The paper introduces Critic's Eye, a method for computational motif discovery that models a play as a heterogeneous Narrative Topology Graph and performs discriminative inference over global structural evidence. It achieves 84.8% Hit@5 on a benchmark, nearly quadrupling the performance of large foundation models such as Gemini-3-pro-preview. The authors argue that explicit structural priors outperform mere parameter scaling for decoding the complex architecture of long‑form literature.
arXiv:2606. 05724v1 Announce Type: cross Abstract: Long-form narrative QA requires reasoning over evolving story worlds rather than isolated passages: answers may depend on earlier goals, changing character states, social relations, causal triggers, temporal position, and later consequences.
arXiv:2607. 09328v2 Announce Type: replace-cross Abstract: Answering complex questions over long documents frequently requires integrating evidence that the source itself disperses naturally across distant passages.
arXiv:2607. 09328v1 Announce Type: cross Abstract: Answering complex questions over long documents frequently requires integrating evidence that the source itself disperses naturally across distant passages.
arXiv:2609.38406v1 Announce Type: new Abstract: Access to real-world information is often noisy and fragmented. Constructing a coherent narrative from such fragments requires models to reconstruct mi...
arXiv:2605.27156v2 Announce Type: replace-cross Abstract: Retrieval-Augmented Generation (RAG) enhances Large Language Models (LLMs) by incorporating external knowledge, particularly for long-tail do...
GraphLit introduces Dynamic Heterogeneous Character Networks (DHCNs) that structure long novels into temporally localized graphs aligning characters with their textual contexts. Using a masked graph autoencoder, GraphLit learns rich literary representations and outperforms text-only, graph-only, and prior hybrid baselines on 12 character-related tasks. Ablation studies show that grounding characters in context is the primary performance driver, while narrative order and character relationships offer task-dependent gains, and the framework is applied to analyze the link between narrative non-linearity and dynamic social features.
The paper introduces a three-tier evaluation framework—recovery, mining, and discovery—for unsupervised narrative label generation in disinformation datasets. It compares clustering-based and graph-community-based pipelines across seven datasets, finding that clustering can underrepresent prominent topics while graph methods produce many singletons that human annotators recognize as valid narratives. The authors release human-validated narrative candidate labels for the Climate Obstruction and PolyNarrative datasets to aid taxonomy development and dataset expansion.
The paper presents a detailed examination of narrative elements—agency, setting, and events—within the Dolma web-scale pretraining corpus. Using a framework of 11 interpretable dimensions, the authors hand‑annotated 400 passages, expanded this to a 25,000‑passage LLM‑labeled dataset, and trained NarraBERT models to predict narrative features across 13 million passages, producing the NarraDolma dataset. The study reveals that narrative structure is measurable at scale and that narrative qualities vary unevenly across different data sources, topics, and formats, highlighting gaps in current data curation practices.
arXiv:2605.20364v2 Announce Type: replace Abstract: Automatic evaluation of long-form literary writing remains challenging, particularly for creativity-related dimensions such as originality and flex...
arXiv:2610.32002v1 Announce Type: new Abstract: The rapid growth of microdramas has increased demand for high-quality script production. Long-form microdrama script generation with LLMs requires both...
arXiv:2601. 01095v3 Announce Type: replace-cross Abstract: Multimodal large language models (MLLMs) have achieved impressive progress in vision-language reasoning, yet their ability to understand temporally unfolding narratives in videos remains underexplored.
The paper introduces a modular framework that builds an entity‑interaction graph from narratives without relying on annotated node labels. It uses a domain‑adapted BERT encoder to identify entity spans, constructs a bipartite graph, and classifies it with a Heterogeneous Graph Transformer. By casting the graph as a hypergraph and applying counterfactual distillation, the method estimates the Individual Treatment Effect of each node, pruning extraneous entities to produce a minimal predictive subgraph while maintaining high accuracy on the PAN Oppositional Narratives benchmark.