Abstract:
Citation Text Generation (CTG) is the task of generating the text that explains the relationship between a pair of scientific documents. Given the context of a source paper (citing paper) X and a cited paper Y, the task generates text in natural language to describe Y in the given context of X. However, previous works in this field are mainly based on standard summarization of both documents. The plan is to understand the influence/ contribution of various parts of citing paper and cited paper (when provided as input) while generating citation text using various prompting techniques in various LLMs. Our research contributes to generating more accurate citations and effective LLM interaction.
