alleryStart learning →The alignment goal is summarised as the model being
AFast, cheap and small
BOpen, transparent and audited
CMultilingual, multimodal and grounded
DHelpful, honest and harmless
Answer & Solution
Correct answer: D. Helpful, honest and harmless
Following the framework of Askell et al. (2021), InstructGPT aligns toward the HHH triad: helpful, honest, harmless.
These shape both training data and evaluation criteria.
Keep practising InstructGPT RLHF
Browse all InstructGPT RLHF questions for Claude, or explore more Aifoundations topics.
Related questions
Knock out the false alignment claimsRM models in InstructGPT are sized atInter-annotator agreement among training labelers is roughlyRM training uses all C(K,2) comparisons from a single prompt asThe RM training pairwise loss function uses'PPO-ptx' differs from plain PPO byThe final RL step uses which algorithm to optimise the policy?The reward model is trained on