引導式測驗 · 中等 等級

法學碩士測驗中的逆轉詛咒

測試定向事實回憶、反向查詢、訓練順序、實驗控制和反向詛咒發現的限制。

問題 1 的 8

模型學習了“A 是 B”,但未能回答“B 是 A”。這叫什麼?

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。

  1. 模型學習了“A 是 B”,但未能回答“B 是 A”。這叫什麼?
    • The lost-in-the-middle effect
    • 災難性遺忘
    • The reversal curse
    • 模式崩潰
  2. 方向關聯的逆轉詛咒顯示了什麼?
    • Learning one word order does not guarantee reliable recall in the reverse order
    • Language models automatically invert every relation
    • A fact is always stored as a symmetric database edge
    • Fine-tuning removes all directional errors
  3. 逆轉詛咒只是因為模型從未看到事實而失敗嗎?
    • Yes, the model must not have read it
    • It is caused only by a short context window
    • It occurs only when facts are false
    • No; it can fail in reverse even after training on the forward statement
  4. 為什麼下一個令牌訓練可能有助於這種效果?
    • It encodes all facts into an unordered table
    • It trains on token order, so A-followed-by-B does not automatically teach B-followed-by-A
    • It optimizes reverse queries by default
    • It prevents memorization of entity names
  5. How can the effect be tested?
    • Compare model performance on matched forward and reverse versions of facts
    • Measure only the forward statement
    • Change both the entities and question at once
    • Use only a general language benchmark
  6. 反向查詢失敗對模型知識意味著什麼?
    • The model has no information about either entity
    • The model is necessarily incapable of reasoning
    • The model’s learned association may not support flexible inversion
    • Every entity relation is symmetric
  7. Why is a reverse-direction test useful for a factual QA system?
    • All questions use the same word order
    • Users may ask for either endpoint of a relation
    • It measures image classification accuracy
    • It guarantees factual truth
  8. What should be inferred from one model’s reversal-cursed answer?
    • All language models fail on every reversed fact
    • The model has no semantic representations
    • The relation must be false
    • A documented failure mode to test, not proof that every model or fact behaves identically