Chapter 1: Background and Fundamental Concepts
1.1 Background
Definition and Evolution of Zero-Shot CoT
Zero-Shot CoT, short for Zero-Shot Coreference Tracking, is a natural language processing (NLP) technique that enables the identification and linking of entities across sentences without prior training on specific entity sets. Unlike traditional named entity recognition (NER) and coreference resolution systems that require annotated data for the entities they are trained on, Zero-Shot CoT systems can generalize to unseen entities and contexts. This capability is particularly important in scenarios where labeled data is scarce or expensive to obtain.
The concept of