引導式測驗 · 簡單 等級
深度研究代理如何工作測驗
測試深度研究代理循環:規劃子問題、迭代搜尋和閱讀、引用綜合以及常見故障模式(如錯誤歸因)。
問題 1 的 8
深度研究代理與使用網路搜尋的普通聊天機器人答案的最大區別是什麼?
預覽每個問題8 問題
當您參加測驗時,就會出現答案和解釋。
- 深度研究代理與使用網路搜尋的普通聊天機器人答案的最大區別是什麼?
- It never uses the web
- It runs a long iterative loop of searching and reading many sources before writing
- It only answers yes or no questions
- It writes answers without citations
- 為什麼深度研究者會在一開始就提出澄清問題?
- To pin down scope, audience or time frame before planning
- To delay the answer
- Because it cannot search without a password
- To collect the user's personal data
- OpenAI於2025年2月在ChatGPT推出了哪些產品?
- Code Interpreter
- Operator Search
- Deep Research
- Gemini Deep Research
- 為什麼深入的研究報告經常錯過重要的來源?
- They refuse to read PDFs
- They only read one page
- They ignore search engines
- Paywalled, login-gated or bot-blocking sites are often unreachable
- 報告在統計數據旁邊放置了引文。指南對此有何評論?
- The citation proves the statistic is correct
- The source may not actually say that, so spot-checking is needed
- Citations are only decorative
- Only the first citation in a report matters
- How do agents typically handle the problem that raw web pages are too large to keep in context?
- They extract notes or summaries after each read and store them for later
- They delete every page after reading without notes
- They only read page titles
- They increase the font size
- What is a trade-off of multi-agent deep research designs?
- They cannot use search
- They reduce breadth
- They increase breadth but multiply token cost
- They remove the need for citations
- Which description best matches the guide's recommended mental model for these tools?
- A final authority whose reports need no review
- A replacement for all expert judgment
- A search engine with no reasoning
- A fast research assistant that produces a first draft and leads