⚡闪卡盲测➔
L3S2 · 数据与检索S3 · 模型与编排S4 · 交互与 HITLS5 · 评测与观测S6 · 安全与成本Verified: 2026-11-15

可验证财务证据链的 SEC 10-K 投研 Copilot(实战交付项目)

集成本地/云端可运行架构的高含金量作品集参考项目:覆盖 Docling 财报入库、BM25+Vector RRF + BGE-Reranker 检索、LangGraph 确定性财务计算节点与 60 题黄金 Eval 回归门禁,直接对标 AI Engineer / Applied AI 面试高压追问。

适用场景 (Best For)

需要一份能扛住大厂系统设计深挖(表格跨页切分、幻觉拒答、数值零误差计算、成本延迟权衡)的后端/全栈/DS 转 AI 标志性作品。

人机边界 (Human-AI Boundary)

保留分析师人工覆盖入口(Override Citation / Verify Period)并将误召回样本一键追加进 `eval_golden_set.jsonl` 构成飞轮。

常见反模式与避坑 (Anti-Patterns)

  • Presenting only UI screenshots without before/after retrieval NDCG@5 and numerical hallucination benchmark deltas
  • Skipping negative/unanswerable query suites where a model fabricates Q4 numbers from Q3 footnotes

延迟成本与评测门禁

P50 2.1s / P95 4.2s E2E with streaming citations · Reranking + generation tokens (~$0.004-0.012 per query with prompt caching)

  • Golden 10-K Benchmark Accuracy: Baseline naive chunking ~58% -> Target pipeline >= 94% (60 real 10-K comparison questions + 15 adversarial unanswerable cases)
  • Ungrounded Numerical Hallucination Rate: 0% (strict deterministic verifier pass) (Every emitted metric cross-checked against cited JSON table rows)

📦 工程交付清单与作品集产物

  • Docling 10-K structured parser + parent-table metadata indexer
  • Hybrid BM25/Vector + BGE-Reranker service with score-threshold abstention
  • LangGraph state machine enforcing deterministic Python financial ratio calculation node
  • 60-question benchmark report comparing Naive Chunking vs Hybrid + Rerank + Calculator

🎤 面试高频追问与架构答辩题

  1. Why use Docling + table-aware chunking plus BM25 RRF instead of purely increasing embedding context length?
  2. How do you guarantee 0% arithmetic hallucination and handle fiscal calendar mismatches across Apple/Microsoft 10-K filings?
  3. When calibrated reranker scores drop below 0.35, how does the agent recover or abstain safely without hurting UX?
⚡
想检验本题的肌肉记忆?进入 3 阶闪卡主动回忆盲测
隐藏答案遮罩,自测:审题锚点 ➔ 数理公式机理 ➔ 工业标杆代码与连环追问。3 分钟攻破认知盲区!
🎯 立即在闪卡中盲测此题➔

上下游关联知识链路

[HubB] 岗位能力透镜 · AI 应用与智能体系统工程师 (AI Engineer)[HubB] 岗位能力透镜 · AI Infra / LLM Inference Systems Engineer[HubB] 岗位能力透镜 · Applied Scientist (研究落地、对齐与实验归因)[HubB] 岗位能力透镜 · AI 决策科学、实验因果与评测架构师 (Data Scientist)[HubB] 岗位能力透镜 · Forward Deployed Engineer (FDE 企业级 AI 交付架构师)[HubB] 岗位能力透镜 · 机器学习与大模型推理系统工程师 (ML Engineer)[HubB] 岗位能力透镜 · AI 产品经理与商业化架构师 (AI Product Manager)[HubB] 岗位能力透镜 · 后端 / 全栈工程师向 AI 系统迁移视图 (Software Engineer)[L0] Docling 版面布局与复杂表格解析器[L0 Card] BGE-Reranker 模型系列[L0 Card] BM25+向量混合检索与 RRF 融合[L0 Card] LangGraph 状态图与持久化[L0 Card] RAG 三元组评估 (Ragas/TruLens)[L1] 复杂长 PDF 混合解析与结构化重排入库模式[L2] SEC 10-K 财报可验证溯源投研 Copilot 全流程剖析

站内交互工具与信源依据