Home › Claude › aifoundations › instructgpt_rlhf › The FLAN and T0 datasets are mentioned as baseli…
The FLAN and T0 datasets are mentioned as baselines because they
AWere the only datasets publicly available at the time
BCover translation between English and other languages
CCombine many NLP tasks with natural language instructions
DUse only human-written prompts and reject API data
Answer & Solution
Correct answer: C. Combine many NLP tasks with natural language instructions
FLAN and T0 are instruction-tuned datasets that combine many NLP tasks with natural language instructions per task. They are the closest non-RLHF baselines for InstructGPT — instruction-only fine-tuning without preference modelling.
Related questions
Why 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 byThe PPO objective includes a KL penalty against the SFT model to