Ikibazo gihuza ibibazo · Hagati Urwego

Ikizamini-Igihe cyo Kubara Ikizamini

Ongera usuzume igihe-cyo kubara, gushakisha abakandida, kugabanganya guhuza n'imihindagurikire y'ibiciro-byihuta mu bipimo by'ibizamini.

Inzira zijyanye no kuyoboraIkizamini Igihe cyo Kubara
Ikibazo 1 Bya 8

'Ikizamini-gihe cyo kubara' bivuga iki?

Reba buri kibazo8 ibibazo

Ibisubizo nibisobanuro bigaragara mugihe ufata ikibazo.

  1. 'Ikizamini-gihe cyo kubara' bivuga iki?
    • Ibarura ryakoresheje imyitozo yicyitegererezo kuri data
    • Igihe bisaba gukuramo moderi
    • Ibarura ryakoreshejwe mugihe icyitegererezo gisubiza ikibazo
    • Amashanyarazi yakoreshejwe mu gukonjesha ibigo
  2. In test-time scaling studies, what can extra inference compute be used for?
    • Changing the model’s training data during the request
    • Increasing the number of model parameters automatically
    • Guaranteeing correctness without evaluation
    • Searching or revising candidate solutions and checking their quality
  3. Ni ubuhe buryo bwibanze bwo kugurisha-igihe cyo kubara?
    • Ukuri kwukuri muguhana ubukererwe nigiciro
    • Ubuke buke ariko imyitozo itinda
    • Amakuru menshi ariko ntoya
    • Igishushanyo cyiza cyo gukemura hasi
  4. Niki 'cyiza-cya-N' icyitegererezo?
    • Amahugurwa kuri N imibare nini
    • Kubyara N umukandida asubiza no guhitamo icyiza
    • Gukoresha gusa N Namagambo yambere yikibazo
    • Gukoresha icyitegererezo kuri N mudasobwa zitandukanye kugirango byihute
  5. Ni ukubera iki ibizamini-igihe bibarwa bifatwa nk '' igipimo cyo gupima '?
    • Irasimbuza ibikenewe mumahugurwa ayo ari yo yose
    • Ikora gusa kuri moderi nto
    • Itezimbere imikorere yongeramo compte aho gukura gusa ingano yicyitegererezo
    • Igabanya umubare wabakoresha icyitegererezo gishobora gutanga
  6. Nibihe bisubizo bitangaje kubara-igihe cyo kubara?
    • Moderi irashobora gukora nta GPU iyo ariyo yose
    • Amahugurwa yo guhugura aba atari ngombwa
    • Abanyamideli burigihe basubiza vuba
    • Icyitegererezo gito gitekereza igihe kirekire kirashobora guhangana nicyitegererezo kinini gisubiza ako kanya
  7. What is one reason a fixed best-of-N strategy may waste inference compute?
    • It spends the same candidate budget on easy and difficult prompts
    • It modifies model weights on every query
    • It cannot generate multiple candidates
    • It eliminates the need for scoring
  8. What did the test-time compute paper compare with best-of-N sampling?
    • A strategy that disables inference
    • A compute-optimal strategy that allocates effort based on expected solution quality
    • A larger tokenizer vocabulary
    • A method that changes image resolution