Beam Seetlu Quiz

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

1. Lan la 'yaatu traverse' (k) di saytu ci seetug traverse?
   • Ñaata xeetu kàndidaa yuñ denc ci jéego bu nekk ci dekodaas bi
   • Guddaay bi gëna mag ci toppalante biñ defar
   • Tamperatiir biñ jëfandikoo ci probabilite yi
   • Dayo vocabulaire model bi
2. Lan moo wuutale seetlu beam ak dekodaas bu bëgg xaalis?
   • Dafay tann benn token bi gëna am wërsëg
   • Dafay denc ay pàcc yu bari yu am poñ yu bari ci barabu def benn
   • Dafay jël ay jeton ci anam wu mucc
   • Dafay baña bayyi xel ci probabilite token yi
3. Lan moo waral ñuy faral di jëfandikoo normalisasioŋ ci guddaay ci seetug traverse?
   • Ngir gëna gaaw ci defar
   • Yokk vocabulaire bi
   • Ndax boole log-probabilite yi dafay tënk poñ yi ci toppalante yu gëna gàtt
   • Yokk aleatoire ci génne yi
4. Ban liggéey la seetlu beam bi gëna baax?
   • Nettali jaar-jaar bu ubbeeku
   • Defar ay mbind yu bari te wuute
   • Defar taalif buñu xamul
   • Tekki làkku masin, di barab bu ñu bëggee génne gu wóor
5. Lan mooy njariñu jëfandikoo yaatuwaayu traverse bu yaatu lool ngir defar lu ubbeeku?
   • Mën na génne mbind yu dëgërul, yu baamtu, wala yu yàqu
   • Dafay garanti génne gi gëna am gis-gis
   • Dina dindi soxla bépp xeetu probabilite
   • Dafay tax model bi baña bàyyi xel ci liñu dugal
6. Ci jéego bu nekk, naka la beam search di doxee ci kàndidaa yiñ wara tëye?
   • ci rang alfabetu token yi
   • Ci poñ yiñ dajale (sommu log-probabilite yi), tëye k bi gëna kawe
   • Soo tànnee kàndidaa yi gëna gudd
   • Ci tànneef ci tànneef ci kàndidaa yépp
7. What problem does GNMT’s coverage penalty aim to reduce in translation?
   • Repeatedly encoding the same source vocabulary
   • The use of subword units
   • Output translations that fail to cover source words
   • The number of encoder layers
8. What does diverse beam search modify to encourage different candidate outputs?
   • It replaces the model probabilities with uniform random scores
   • It removes all but one candidate at the first step
   • It changes the training tokenizer after decoding
   • It penalizes similarity or overlap among candidates in different beam groups