Idanwo ti o sopọ mọ itọsọna · Lile Ipele

Awọn olutọpa Iṣẹ ati Awọn Aṣoju Iṣẹ-ṣiṣe

Study how function vectors are extracted from attention heads, injected into hidden states, and tested across task contexts.

Jẹmọ itọsọna awọn ọnaAwọn olutọpa Iṣẹ ati Awọn aṣoju Iṣẹ
Ibeere 1 ti 8

What does a function vector represent in the studied language models?

Awotẹlẹ gbogbo ibeere8 ibeere

Awọn idahun ati awọn alaye han bi o ṣe n mu ibeere naa.

  1. What does a function vector represent in the studied language models?
    • A compact activation-space representation associated with a demonstrated task
    • The model’s full weight matrix
    • A single token embedding in every setting
    • The optimizer’s learning rate
  2. Bawo ni fekito iṣẹ ṣe n ṣe deede bi?
    • Nipa akopọ gbogbo awọn ifibọ ọrọ ninu awọn fokabulari
    • Nipa aropin awọn abajade ti iṣẹ-ṣiṣe kan pato-gbigbe akiyesi awọn ori lori apẹẹrẹ ta
    • Nipa itanran-yiyi awoṣe lori ọkan apẹẹrẹ
    • Nipa kika awọn iṣeeṣe softmax ikẹhin
  3. What did the function-vector paper observe after injecting a vector into a zero-shot prompt?
    • The model always crashed
    • It guaranteed correct answers on any task
    • It could trigger the associated task in the tested settings
    • It permanently changed the model weights
  4. Ilana itupalẹ wo ni o ṣe atilẹyin wiwa awọn ori ti o gbe fekito iṣẹ kan?
    • Ilọlẹ gradient lori pipadanu
    • Iwadi ina
    • Isọdọtun yiyọ kuro
    • Ayẹwo ilaja idi
  5. What do function-vector experiments provide evidence for?
    • Some models form reusable internal representations associated with tasks
    • All language models only memorize text
    • Every model has one universal task vector
    • Tokenization alone determines every behavior
  6. Why are function vectors interesting to steering research?
    • They guarantee safe or correct behavior
    • They provide an activation-level intervention that can influence task behavior in tested settings
    • They eliminate prompts for every application
    • They make all model decisions fully transparent
  7. At what part of the transformer did the paper report strong function-vector effects?
    • Only the input embedding table
    • Only the final output logits
    • Middle layers in the tested models
    • Awọn fokabulari tokenizer
  8. What happened when researchers summed some function vectors?
    • The vectors always canceled to zero
    • The model’s weights were permanently updated
    • Every possible task was composed perfectly
    • Some combinations triggered new complex tasks in the tested experiments