quiz bu lëkkale ak tegtal · Diggante Niveau

Quiz dekodaas bu wuute

Test poñ yi eksper–amateur yi, dindi token yuñ bokk, segg plausibilite, jëfandikoo waxtu dekodaas, ak jàngat defar xeetu wuute.

Yooni njiit yu jëm ciDekodaas buy wuutale
Laaj 1 de 8

Ci dekodaas bu wuute, ban cër la xeetu 'amateur' bu ndaw bi di def?

Xoolaatal laaj bu nekk8 laaj

Sooy def quiz bi, tontu yi ak leeral yi dañuy feeñ.

  1. Ci dekodaas bu wuute, ban cër la xeetu 'amateur' bu ndaw bi di def?
    • log-probabilité yi deñukoy dindi ngir dindi anam yu ñuy bokk
    • Dafay defar ci saasi mbindu génne gu mujj gi
    • Dañu koy méngale ak njuumti eksper bi
    • Dafay wecci eksper bi sudee eksper bi wóorul
  2. Lan moo waral dindi probabilite amateur yi di wàññi baamtu yi ak génne gu neex gi?
    • Amateur bi dañu ko tàggat mu baña baamtu kàddu
    • Generic ak jetons repetitif ñoo gëna bëgg ci xeetu model yépp, moo tax contrast bi daf leen di fomm
    • Amateur bi dafay efaase ñaari token ginaaw biñu ko defaree
    • Baamtu mënul am sudee ñaari xeetu wote
  3. What does the adaptive plausibility constraint do in contrastive decoding?
    • It guarantees every remaining continuation is coherent and true
    • It replaces the expert distribution with the amateur distribution
    • It filters out candidates whose expert probability is too low relative to the expert’s top token
    • It forces every token to have equal probability
  4. Ci gàttal lan mooy formul biy joxe poñ ci dekodaas bu wuute?
    • log p_amateur dindi log p_ekspert
    • p_expert jamono p_amateur
    • Moyenne ñaari probabilite yi
    • log p_expert dindi ab koeffisien yoon log p_amateur
  5. Benn ci njariñ yi gëna am solo ci dekodaas kontrastif mooy:
    • Liggéey ci waxtu inference te amul beneen tàggat
    • Dafay laaj ñu tàggataat ñaari model yi
    • Liggéey ci xaaj rek lay liggéey
    • Dafay dindi soxla model bu rëy
  6. Ban njuréef ci ginaaw mooy yokk xalaat bu wuute bi ci benn model boppam?
    • Seetug limyéer
    • DoLa, di wuutale diisaay yu njëkk yi ak yu mujj yi
    • Decodage bu bëgg xaalis
    • Eskalaasioŋ tàngoor
  7. The expert assigns probability 0.4 to its top token and the plausibility threshold is 0.1 of that value. What is the minimum probability for a candidate to remain eligible?
    • 0.4
    • 0.1
    • 0.04
    • 0.004
  8. Why can contrastive decoding require more inference work than using the expert model alone?
    • It must retrain both models for every prompt
    • It replaces the vocabulary with image patches
    • It runs only the amateur model
    • It scores candidates with both an expert and an amateur model at inference time