Agenda-based Narrative Extraction: Steering Pathfinding Algorithms with Large Language Models
Summary: arXiv:2603.29661v1 Announce Type: cross
Abstract: Existing narrative extraction methods face a trade-off between coherence, interactivity, and multi-storyline support. Narrative Maps supports rich interaction and generates multiple storylines as a byproduct of its coverage constraints, though this comes at the cost of individual path coherence. Narrative Trails achieves high coherence through maximum capacity path optimization but provides no mechanism for user guidance or multiple perspectives. We introduce agenda-based narrative extraction, a method that bridges this gap by integrating large language models into the Narrative Trails pathfinding process to steer storyline construction toward user-specified perspectives.
Our approach uses a large language model (LLM) at each step to rank candidate documents based on their alignment with a given agenda while maintaining narrative coherence. Running the algorithm with different agendas yields different storylines through the same corpus. We evaluated our approach on a news article corpus using LLM judges with Claude Opus 4.5 and GPT 5.1, measuring both coherence and agenda alignment across 64 endpoint pairs and 6 agendas.
Key Findings
Our research produced several significant outcomes:
- LLM-driven steering achieves 9.9% higher alignment than keyword matching on semantic agendas (p=0.017).
- There is a 13.3% improvement on the Regime Crackdown agenda specifically (p=0.037).
- Keyword matching remains competitive on agendas with literal keyword overlap.
- The coherence cost of LLM steering is minimal, reducing coherence by only 2.2% compared to the agenda-agnostic baseline.
- Counter-agendas that contradict the source material scored uniformly low (2.2-2.5) across all methods, confirming that steering cannot fabricate unsupported narratives.
Conclusion
Agenda-based narrative extraction offers a promising advancement in the field of narrative construction and pathfinding algorithms by utilizing the capabilities of large language models. By effectively steering storyline construction towards user-defined perspectives, this method enhances the coherence of narratives while also enabling multiple storylines to emerge from the same source material. Our findings indicate that this approach not only improves alignment with specific agendas but also maintains a high level of narrative coherence, making it a valuable tool for applications in journalism, storytelling, and interactive media.
As we continue to explore the integration of AI in narrative generation, further research will be necessary to refine these algorithms and assess their impact across various domains. The potential for enhanced interactivity and user engagement in narrative experiences represents an exciting frontier for both researchers and practitioners in the AI landscape.