Maswali yanayounganishwa na mwongozo · Kati Kiwango

Maswali ya Kurekebisha Sampuli ya Kukataliwa

Inashughulikia uzalishaji na uchujaji wa mgombea, zawadi zinazotegemea kithibitishaji, mafunzo upya yanayosimamiwa na hatari kutokana na kuboresha mfungaji mwenye dosari.

Njia za mwongozo zinazohusianaKukataliwa Sampuli Fine Tuning
Swali 1 ya 8

Which loop best describes how Rejection Sampling Fine-Tuning improves a model?

Hakiki kila swali8 maswali

Majibu na maelezo huonekana unapojibu maswali.

  1. Which loop best describes how Rejection Sampling Fine-Tuning improves a model?
    • Tengeneza majibu mengi ya watahiniwa, weka yale yaliyoweka alama bora zaidi, na ujifunze upya kuyahusu
    • Futa safu za mfano zinazofanya vibaya zaidi
    • Kataa data zote za mafunzo na anza upya
    • Sampuli mara moja na ukubali matokeo kila wakati
  2. Katika vikoa vinavyoweza kuthibitishwa kama vile hesabu au msimbo, ni zawadi gani inaweza kutumika kama zawadi?
    • Mfano wa zawadi uliojifunza pekee
    • Ikiwa jibu ni sahihi au majaribio yamepita
    • Urefu wa majibu
    • Idadi ya tokeni zinazozalishwa
  3. Which model family does the cited paper describe using rejection sampling in its alignment pipeline?
    • GPT-4o
    • BERT Base
    • Meta’s Llama 2-Chat
    • Whisper Large
  4. Why might a team choose rejection-sampling fine-tuning over a policy-gradient approach for some tasks?
    • It never needs a candidate generator or scoring rule
    • It guarantees no reward hacking
    • It always requires less compute than every RL method
    • It can use supervised learning on selected generations instead of updating through policy-gradient RL
  5. Mafunzo juu ya sampuli za malipo ya juu kwa ufanisi hufanya nini?
    • Hupunguza ubora wa N katika matokeo ya sampuli moja ya modeli
    • Huongeza ukubwa wa msamiati
    • Huondoa uwezo wa muundo kutengeneza maandishi
    • Huongeza idadi ya vigezo mara mbili
  6. What should teams monitor when repeatedly fine-tuning on filtered model-generated outputs?
    • Idadi tu ya vigezo vya mfano
    • Output diversity alongside held-out correctness
    • Whether every answer uses the same wording
    • Only the training loss on selected samples
  7. What is a limitation of rejection-sampling fine-tuning’s candidate selection?
    • It can only be used with image labels
    • It always changes the tokenizer
    • It cannot train on a high-quality answer that was never generated or retained
    • It makes the scoring function unnecessary
  8. What risk arises if the scoring model used to filter candidates is flawed?
    • The policy becomes unable to generate any text
    • The tokenizer automatically changes vocabulary
    • The reference labels become unnecessary
    • The policy may learn to exploit the scorer rather than improve the intended quality