引導式測驗 · 中等 等級
Claude 模範家庭測驗
查看 Claude 的憲法 AI、長上下文文件工作、模型層以及 Anthropic 模型系列中的實際編碼和分析用途。
問題 1 的 8
哪家公司開發了 Claude 系列型號?
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。
- 哪家公司開發了 Claude 系列型號?
- Anthropic
- Meta
- Amazon
- Adobe
- 哪種訓練技術與 Claude 密切相關?
- 卷積池化
- 憲法人工智慧
- 遺傳演算法
- 僅手動規則腳本
- Claude 尤其以何種功能聞名?
- 生成物理機器人
- 列印 3D 模型
- 處理長文檔的非常大的上下文視窗
- 挖掘加密貨幣
- Anthropic 將 Claude 的目標描述為有幫助、無害,還有什麼?
- 搞笑
- 隱藏
- 硬體限制
- 誠實
- 根據 Claude Academy,針對這些任務類型建議採用哪種配對?
- Haiku for quick lookups and Sonnet for coding or multi-step workflows
- Haiku for complex reasoning and Opus for only short tasks
- Every model for the same workload with identical latency
- Opus only for image compression
- What does Anthropic describe Claude Code as helping developers do?
- Diagnose medical images without clinician review
- Work with code from a command-line development environment
- Issue insurance policies from aerial photographs
- Replace a source-control system without developer input
- In Constitutional AI’s second phase, how does AI feedback contribute to training?
- A human writes every training response from scratch
- The model removes the language model and trains only a classifier
- A model uses written principles to compare response pairs and provide preference labels
- The system uses product pricing to select a model tier
- What outcome did Anthropic report from Constitutional AI harmlessness training?
- It proved the model cannot produce harmful content
- It guaranteed that every answer is honest and correct
- It eliminated the need for preference evaluation
- It reduced harmful responses in evaluation while the paper also discusses evasiveness tradeoffs