Une évaluation générée réelle, ouverte

A new developer joins a team using FastAPI and needs to quickly understand the API endpoints and their associated data models without sifting through source code. How does FastAPI, by default, provide interactive documentation to facilitate this process?

1. It generates static HTML files containing API specifications that must be manually hosted on a separate web server.
2. It automatically generates and serves interactive Swagger UI (OpenAPI) and ReDoc documentation at default URL paths like `/docs` and `/redoc`.✓ Réponse attendue
3. It requires explicit installation and configuration of a third-party documentation generator, along with manual definition of API schemas.
4. It integrates with a custom decorator system where each endpoint needs a `@document` decorator to appear in a generated text-based API reference.

For an internal backend application integrating with a third-party LLM API (e.g., OpenAI, Anthropic) using API keys for authentication, what is a typical maximum recommended interval for rotating these API keys to maintain good security hygiene? days

True or false: In a RAG pipeline, if the vector database is unavailable, the system will fail to generate a response.

In a new Python FastAPI service, you are instrumenting with OpenTelemetry to capture traces. The team wants to add several high-cardinality attributes (e.g., full request paths, unique session IDs, or specific LLM prompt contents) to every span for detailed debugging. Describe one potential performance impact of this approach and propose a strategy to mitigate this while still retaining useful debugging information.

Rank the four projects from highest to lowest priority using the stated rule. Priority = (benefit if completed) divided by (hours required); rank highest ratio first. A team has 10 hours to choose projects. Project W: benefit 40, 8 hours. Project X: benefit 30, 2 hours. Project Y: benefit 54, 9 hours. Project Z: benefit 12, 4 hours.

1. 1Project W (40 / 8)
2. 2Project X (30 / 2)
3. 3Project Y (54 / 9)
4. 4Project Z (12 / 4)

A model processes 800 requests per minute. After an optimization, the processing time per request is reduced by 25%. How many requests per hour can the model handle now? requests

A company uses a sequence of codes to track inventory. The first three codes are 123, 234, and 345. Based on the pattern, which code should come next?

1. 456✓ Réponse attendue
2. 567
3. 345
4. 123

Describe a specific situation where you used an AI code assistant (e.g., GitHub Copilot, Cursor, etc.) to generate or complete a piece of backend code, such as a utility function, an API client, or a data processing script. What was the task, which AI tool did you use, and how did you interact with it? Detail a non-trivial issue or flaw you encountered in the AI's output (e.g., a logic bug, an inefficient approach, or a deprecated API call), and explain exactly how you identified and corrected it.

You're using an AI assistant to draft a marketing email for a new product launch. The initial draft is too generic, lacks the enthusiastic and benefit-driven tone you need, and misses a specific call to action. Which approach best leverages the AI for effective refinement?

1. Ask the AI to 'make it better' without providing any additional specific instructions.
2. Completely rewrite the email yourself, abandoning the AI's initial draft.
3. Provide specific, actionable feedback to the AI, detailing the desired tone, key benefits to highlight, and the exact call to action.✓ Réponse attendue
4. Start a brand new conversation with the AI and ask it to draft a new email from scratch.

You are using an AI assistant to draft a short summary of a complex industry report for an internal audience. The AI generates a summary that includes a specific statistic and a quote attributed to a well-known industry expert, neither of which you recall seeing in the original report. What is the most important next step to take before sharing this summary?

1. Assume the AI found the information from a broader knowledge base and include it.
2. Edit the summary to make the language more formal and professional.
3. Verify the statistic and quote against the original industry report to confirm their accuracy.✓ Réponse attendue
4. Ask the AI to explain where it found the specific statistic and quote.

Your AI writing assistant generates a marketing draft containing culturally insensitive stereotypes you hadn't explicitly requested. Which response is MOST effective? Which is LEAST effective? A Revise the prompt to explicitly prohibit stereotypical language and regenerate the draft Best Worst B Review the output for biased patterns using the organization's ethical guidelines and flag for human review Best Worst C Accept the draft as non-malicious AI behavior and proceed with minor editing Best Worst D Ignore the issue since the AI doesn't explicitly violate any stated policies Best Worst

Rank these approaches for drafting a client-facing project proposal with minimal initial guidance Rank from most to least appropriate for this task: You need to generate a professional proposal document but only have a vague brief: 'Create a 10-page operations improvement plan for a mid-sized retail business.'

1. 1Send a single prompt with the full instruction to an AI writer
2. 2Start with a zero-shot prompt, then iteratively refine outputs based on feedback
3. 3Use a pre-built proposal template with fill-in-the-blank sections
4. 4Generate multiple variations and manually combine elements

A backend engineer is using an AI assistant to get information related to their work. Which of the following statements provided by the AI assistant is MOST likely to be an obvious factual error or unreliable without immediate human verification?

1. Using a database transaction ensures that a series of operations are either all completed successfully or all rolled back.
2. The latest stable version of Node.js, as of today, is 22.1.0, which was released on May 1st, 2024.✓ Réponse attendue
3. RESTful APIs typically use HTTP status codes to indicate the outcome of a request, such as 200 for success or 404 for not found.
4. Asynchronous programming allows a program to initiate long-running tasks without blocking the execution of other tasks.

A junior analyst is using an AI assistant to draft a summary of a confidential internal report. They input the full report text and ask the AI to summarize it. However, the AI's output includes general best practices and common knowledge about the topic that were not present in the original report. Which conceptual approach is most effective to ensure the AI only uses information directly from the provided report text?

1. Ask the AI to generate a much longer summary, believing more detail will inherently stick closer to the source material.
2. Instruct the AI to 'Only use the information provided in the preceding text and do not introduce any external details or assumptions.'✓ Réponse attendue
3. Break the report into several smaller sections and ask the AI to summarize each section individually.
4. Tell the AI to summarize the report as if explaining it to a specific, less knowledgeable audience, such as 'a new intern.'

You review AI-generated code for a user authentication feature and notice the implementation disproportionately fails for non-Western calendar systems. What is the MOST effective action?

1. Commit the code with a note to fix later, as the primary use case works correctly.
2. Run the code through a bias detection tool and document specific failure scenarios for review.
3. Query the AI for alternative implementations that handle diverse calendar systems.
4. Merge the code as-is, assuming edge cases will be addressed in later sprints.

You're using an AI assistant to draft a short summary of a recent company announcement. After reviewing the initial draft, you notice several minor inaccuracies in dates and product names. What is the most effective next step to get a more accurate summary?

1. Start a brand new conversation and re-enter the original prompt and information.
2. Tell the AI that its previous response was incorrect and ask it to try again without providing specific corrections.
3. Point out the specific inaccuracies (e.g., 'The launch date was July 15th, not July 1st') and ask the AI to revise the summary.✓ Réponse attendue
4. Copy the AI's incorrect summary, paste it into a new document, and manually edit all the errors yourself.

How frequently do you use AI tools to assist with coding tasks such as generating snippets, debugging, or optimizing performance?

LISTENING QUESTION Preview — replay this clip as many times as you like. Play Audio 音声を聞いて、次の二点について書きなさい。（1）新しい在宅勤務の方針の主な内容を、100字程度で要約しなさい。（2）この方針についてのあなたの意見を、理由を挙げながら150字程度で述べなさい。

コンビニのレジで、店員に商品を一つ買いたいと伝えます。最も適切な言い方はどれですか。

1. これをお願いします。✓ Réponse attendue
2. これをレジに持っていきます。
3. これをあげてください。
4. これを買いましょう。

会議室の使用予約についてのお知らせ：今週水曜日の午後3時から4時まで、3階第2会議室を使用希望者がいます。事前申し込みが必要です。詳細は人事部までお問い合わせください。 次の文章に基づき、正しい対応を選んでください。「本日中に提出が必要な資料がシステムにアップロードされました。確認後、所属課長宛てにメールで連絡してください。」

1. 所属課長に電話で報告する
2. 明日までにメールを送る
3. 資料を紙で印刷して提出する
4. システムからデータをダウンロードする✓ Réponse attendue

訪問者が店内を歩いているとき、受付で「トイレの場所を教えていただけますか？」と尋ねました。 店舗内で案内を求める際、最も適切な言い回しはどれですか？

1. トイレがどこですか？
2. トイレの場所を教えていただけますか？✓ Réponse attendue
3. トイレはありますか？
4. トイレを探しています。

スーパーで牛乳の売り場が分からないとき、店員に尋ねる最も適切な言い方はどれですか。

1. すみません、牛乳はどこにありますか。✓ Réponse attendue
2. 牛乳の場所を教えろ。
3. 牛乳はどこ？
4. 牛乳を知っていますか。

「彼は本を読むました。」という文の「読むました」を正しい形に直すとどうなりますか？

次の語句を正しい順序に並べ替えて、意味の通る一文を完成させてください。 正しい順序に並べ替えてください：

1. 1本を
2. 2昨日
3. 3借りに行きました
4. 4私は
5. 5図書館へ

レストランで飲み物を注文するとき、最も適切な言い方はどれですか。

1. コーヒーをお願いします。✓ Réponse attendue
2. コーヒーをもらってください。
3. コーヒーにしましょうか。
4. コーヒーを飲んでください。

会社で同僚に資料のコピーを頼むとき、最も丁寧な言い方はどれですか。

1. この資料をコピーしていただけますか。✓ Réponse attendue
2. この資料をコピーしろ。
3. この資料をコピーしてあげてください。
4. この資料をコピーしますか。