANCE_PLAYBOOK.md | 把 voice bot、实时转写、agent assist、call summary、QA、披露边界、投诉和坐席治理做成 contact center AI 控制系统 |
../AI_DIGITAL_IDENTITY_WALLET_VERIFIABLE_CREDENTIALS_TRUST_PLAYBOOK.md | 把 digital wallet、verifiable credentials、DID、WebAuthn、selective disclosure、revocation 和 trust policy 转成身份信任架构 |
../AI_OPEN_BANKING_OPEN_FINANCE_CONSENTED_DATA_SHARING_PLAYBOOK.md | 把开放银行/开放金融、客户授权、数据最小化、撤回、API 契约、第三方风险和 AI 使用边界做成数据共享架构 |
../AI_PERSONALIZED_PRICING_OFFER_DECISIONING_GOVERNANCE_PLAYBOOK.md | 把个性化价格、费率、额度、offer、实验、解释、投诉和 surveillance pricing 风险做成决策治理系统 |
../AI_DOCUMENT_INTELLIGENCE_UNSTRUCTURED_DATA_EVIDENCE_QUALITY_PLAYBOOK.md | 把 OCR、layout、字段抽取、置信度、人工复核、记录留存、篡改检测和 workflow evidence 做成文档智能架构 |
../AI_PRIVACY_CLEAN_ROOM_DATA_COLLABORATION_MEASUREMENT_PLAYBOOK.md | 把 clean room、PEC、聚合、差分隐私、合成数据、合作方测量、输出审查和目的限制做成数据协作产品 |
../AI_CREDIT_LIFECYCLE_UNDERWRITING_LINE_MANAGEMENT_GOVERNANCE_PLAYBOOK.md | 把授信、审批、额度增减、账户管理、adverse action、fair lending、组合监控和投诉证据做成信用生命周期治理 |
../AI_WEALTH_ADVICE_ROBO_ADVISOR_BEST_INTEREST_BOUNDARY_PLAYBOOK.md | 把财富建议、robo-advisor、教育/建议/执行边界、风险画像、人工升级和监督证据做成投顾 AI 架构 |
../AI_TREASURY_LIQUIDITY_ALM_FORECASTING_STRESS_EVIDENCE_PLAYBOOK.md | 把流动性预测、存款流失、ALM、压力测试、FTP、委员会决策和董事会 MI 做成 Treasury AI 证据架构 |
../AI_COMPLAINT_INTELLIGENCE_ROOT_CAUSE_REGULATORY_RESPONSE_PLAYBOOK.md | 把投诉分类、伤害识别、根因、产品缺陷、监管响应、CAPA 和整改证据做成 complaint intelligence 架构 |
../AI_ACCOUNT_OPENING_KYC_ONBOARDING_DECISION_PLAYBOOK.md | 把开户、KYC/CIP/CDD、身份核验、风险分层、例外队列、欺诈/AML 交接和客户伤害证据做成 onboarding decision 架构 |
../AI_AML_ALERT_TRIAGE_INVESTIGATION_WORKBENCH_PLAYBOOK.md | 把 AML 告警队列、实体解析、案件组装、分析师 copilot、QA、SAR 升级和调优反馈做成调查工作台 |
../AI_PAYMENT_OPERATIONS_RECONCILIATION_SETTLEMENT_EXCEPTION_PLAYBOOK.md | 把支付处理、对账、清算异常、repair queue、suspense、双控、SLA 和运营证据做成支付运营架构 |
../AI_REGULATORY_REPORTING_RISK_DATA_AGGREGATION_ATTESTATION_PLAYBOOK.md | 把风险数据聚合、报表血缘、转换控制、对账、签证、重报和监管提交证据做成 reporting architecture |
../AI_CORE_BANKING_MODERNIZATION_STRANGLER_EVIDENCE_PLAYBOOK.md | 把核心银行现代化、遗留系统绞杀、ACL、dual-run、数据迁移、cutover 和回滚证据做成转型架构 |
../AI_REQUIREMENTS_MINING_PROCESS_KNOWLEDGE_EXTRACTION_PLAYBOOK.md | 把 BRD/PRD/SOP/Jira/日志/代码/API/test case 转成需求图谱、流程知识、验收标准和 SME 校验闭环 |
../AI_UAT_REGRESSION_CERTIFICATION_BUSINESS_ACCEPTANCE_PLAYBOOK.md | 把 UAT、golden journey、synthetic transaction、AI regression、业务验收、例外和 release evidence 做成认证架构 |
../AI_PRODUCT_PORTFOLIO_RATIONALIZATION_CAPABILITY_DECOMMISSION_PLAYBOOK.md | 把产品组合、能力重叠、成本风险、依赖图、客户迁移、feature/API/model sunset 和下线证据做成组合治理 |
../AI_ADOPTION_ANALYTICS_BEHAVIOR_CHANGE_VALUE_REALIZATION_PLAYBOOK.md | 把 AI 采用、行为改变、work-as-done、价值兑现、cohort 分析和控制影响做成 adoption evidence 架构 |
../AI_EXECUTIVE_INVESTMENT_NARRATIVE_BUSINESS_CASE_PLAYBOOK.md | 把 AI 架构、产品价值、风险偏好、收益兑现和 stop/scale/pivot gate 讲成高管可投资决策包 |
../AI_STAKEHOLDER_DECISION_COALITION_INFLUENCE_RISK_PLAYBOOK.md | 把审批者、阻断者、出资者、运营者、审计者、采用者和受影响方组织成 decision coalition 和 evidence forum |
../AI_WORKFORCE_CAPABILITY_ACADEMY_ROLE_SKILL_TRANSFORMATION_PLAYBOOK.md | 把 AI 角色、技能缺口、能力证据、场景评估和 adoption telemetry 做成 workforce capability system |
../AI_DELIVERY_ASSURANCE_CONTROL_TOWER_RELEASE_READINESS_PLAYBOOK.md | 把 AI discovery、pilot、release、scale、post-release assurance 做成 evidence-based delivery control tower |
../AI_PRODUCT_OPERATIONS_OPERATING_CADENCE_OUTCOME_REVIEW_PLAYBOOK.md | 把周运营、月度价值复盘、季度组合复盘、实验发布日历和 incident-to-roadmap 做成 AI Product Ops 节奏 |
../AI_REFERENCE_IMPLEMENTATION_PATTERN_LIBRARY_REUSE_ASSURANCE_PLAYBOOK.md | 把 RAG、copilot、tool gateway、human review、eval 和 observability 模式沉淀成可复用参考实现和证据包 |
../AI_TECHNOLOGY_RADAR_SCENARIO_PLANNING_STRATEGIC_FORESIGHT_PLAYBOOK.md | 把 AI 技术、监管、供应商、成本、安全和 workforce 信号转成雷达、情景、选项组合和架构 runway |
../AI_PRODUCT_RESPONSIBILITY_ACCOUNTABILITY_BOUNDARY_PLAYBOOK.md | 把产品承诺、决策权、行动权、人工责任、供应商边界、客户救济和残余风险 owner 做成责任架构 |
../AI_CONTEXT_SUPPLY_CHAIN_PROVENANCE_POISONING_DEFENSE_PLAYBOOK.md | 把 prompt、RAG、tool output、memory、metadata 和 workflow state 做成有 provenance、权限、质量和投毒防御的上下文供应链 |
../AI_AGENTIC_PROCESS_AUDIT_WORKFLOW_REPLAY_ASSURANCE_PLAYBOOK.md | 把 agent plan/action/observation/approval/output 事件做成可审阅、可重放、可抽样、可质询的流程审计证据 |
../AI_ORGANIZATIONAL_MEMORY_KNOWLEDGE_ASSET_GOVERNANCE_PLAYBOOK.md | 把组织记忆、知识资产、决策记录、遗忘/保留、引用、版本和知识债做成 AI memory governance |
../AI_EVAL_DATASET_LIFECYCLE_GOLDEN_SET_TEST_DATA_FACTORY_PLAYBOOK.md | 把 golden set、eval data lifecycle、coverage、synthetic/real cases、label governance 和 regression evidence 做成测试数据工厂 |
../AI_SYNTHETIC_USER_SIMULATION_PERSONA_SCENARIO_LAB_PLAYBOOK.md | 把 synthetic users、persona scenarios、journey simulation、edge cases 和 behavior testbed 做成 AI 产品发现与验证实验室 |
../AI_HUMAN_FACTORS_OPERATIONS_COGNITIVE_LOAD_AUTOMATION_BIAS_PLAYBOOK.md | 把认知负载、自动化偏差、信任校准、人工复核负担和操作员质量做成人因运营架构 |
../AI_PROCUREMENT_INTAKE_VENDOR_EVALUATION_SANDBOX_BUILD_BUY_PLAYBOOK.md | 把供应商 intake、sandbox eval、build/buy/partner 取舍、合同证据和退出约束做成采购决策架构 |
../AI_REASONING_BUDGET_TEST_TIME_COMPUTE_VERIFIER_CASCADE_PLAYBOOK.md | 把推理预算、test-time compute、验证级联、成本/延迟/SLO 和升级证据做成可治理推理架构 |
../AI_MODEL_PORTFOLIO_BENCHMARKING_CAPABILITY_SCORECARD_SELECTION_GOVERNANCE_PLAYBOOK.md | 把模型组合、能力评分卡、benchmark、champion/challenger、退役触发和选型记录做成持续治理 |
../AI_UNCERTAINTY_UX_ABSTENTION_CONFIDENCE_ESCALATION_PLAYBOOK.md | 把不确定性、置信表达、拒答/弃权、补充信息、人工升级和客户沟通测试做成体验控制 |
../AI_RISK_QUANTIFICATION_SCENARIO_LOSS_CONTROL_ROI_PLAYBOOK.md | 把 AI 风险场景、损失范围、控制有效性、残余风险、控制 ROI 和管理行动做成决策证据 |
../AI_EVIDENCE_INTEGRITY_TAMPER_EVIDENT_AUDIT_LEDGER_PLAYBOOK.md | 把证据对象、hash chain、签署链、防篡改测试、法律保全和审计查询做成 AI evidence integrity |
../AI_SHADOW_MODE_COUNTERFACTUAL_EVALUATION_SILENT_LAUNCH_PLAYBOOK.md | 把影子模式、反事实事件、标签延迟、泄漏控制、分群公平性和 rollout gate 做成静默上线架构 |
../AI_SYNTHETIC_DATA_GOVERNANCE_PRIVACY_UTILITY_FIDELITY_PLAYBOOK.md | 把合成数据 intake、源数据权限、隐私攻击测试、效用/保真度、偏差覆盖和 allowed-use 做成治理体系 |
../AI_THREE_LINES_GOVERNANCE_DECISION_RIGHTS_ASSURANCE_PLAYBOOK.md | 把三道防线、生命周期决策权、evidence forum、challenge memo、issue log 和 assurance packet 做成 AI 治理运营模型 |
Reading Output Standard
每篇论文解读完成后，优先沉淀 5 类输出：
- Concept diagram: 用自己的图解释核心机制，不只复述术语。
- Architecture mapping: 映射到 RAG / Agent / Copilot / Eval / Governance 的组件、边界和责任。
- Financial retail case: 至少一个 AML、KYC、信贷、支付、客服、合规或财富场景，说明输入、决策、证据和人工控制点。
- ADR draft: 写清技术取舍、适用条件、反转条件和不可接受风险。
- Evidence pack: 明确 eval dataset、threshold、release gate、monitoring signal、owner 和 review cadence。
Preservation Rule
旧的 docs/ai/ 和 docs/llm/ 内容继续保留。这个目录补的是“经典论文精读 + AI 产品架构 + 评估治理 + 金融零售场景化”的系统学习材料，不是清理旧笔记。