# Fine-Grained Modeling of Narrative Context: A Coherence Perspective via Retrospective Questions

Source: [Fine-Grained Modeling of Narrative Context: A Coherence Perspective via Retrospective Questions](<https://www.sota2.com/research/paper/68fa73771e209131cea5cb67>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-01-04T04:40:41.217Z

Authors: Liyan Xu, Jiangnan Li, Mo Yu, Jie Zhou

Published: 2024-02-21T06:14:04.000Z

## Abstract

This work introduces an original and practical paradigm for narrative comprehension, stemming from the characteristics that individual passages within narratives tend to be more cohesively related than isolated. Complementary to the common end-to-end paradigm, we propose a fine-grained modeling of narrative context, by formulating a graph dubbed NarCo, which explicitly depicts task-agnostic coherence dependencies that are ready to be consumed by various downstream tasks. In particular, edges in NarCo encompass free-form retrospective questions between context snippets, inspired by human cognitive perception that constantly reinstates relevant events from prior context. Importantly, our graph formalism is practically instantiated by LLMs without human annotations, through our designed two-stage prompting scheme. To examine the graph properties and its utility, we conduct three studies in narratives, each from a unique angle: edge relation efficacy, local context enrichment, and broader application in QA. All tasks could benefit from the explicit coherence captured by NarCo.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [Recap Identification on RECIDENT NDP (test)](<https://www.sota2.com/research/sota/recap-identification-on-recident-ndp-test.md>): Recap Identification; RECIDENT NDP (Notre-Dame de Paris) (test); Precision@5: 29.11.

- [Recap Identification on RECIDENT GOT (Game of Thrones) (test)](<https://www.sota2.com/research/sota/recap-identification-on-recident-got-game-of-thrones-test.md>): Recap Identification; RECIDENT GOT (Game of Thrones) (test); Precision@5: 34.9.

- [Question Answering on QuALITY ZeroSCROLLS leaderboard (test)](<https://www.sota2.com/research/sota/question-answering-on-quality-zeroscrolls-leaderboard-test.md>): Question Answering; QuALITY ZeroSCROLLS leaderboard (test); Accuracy: 72.8.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2402.13551>)

[PDF](<https://arxiv.org/pdf/2402.13551.pdf>)
