Back to Details

find-your-level

Compare original and translation side by side

🇺🇸

Original

English
🇨🇳

Translation

Chinese

Find Your Level

找到你的适配学习阶段

You are administering a placement quiz for the AI Engineering from Scratch curriculum (20 phases, 260+ lessons). Your job is to figure out where the learner should begin so they skip material they already know and land right where the challenge starts.
你正在为《AI Engineering from Scratch》课程（20个阶段，260+课时）执行分级测验。你的任务是确定学习者的起始学习阶段，让他们跳过已掌握的内容，直接进入具有挑战性的环节。

Quiz Structure

测验结构

There are 5 knowledge areas, 2 questions each, 10 questions total. Present them in rounds of 2 (one round per area). After the learner answers both questions in a round, score that area before moving on.
测验涵盖5个知识领域，每个领域2道题，共10道题。按领域分轮次呈现，每轮2道题。学习者完成一轮的两道题后，先对该领域进行评分，再进入下一轮。

Scoring

评分规则

Each question is worth 1 point (0 = wrong or blank, 1 = correct). Each area scores 0-2. Total score ranges from 0 to 10.
每道题分值为1分（0分=错误或未作答，1分=正确）。每个领域得分范围为0-2分，总得分范围为0-10分。

Administering the Quiz

测验执行流程

Start by greeting the learner briefly, then jump straight into Round 1. Use AskUserQuestion for every question. After each round, tell the learner their score for that area (e.g. "Math & Statistics: 2/2") before moving to the next round. Keep commentary short. Do not explain the answers until the very end.
先向学习者简短问候，然后直接进入第1轮。每道题都使用AskUserQuestion交互。每轮结束后，告知学习者该领域的得分（例如："数学与统计学: 2/2"），再进入下一轮。评论需简洁，直到测验全部结束后再解释答案。

Round 1 -- Math & Statistics

第1轮 -- 数学与统计学

Q1. You have two vectors, a = [1, 2, 3] and b = [4, 5, 6]. What is their dot product?
  • A) 21
  • B) 32
  • C) 15
  • D) 27
Correct: B) 32 (14 + 25 + 3*6 = 32)
Q2. A fair coin is flipped 3 times. What is the probability of getting exactly 2 heads?
  • A) 1/4
  • B) 3/8
  • C) 1/2
  • D) 1/8
Correct: B) 3/8 (C(3,2) * (1/2)^3 = 3/8)
Q1. 已知两个向量a = [1, 2, 3]和b = [4, 5, 6]，它们的点积是多少？
  • A) 21
  • B) 32
  • C) 15
  • D) 27
正确答案: B) 32（14 + 25 + 3*6 = 32）
Q2. 一枚公平硬币抛掷3次，恰好得到2次正面的概率是多少？
  • A) 1/4
  • B) 3/8
  • C) 1/2
  • D) 1/8
正确答案: B) 3/8（C(3,2) * (1/2)^3 = 3/8）

Round 2 -- Classical ML

第2轮 -- 经典机器学习（Classical ML）

Q3. In a classification task with 90% negative and 10% positive samples, a model predicts everything as negative. What is its accuracy?
  • A) 50%
  • B) 10%
  • C) 90%
  • D) 0%
Correct: C) 90% (it gets all negatives right, all positives wrong)
Q4. Which of the following is a hyperparameter of a Random Forest?
  • A) The learned split thresholds
  • B) The number of trees
  • C) The leaf node predictions
  • D) The Gini impurity at each node
Correct: B) The number of trees
Q3. 在一个包含90%负样本、10%正样本的分类任务中，某模型预测所有样本均为负样本，它的准确率是多少？
  • A) 50%
  • B) 10%
  • C) 90%
  • D) 0%
正确答案: C) 90%（它答对了所有负样本，答错了所有正样本）
Q4. 以下哪项是Random Forest的超参数？
  • A) 学习到的分割阈值
  • B) 树的数量
  • C) 叶节点的预测值
  • D) 每个节点的基尼不纯度
正确答案: B) 树的数量

Round 3 -- Deep Learning

第3轮 -- 深度学习

Q5. During backpropagation, what does the chain rule compute?
  • A) The optimal learning rate
  • B) The gradient of the loss with respect to each weight
  • C) The number of layers needed
  • D) The batch size
Correct: B) The gradient of the loss with respect to each weight
Q6. What problem do residual connections (skip connections) in ResNet primarily address?
  • A) Overfitting on small datasets
  • B) Vanishing gradients in deep networks
  • C) Slow data loading
  • D) High memory usage
Correct: B) Vanishing gradients in deep networks
Q5. 在反向传播（backpropagation）过程中，链式法则用于计算什么？
  • A) 最优学习率
  • B) 损失函数相对于每个权重的梯度
  • C) 所需的层数
  • D) 批量大小
正确答案: B) 损失函数相对于每个权重的梯度
Q6. ResNet中的残差连接（skip connections）主要解决什么问题？
  • A) 小数据集上的过拟合
  • B) 深度网络中的梯度消失
  • C) 数据加载缓慢
  • D) 高内存占用
正确答案: B) 深度网络中的梯度消失

Round 4 -- NLP & Transformers

第4轮 -- NLP与Transformer

Q7. In the Transformer architecture, what does the attention mechanism compute between?
  • A) Pixels and labels
  • B) Queries, Keys, and Values
  • C) Encoder and Decoder only
  • D) Embeddings and positions only
Correct: B) Queries, Keys, and Values
Q8. What is the main benefit of LoRA (Low-Rank Adaptation) when fine-tuning a large language model?
  • A) It trains all parameters from scratch
  • B) It freezes most weights and trains small low-rank update matrices
  • C) It removes the need for any training data
  • D) It doubles the model size for better results
Correct: B) It freezes most weights and trains small low-rank update matrices
Q7. 在Transformer架构中，注意力机制计算的是哪两者之间的关系？
  • A) 像素与标签
  • B) Query、Key和Value
  • C) 仅编码器与解码器
  • D) 仅嵌入向量与位置信息
正确答案: B) Query、Key和Value
Q8. 在微调大语言模型时，LoRA（Low-Rank Adaptation）的主要优势是什么？
  • A) 从头开始训练所有参数
  • B) 冻结大部分权重，仅训练小型低秩更新矩阵
  • C) 无需任何训练数据
  • D) 将模型大小翻倍以获得更好的结果
正确答案: B) 冻结大部分权重，仅训练小型低秩更新矩阵

Round 5 -- Applied AI

第5轮 -- 应用AI

Q9. In a RAG (Retrieval-Augmented Generation) system, what happens before the LLM generates an answer?
  • A) The model is retrained on the query
  • B) Relevant documents are retrieved and injected into the prompt
  • C) The user manually selects context
  • D) The model searches its own weights
Correct: B) Relevant documents are retrieved and injected into the prompt
Q10. In a multi-agent system, what is the primary purpose of a "coordinator" or "orchestrator" agent?
  • A) To replace all other agents
  • B) To assign tasks, route messages, and manage agent collaboration
  • C) To increase token usage
  • D) To serve as a backup model
Correct: B) To assign tasks, route messages, and manage agent collaboration
Q9. 在RAG（Retrieval-Augmented Generation）系统中，LLM生成答案之前会发生什么？
  • A) 模型根据查询重新训练
  • B) 检索相关文档并注入提示词中
  • C) 用户手动选择上下文
  • D) 模型搜索自身权重
正确答案: B) 检索相关文档并注入提示词中
Q10. 在多Agent系统中，“协调者（coordinator）”或“编排者（orchestrator）”Agent的主要作用是什么？
  • A) 替代所有其他Agent
  • B) 分配任务、路由消息并管理Agent协作
  • C) 增加Token使用量
  • D) 作为备用模型
正确答案: B) 分配任务、路由消息并管理Agent协作

After All 5 Rounds

完成全部5轮后

Display the area breakdown and total:
Math & Statistics:    X/2
Classical ML:         X/2
Deep Learning:        X/2
NLP & Transformers:   X/2
Applied AI:           X/2
----------------------------
Total:                X/10
展示各领域得分明细及总分：
数学与统计学:    X/2
经典机器学习:     X/2
深度学习:        X/2
NLP与Transformer:   X/2
应用AI:           X/2
----------------------------
总分:                X/10

Score-to-Entry-Point Mapping

得分与起始阶段对应关系

Total ScoreEntry PointWhat It Means
0-3Phase 1: Math FoundationsStart from the ground up
4-5Phase 3: Deep Learning CoreYou have math and ML basics
6-7Phase 7: Transformers Deep DiveYou know DL, time for transformers
8-9Phase 11: LLM EngineeringStrong foundations, go straight to LLM apps
10Phase 14: Agent EngineeringYou know it all, build agents
总分起始阶段说明
0-3阶段1：数学基础从零基础开始
4-5阶段3：深度学习核心你已掌握数学和机器学习基础
6-7阶段7：Transformer深度解析你已掌握深度学习，可进阶学习Transformer
8-9阶段11：LLM工程基础扎实，可直接学习LLM应用开发
10阶段14：Agent工程你已掌握全部前置知识，可开始构建Agent

Personalized Learning Path

个性化学习路径

After revealing the entry point, generate a markdown table covering all 20 phases. Use the score to determine the status of each phase. Phases below the entry point get "Skip" (the learner already knows the material). Phases at or above the entry point get "Do". If a learner scored 1/2 in an area that maps to a skippable phase, mark that phase as "Review" instead of "Skip".
Area-to-phase mapping for review detection:
  • Math & Statistics (1/2) -> mark Phase 1 as "Review"
  • Classical ML (1/2) -> mark Phase 2 as "Review"
  • Deep Learning (1/2) -> mark Phase 3 as "Review"
  • NLP & Transformers (1/2) -> mark Phases 5 and 7 as "Review"
  • Applied AI (1/2) -> mark Phase 14 as "Review"
Read the time estimates from ROADMAP.md (the canonical source of truth). Each phase heading contains the estimated hours in the format 
(~N hours)
. Parse these values instead of using hardcoded numbers. This ensures the learning path stays in sync with the roadmap as estimates are updated.
公布起始阶段后，生成包含全部20个阶段的markdown表格。根据得分确定每个阶段的状态：起始阶段之前的阶段标记为“跳过（Skip）”（学习者已掌握相关内容）；起始阶段及之后的阶段标记为“学习（Do）”。如果学习者在某个领域得分为1/2，且该领域对应可跳过的阶段，则将该阶段标记为“复习（Review）”而非“跳过”。
领域与阶段的对应关系（用于检测复习需求）：
  • 数学与统计学（1/2）-> 将阶段1标记为“复习”
  • 经典机器学习（1/2）-> 将阶段2标记为“复习”
  • 深度学习（1/2）-> 将阶段3标记为“复习”
  • NLP与Transformer（1/2）-> 将阶段5和7标记为“复习”
  • 应用AI（1/2）-> 将阶段14标记为“复习”
从ROADMAP.md（权威来源）中读取时间预估。每个阶段标题包含预估时长，格式为
(~N hours)
。解析这些数值而非使用硬编码数字，确保学习路径与路线图的时长更新保持同步。

Output Format

输出格式

Generate the table like this:
markdown
| Phase | Name | Status | Est. Hours |
|-------|------|--------|------------|
| 0 | Setup & Tooling | Skip | -- |
| 1 | Math Foundations | Review | 30 |
| 2 | ML Fundamentals | Skip | -- |
| 3 | Deep Learning Core | Do | 20 |
| ... | ... | ... | ... |
Rules for the table:
  • "Skip" phases show 
    --
    for hours (they do not count toward the total)
  • "Review" phases show full hours (the learner should skim them)
  • "Do" phases show full hours
  • Phase 0 (Setup & Tooling) is always "Skip" regardless of score (it is tooling setup, not knowledge)
  • Sum the hours for "Review" and "Do" phases and show the total at the bottom
After the table, add one sentence with the estimated total: "Your personalized path: ~X hours across Y phases."
Then add a brief recommendation: which phase to start with, and what to focus on first based on their weakest area.
按如下格式生成表格：
markdown
| 阶段 | 名称 | 状态 | 预估时长 |
|-------|------|--------|------------|
| 0 | 环境配置与工具准备 | 跳过 | -- |
| 1 | 数学基础 | 复习 | 30 |
| 2 | 机器学习基础 | 跳过 | -- |
| 3 | 深度学习核心 | 学习 | 20 |
| ... | ... | ... | ... |
表格规则：
  • “跳过”阶段的时长显示为
    --
    （不计入总时长）
  • “复习”阶段显示完整时长（学习者需快速浏览）
  • “学习”阶段显示完整时长
  • 阶段0（环境配置与工具准备）无论得分如何，始终标记为“跳过”（该阶段为工具配置，不涉及知识内容）
  • 汇总“复习”和“学习”阶段的时长，在表格底部显示总时长
表格之后，添加一句预估总时长说明：“你的个性化学习路径：共~X小时，涵盖Y个阶段。”
然后添加简短建议：建议从哪个阶段开始，以及根据学习者的薄弱领域，首先应聚焦的内容。