The reward model is trained on
APairwise comparisons of model outputs
BSingle absolute scores per output
CRule-based outputs from a regex set
DHuman-written critiques of each token
Answer & Solution
Correct answer: A. Pairwise comparisons of model outputs
Labelers rank K outputs to a prompt (K is 4 to 9). All C(K,2) pairs become preference labels. The RM learns to assign higher scalar rewards to the preferred completions.
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