Ikibazo gihuza ibibazo · Hagati Urwego
Kwangwa Gutoranya Ikibazo Cyiza
Hindura ibisekuruza byabakandida no kuyungurura, kugenzura ibihembo bishingiye, kugenzura imyitozo, hamwe ningaruka zo guhitamo amanota atagira inenge.
Ikibazo 1 Bya 8
Nuwuhe muzingo usobanura neza uburyo Kwangwa Gutoranya Cyiza-Guhuza bitezimbere icyitegererezo?
Reba buri kibazo8 ibibazo
Ibisubizo nibisobanuro bigaragara mugihe ufata ikibazo.
- Nuwuhe muzingo usobanura neza uburyo Kwangwa Gutoranya Cyiza-Guhuza bitezimbere icyitegererezo?
- Tanga ibisubizo byinshi byabakandida, komeza utange amanota meza, kandi ubigumane
- Siba ibyakozwe nabi cyane
- Wange amakuru yose yo guhugura hanyuma utangire hejuru
- Icyitegererezo rimwe kandi burigihe wemere ibisohoka
- Mubishobora kugenzurwa nkimibare cyangwa code, niki gishobora kuba ibihembo?
- Icyitegererezo cyo kwiga ibihembo gusa
- Niba igisubizo ari cyo cyangwa ibizamini biratsinda
- Uburebure bw'igisubizo
- Umubare wibimenyetso byatanzwe
- 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
- 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
- Amahugurwa kuri sample-bihembo byintangarugero akora iki?
- Kugaragaza ibyiza-bya-N ubuziranenge bw'icyitegererezo kimwe-icyitegererezo gisohoka
- Yongera ingano yamagambo
- Kuraho ubushobozi bwicyitegererezo cyo kubyara inyandiko
- Kabiri ibipimo bibarwa
- What should teams monitor when repeatedly fine-tuning on filtered model-generated outputs?
- Gusa umubare wikitegererezo
- Output diversity alongside held-out correctness
- Whether every answer uses the same wording
- Only the training loss on selected samples
- 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
- 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