QuAREL

anchor

Findings

• IC-1499Self-rationalization quality and task accuracy scale with model size across GPT-3, FLAN-T5, and LLaMA on five QA datasets [eval]