Idanwo ti o sopọ mọ itọsọna · Lile Ipele

Aṣoju Iṣapeye adanwo

Checks typical surprisal, entropy-based ranking, the typical_p mass threshold, and differences from probability-ranked top-p sampling.

Jẹmọ itọsọna awọn ọnaApejuwe Aṣoju
Ibeere 1 ti 8

Ohun ti mojuto opoiye wo ni aṣoju iṣapẹẹrẹ afiwe kọọkan oludije àmi lodi si?

Awotẹlẹ gbogbo ibeere8 ibeere

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

  1. Ohun ti mojuto opoiye wo ni aṣoju iṣapẹẹrẹ afiwe kọọkan oludije àmi lodi si?
    • Awọn akoonu alaye ti o ti ṣe yẹ (entropy) ti pinpin
    • Iwọn fokabulari gbogbogbo ti awoṣe
    • Ipo àmi ninu gbolohun ọrọ
    • Paramita iwọn otutu nikan
  2. Bawo ni iṣapẹẹrẹ aṣoju ṣe tọju aami kan ti o ṣeeṣe julọ julọ?
    • Nigbagbogbo o yan ni akọkọ
    • O le yọkuro ti iyalẹnu rẹ ba wa ni isalẹ iye ti a reti
    • O sekeji awọn oniwe-iṣeeṣe
    • O foju pa iṣeeṣe patapata
  3. Apẹẹrẹ wo ni o ni iwuri nipa ede eniyan?
    • Awọn eniyan nigbagbogbo yan awọn ọrọ to ṣeeṣe ti o ṣọwọn
    • Gbẹtọ lẹ ma nọ dọho gbede
    • Awọn eniyan ibasọrọ sunmọ iwọn alaye ti o duro ni aijọju
    • Awọn eniyan sọrọ ni awọn gbolohun ọrọ gigun
  4. Kini paramita tau ni iṣakoso iṣapẹẹrẹ aṣoju?
    • Oṣuwọn ẹkọ lakoko ikẹkọ
    • Awọn nọmba ti fẹlẹfẹlẹ ninu awọn awoṣe
    • Awọn ti o pọju gbolohun ipari
    • Ibi-iṣe iṣeeṣe akopọ ti a gba sinu eto aṣoju
  5. Compared with greedy decoding, what does typical sampling try to change?
    • It samples from tokens whose surprisal is closer to the distribution’s expected surprisal
    • It always selects the most probable token
    • It guarantees every continuation is factually correct
    • It removes probability scores from decoding
  6. Ninu iru sọfitiwia wo ni iṣapẹẹrẹ aṣoju ti o wọpọ si awọn olupilẹṣẹ?
    • Awọn ohun elo lẹja
    • Awọn ile ikawe itọka LLM orisun-ìmọ bii Awọn Ayirapada Oju famọra
    • Awọn irinṣẹ ṣiṣatunṣe aworan
    • Awọn ẹrọ ibeere aaye data
  7. How does typical sampling rank candidate tokens differently from top-p sampling?
    • By alphabetical order of token strings
    • By token position in the prompt
    • By distance between token surprisal and expected surprisal, rather than probability rank alone
    • By the model’s training loss
  8. Which value does typical_p control?
    • The model’s vocabulary size
    • The number of transformer layers
    • The probability that the answer is factually correct
    • The target cumulative probability mass collected in the typical set