Home › Claude › aifoundations › instructgpt_rlhf › InstructGPT is trained by which three high-level…
InstructGPT is trained by which three high-level steps?
APretrain, distil, ship to product
BSupervised fine-tune, train RM, PPO
CQuantise, prune, deploy on the edge
DTokenise, embed, classify per token
Answer & Solution
Correct answer: B. Supervised fine-tune, train RM, PPO
The InstructGPT recipe: (1) collect labeler demonstrations and fine-tune GPT-3 (SFT), (2) collect preference comparisons and train a reward model (RM), (3) fine-tune the SFT policy against the RM with PPO.
Related questions
The FLAN and T0 datasets are mentioned as baselines because theyWhy is honesty measured via 'truthfulness' rather than directly?Prioritising helpfulness during training while prioritising harmlessness at evaluation is RM 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 by