Language Model Behavior: A Comprehensive Survey (2303.11504v2)
Abstract: Transformer LLMs have received widespread public attention, yet their generated text is often surprising even to NLP researchers. In this survey, we discuss over 250 recent studies of English LLM behavior before task-specific fine-tuning. LLMs possess basic capabilities in syntax, semantics, pragmatics, world knowledge, and reasoning, but these capabilities are sensitive to specific inputs and surface features. Despite dramatic increases in generated text quality as models scale to hundreds of billions of parameters, the models are still prone to unfactual responses, commonsense errors, memorized text, and social biases. Many of these weaknesses can be framed as over-generalizations or under-generalizations of learned patterns in text. We synthesize recent results to highlight what is currently known about LLM capabilities, thus providing a resource for applied work and for research in adjacent fields that use LLMs.