Identifying and Analyzing Task-Encoding Tokens in Large Language Models

arXiv:2401.11323v2 Announce Type: replace
Abstract: In-context learning (ICL) has become an effective solution for few-shot learning in natural language processing. However, our understanding of ICL's working mechanisms is limited, specifically regarding how models learn to perform tasks from ICL demonstrations. For example, unexpectedly large changes in performance can arise from small changes in the prompt, leaving prompt design a largely empirical endeavour. In this paper, we investigate this problem by identifying and analyzing task-encoding tokens on whose representations the task …

abstract arxiv become context cs.cl encoding example few-shot few-shot learning in-context learning language language models language processing large language large language models learn natural natural language natural language processing performance processing small solution tasks tokens type understanding