arXiv cs.AI· David Y. Liu, Xanthe Muston, Dipankar Srirag, Aditya Joshi, Sebastian Sequoiah-Grayson·· 3 天前AI 评分22
Retell, Reward, Repeat：用叙事理论驱动的强化学习教 LLM 讲因果故事
Retell, Reward, Repeat: Reinforcement Learning for Narrative Theory-Informed Story Retelling
AI 导读
研究提出 Retell, Reward, Repeat（RRR）强化学习流水线，通过人类标注的叙事均衡阶段与 LLM-as-judge 奖励信号，教 LLM 生成具备因果叙事结构的文本。
正文 · 原文
Abstract:Counterfactual storytelling exposes LLM shortcomings in causal reasoning, where few-shot prompting and SFT fail to produce logical and rational narratives. This case study introduces Retell, Reward, Repeat (RRR), a narrative theory-informed RL pipeline that teaches causal story structure. We extend an existing counterfactual stories dataset with human-annotated stages of narrative equilibrium, synthesising Structuralist Narratology with scalar narrativity. An aligned LLM-as-judge automates this framework to output reward signals, teaching the policy LLM to generate texts with causal narrative structure. Evaluations show that RRR-trained 7/8B LLMs outperform few-shot and SFT baselines in logic, rationality, and completeness, also aligning with a blind human preference subset. RRR offers a linguistically grounded, cost-effective post-training mechanism for LLM storytelling using a small, query-only dataset, showing the value of integrating linguistic theories into contemporary NLP.
| Comments: | 9 Pages, 7 figures |
| Subjects: | Computation and Language (cs.CL); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2601.17226 [cs.CL] |
| (or arXiv:2601.17226v3 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2601.17226 arXiv-issued DOI via DataCite |
Submission history
From: David Liu [view email]
[v1]
Fri, 23 Jan 2026 23:23:42 UTC (5,261 KB)
[v2]
Wed, 17 Jun 2026 01:09:11 UTC (4,603 KB)
[v3]
Sat, 10 Oct 2026 04:15:26 UTC (4,603 KB)
来源：arXiv cs.AI · arxiv.org