Bibliography (3):

1. Can Language Models Encode Perceptual Structure Without Grounding? A Case Study in Color

2. Language Models are Unsupervised Multitask Learners

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