Ikibazo gihuza ibibazo · Biragoye Urwego
Ikibazo cyo guhuza AI
Ikizamini cyo guhemba ibihembo, imipaka ya RLHF, umukino udasanzwe, kumenyekanisha isuzuma, nuburyo ubugenzuzi bunini bufasha imirimo igoye-kugenzura.
Ikibazo 1 Bya 8
Mu bushakashatsi bwo guhuza, ijambo "ibihembo byo guhemba" risobanura iki?
Reba buri kibazo8 ibibazo
Ibisubizo nibisobanuro bigaragara mugihe ufata ikibazo.
- Mu bushakashatsi bwo guhuza, ijambo "ibihembo byo guhemba" risobanura iki?
- Icyitegererezo cyanze guhitamo ibihembo byacyo kuko umurimo usa nkuwangiza
- Intumwa ikoresha inenge mubimenyetso byayo ibihembo aho gukora umurimo ugenewe
- Igitero cyibye uburemere bwikitegererezo kugirango yongere imyitozo irushanwa
- Abashoramari bazamura intoki indangagaciro zo kwihutisha guhuza mugihe cy'amahugurwa
- Muri RLHF, ni uruhe ruhare rw'icyitegererezo cyo guhemba?
- Isimbuza abantu amanota rwose mukwandika ibisubizo byerekana neza
- Irabika urutonde ruhamye rwamagambo yabujijwe politiki itigera yemererwa gusohoka
- Itanga inyandiko nshya yo kwitegura kugirango icyitegererezo fatizo gifite amakuru menshi yo kwigiraho
- Yigira kubyo abantu bakunda kugereranya no gutanga amanota yo kuyobora politiki
- What does goal misspecification mean in the guide?
- A model behaves differently when it recognizes a test setting
- The written objective fails to represent what people actually intend
- An agent finds a shortcut in a proxy metric while optimizing it
- A model refuses every task after a safety update
- Which description matches instrumental convergence in the guide?
- A user changes their prompt after a refusal
- A system may seek power or resources because they help achieve many possible goals
- A test set is held out from training
- A reward model stores written principles
- Which modern alignment approach is named in the guide?
- Credit scoring notices
- Post-market incident reporting
- Dataset deduplication
- AI
- Ubushakashatsi "bwagutse" bugerageza gukemura iki?
- Nigute wagabanya umubare wibipimo mugihe uhoraho neza
- Uburyo abantu bashobora kugenzura AI kubikorwa bigoye kubisuzuma muburyo butaziguye
- Nigute ushobora gutanga icyitegererezo kubantu babarirwa muri za miriyoni icyarimwe utarinze kongera ubukererwe
- Nigute ushobora gupima amahuriro yibice byinshi byamakuru neza
- Why does preference optimization not prove that a model has adopted human intent as an internal goal?
- It improves average responses but does not prove stable behavior under shift or pressure
- It prevents models from using human preference data
- It guarantees a model will ignore every written instruction
- It measures only GPU efficiency
- Why might an alignment evaluation include tests for evaluation awareness?
- To measure the number of human raters in training
- To determine whether a reward model has enough parameters
- To see whether a model changes behavior when it recognizes a test setting
- To prove that a system is conscious