Measuring Inductive Biases of In-Context Learning with Underspecified Demonstrations

Keyphrases

• Contextual Learning 100%
• Inductive Bias 100%
• Feature Bias 50%
• Large Language Models 33%
• New Task 16%
• Feature Combination 16%
• Strong Prior 16%
• Generalization Behavior 16%
• NLP. 16%
• Natural Language Instruction 16%
• Bias in Favor 16%
• GPT-3 16%
• Punctuation 16%
• Lexical Features 16%
• Prior Bias 16%

Computer Science

• Large Language Model 100%
• Generative Pre-Trained Transformer 3 50%

Economics, Econometrics and Finance

• Learning Context 100%