引导式测验 · 中等 等级

比值比偏好优化测验

解释 ORPO 的无参考模型偏好目标、优势比惩罚以及组合监督和偏好训练。

相关引导路径比值比偏好优化
问题 1 的 8

与 DPO 等方法相比，ORPO 的主要实际优势是什么？

预览每个问题8 问题

当您参加测验时，就会出现答案和解释。

  1. 与 DPO 等方法相比，ORPO 的主要实际优势是什么？
    • 它消除了对单独参考模型的需要
    • 它根本不需要训练数据
    • 它仅适用于图像模型
    • 它保证零幻觉
  2. ORPO 中的“比值比”有何不同？
    • 训练速度与准确性
    • 所选响应与拒绝响应的几率
    • 两种不同的学习率
    • 模型大小与数据集大小
  3. What form does an ORPO preference example contain?
    • Only a single unlabeled response
    • A prompt with two identical completions
    • A prompt with one chosen response and one disfavored response
    • A reward score with no associated responses
  4. How does ORPO convert a response probability p into odds?
    • One minus p divided by p
    • The square root of p
    • The logarithm of the prompt length
    • p divided by one minus p
  5. ORPO 损失最好描述为哪种组合？
    • SFT 交叉熵加上加权优势比偏好项
    • 只有奖励模型分数
    • 纯粹的强化学习回报
    • 随机噪声注入
  6. Which response receives the supervised fine-tuning component of ORPO?
    • Only the rejected response
    • The chosen or preferred response
    • A response sampled after deployment
    • No response; ORPO has no language-modeling term
  7. Which parameter controls the weight of ORPO’s preference term relative to the supervised loss?
    • 分词器词汇量大小
    • 注意力头数量
    • The preference-loss coefficient
    • The prompt’s character count
  8. Why should an ORPO result report its preference data and evaluation setup?
    • ORPO results are identical across datasets
    • The objective removes the need for evaluation
    • The number of model layers determines every preference
    • Outcomes depend on the dataset, coefficient, base model, and evaluation