Summary:

The rise of unifying frameworks enabling LLM-LLM collaboration for open-ended tasks has led to the exploration of multi-LLM co-authoring scenarios. CollabStory introduces a dataset of LLM-generated collaborative stories, focusing on single to multi-author settings for story generation. The study extends authorship-related tasks for LLM-LLM collaboration, addressing challenges related to plagiarism detection, credit assignment, and academic integrity.

Importance:

CollabStory contributes to understanding and developing techniques for leveraging multiple LLMs in collaborative story writing tasks. This research is vital for enhancing the capabilities of LLMs in generating diverse and engaging content, while also addressing concerns related to ethical writing practices and integrity in the academic and creative domains.