aiqbench analysis
This research presents RRR, a reinforcement learning pipeline that leverages narrative theory and human-annotated data to improve causal reasoning in LLMs, addressing specific logical gaps where standard prompting fails. By integrating Structuralist Narratology with automated reward modeling, the study demonstrates a cost-effective method for enhancing story coherence and rationality in smaller models without relying on massive datasets.
Retell, Reward, Repeat: Reinforcement Learning for Narrative Theory-Informed Story Retelling
arXiv:2601.17226v3 Announce Type: replace-cross 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.
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.
Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
arXiv:2601.17226 [cs.CL]
(or arXiv:2601.17226v3 [cs.CL] for this version)
https://doi.org/10.48550/arXiv.2601.17226 Focus to learn more arXiv-issued DOI via DataCite
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 224 words
Related coverage
The Alignment Problem: Computer Science Meets Analytic Philosophy
LessWrong · Oct 18, 2026
The J-Space Debate, Agent Swarms, and Pacing Frontier AI - Digital Minds Newsletter #4
LessWrong · Sep 18, 2026
Synthesis Superintelligence: from Semiconductors to Superconductors — Periodic Labs’ Liam Fedus and Ekin Dogus Cubuk
Latent Space · Oct 8, 2026
An unexamined cause of the OpenAI Hugging Face hacking incident: its binary performance metric
LessWrong · Sep 23, 2026