Brief

Researchers have developed a new reinforcement learning pipeline called Retell, Reward, Repeat (RRR) designed to improve Large Language Models' (LLMs) storytelling capabilities. This method integrates Structuralist Narratology with scalar narrativity to train LLMs on logical and rational narrative event generation, addressing shortcomings in current post-training techniques like SFT. RRR utilizes a synthesized TimeTravel dataset and derives training signals from textual features via d-RLAIF, avoiding the need for reference outputs. Evaluations show RRR-trained LLMs outperform existing baselines in logic, rationality, and completeness, offering a cost-effective approach to enhancing LLM storytelling. AI

A new survey paper published on arXiv explores the intersection of narrative theory and large language models (LLMs) for automatic story generation and understanding. The paper categorizes current NLP research based on narratology concepts, highlighting that while understanding tasks are more advanced, generation methods lag in theoretical application and exploring non-fiction narratives. The authors suggest future research should focus on theory-based metrics for narrative attributes, large-scale literary analysis, and generating narratives in situated contexts to validate or refine narrative theories. AI