Quiz de fine-tuning

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

1. Ban bouclage moo gëna mëna fësal ni Rejection Sampling Fine-Tuning di gëna baaxal benn model?
   • Defar tontu yu bari ci kàndidaa yi, nga denc ñi gëna am poñ, nga tàggataat ci ñoom
   • efaase model layers yi gëna bon
   • Bañ bépp done tàggat te tàmbaliwaat
   • Sample benn yoon te nangu saa yu nekk
2. Ci wàll yiñ mëna saytu lu melni math wala kode, lan moo mëna nekk neexal bi?
   • ab xeetu neexal buñu jàng kese
   • Ndax tontu bi jub na wala examen yi jàll nañu
   • Guddaayig tontu bi
   • Limu jeton yiñ defar
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. Taggat ci misaali neexal bu gëna mag, lan lay def?
   • Dafay distill kalite bu gëna baax ci N ci biir model biy génne benn misaal
   • Yokk dayo vokabileer bi
   • Dindi na model bi mëna defar mbind
   • Ñaari yoon limu parametre yi
6. What should teams monitor when repeatedly fine-tuning on filtered model-generated outputs?
   • Limu paramet model yi kese
   • 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