Bibliography (3):

1. GPT-3: Language Models are Few-Shot Learners

2. MegatronLM: Training Billion+ Parameter Language Models Using GPU Model Parallelism

3. MMLU: Measuring Massive Multitask Language Understanding