Kuiz berkaitan panduan · Sederhana Tahap

Kuiz Pengoptimuman Keutamaan Nisbah Odds

Menjelaskan objektif pilihan bebas model rujukan ORPO, penalti nisbah odds dan latihan gabungan penyeliaan dan keutamaan.

Laluan panduan yang berkaitanPengoptimuman Keutamaan Nisbah Odds
soalan 1 daripada 8

Apakah kelebihan praktikal utama ORPO berbanding kaedah seperti DPO?

Pratonton setiap soalan8 soalan

Jawapan dan penjelasan muncul semasa anda mengambil kuiz.

  1. Apakah kelebihan praktikal utama ORPO berbanding kaedah seperti DPO?
    • Ia menghapuskan keperluan untuk model rujukan yang berasingan
    • Ia tidak memerlukan data latihan sama sekali
    • Ia hanya berfungsi pada model imej
    • Ia menjamin sifar halusinasi
  2. Apakah perbandingan 'nisbah kemungkinan' dalam ORPO?
    • Kelajuan latihan berbanding ketepatan
    • Kemungkinan respons yang dipilih berbanding respons yang ditolak
    • Dua kadar pembelajaran yang berbeza
    • Saiz model berbanding saiz set data
  3. What form does an ORPO preference example contain?
    • Only a single unlabeled response
    • A prompt with two identical completions
    • A prompt with one chosen response and one disfavored response
    • A reward score with no associated responses
  4. How does ORPO convert a response probability p into odds?
    • One minus p divided by p
    • The square root of p
    • The logarithm of the prompt length
    • p divided by one minus p
  5. Kehilangan ORPO paling sesuai digambarkan sebagai kombinasi yang manakah?
    • Entropi silang SFT ditambah istilah keutamaan nisbah kemungkinan berwajaran
    • Hanya skor model ganjaran
    • Pulangan pembelajaran peneguhan tulen
    • Suntikan hingar rawak
  6. Which response receives the supervised fine-tuning component of ORPO?
    • Only the rejected response
    • The chosen or preferred response
    • A response sampled after deployment
    • No response; ORPO has no language-modeling term
  7. Which parameter controls the weight of ORPO’s preference term relative to the supervised loss?
    • Saiz perbendaharaan kata tokenizer
    • Bilangan ketua perhatian
    • The preference-loss coefficient
    • The prompt’s character count
  8. Why should an ORPO result report its preference data and evaluation setup?
    • ORPO results are identical across datasets
    • The objective removes the need for evaluation
    • The number of model layers determines every preference
    • Outcomes depend on the dataset, coefficient, base model, and evaluation