Papers › NexusSum: Hierarchical LLM Agents for Long-Form Narrative Summarization

NexusSum: Hierarchical LLM Agents for Long-Form Narrative Summarization

30 May 2025arXiv:2505.24575archive 2025-07-28

Hyuntak Kim, Byung-Hak Kim

Summarizing long-form narratives--such as books, movies, and TV scripts--requires capturing intricate plotlines, character interactions, and thematic coherence, a task that remains challenging for existing LLMs. We introduce NexusSum, a multi-agent LLM framework for narrative summarization that processes long-form text through a structured, sequential pipeline--without requiring fine-tuning. Our approach introduces two key innovations: (1) Dialogue-to-Description Transformation: A narrative-specific preprocessing method that standardizes character dialogue and descriptive text into a unified format, improving coherence. (2) Hierarchical Multi-LLM Summarization: A structured summarization pipeline that optimizes chunk processing and controls output length for accurate, high-quality summaries. Our method establishes a new state-of-the-art in narrative summarization, achieving up to a 30.0% improvement in BERTScore (F1) across books, movies, and TV scripts. These results demonstrate the effectiveness of multi-agent LLMs in handling long-form content, offering a scalable approach for structured summarization in diverse storytelling domains.

PaperPDF

In Syntology View this paper on Syntology: its repositories, every harvested function with whether it ran, its licence and the call to fetch it.

Open this paper in Syntology's Atlas, the map of the papers in Syntology's graph and their citations.

Code

No code repository is listed for this paper in the archive or in Syntology's graph.

Code Syntology ran Syntology

Not run by Syntology. Nothing on this page verifies that the listed code works.

Tasks

DescriptiveFormLong-Form Narrative Summarization

Results from the paper archive 2025-07-28

TaskDatasetModelMetricValueRank at snapshotLeaderboardReport
Long-Form Narrative Summarization BookSum NexusSum (Mistral Large) BERTScore (F1) 70.70 #1 of 10 Archive leaderboard report
Long-Form Narrative Summarization BookSum NexusSum (Mistral Large) ROUGE (geometric mean of 1/2/L) 18.27 #1 of 10 Archive leaderboard report
Long-Form Narrative Summarization BookSum NexusSum (Mistral Large) ROUGE-1 42.51 #1 of 10 Archive leaderboard report
Long-Form Narrative Summarization BookSum NexusSum (Mistral Large) ROUGE-2 10.27 #1 of 10 Archive leaderboard report
Long-Form Narrative Summarization BookSum NexusSum (Mistral Large) ROUGE-L 23.91 #1 of 10 Archive leaderboard report
Long-Form Narrative Summarization BookSum Zero-Shot (Mistral Large) BERTScore (F1) 46.42 #6 of 10 Archive leaderboard report
Long-Form Narrative Summarization BookSum Zero-Shot (Mistral Large) ROUGE-1 19.63 #6 of 10 Archive leaderboard report
Long-Form Narrative Summarization BookSum Zero-Shot (Mistral Large) ROUGE-2 2.99 #6 of 10 Archive leaderboard report
Long-Form Narrative Summarization BookSum Zero-Shot (Mistral Large) ROUGE-L 12.0 #6 of 10 Archive leaderboard report
Long-Form Narrative Summarization BookSum NexusSum (Claude 3 Haiku) ROUGE (geometric mean of 1/2/L) 16.46 #8 of 10 Archive leaderboard report
Long-Form Narrative Summarization MENSA NexusSum (Mistral Large) BERTScore (F1) 65.73 #1 of 10 Archive leaderboard report
Long-Form Narrative Summarization MENSA NexusSum (Mistral Large) ROUGE-1 44.91 #1 of 10 Archive leaderboard report
Long-Form Narrative Summarization MENSA NexusSum (Mistral Large) ROUGE-2 11.43 #1 of 10 Archive leaderboard report
Long-Form Narrative Summarization MENSA NexusSum (Mistral Large) ROUGE-L 19.23 #1 of 10 Archive leaderboard report
Long-Form Narrative Summarization MENSA Zero-Shot (Mistral Large) BERTScore (F1) 54.80 #8 of 10 Archive leaderboard report
Long-Form Narrative Summarization MENSA Zero-Shot (Mistral Large) ROUGE-1 37.43 #8 of 10 Archive leaderboard report
Long-Form Narrative Summarization MENSA Zero-Shot (Mistral Large) ROUGE-2 10.52 #8 of 10 Archive leaderboard report
Long-Form Narrative Summarization MENSA Zero-Shot (Mistral Large) ROUGE-L 21.52 #8 of 10 Archive leaderboard report
Long-Form Narrative Summarization MovieSum NexusSum (Mistral Large) BERTScore (F1) 63.53 #1 of 5 Archive leaderboard report
Long-Form Narrative Summarization MovieSum NexusSum (Mistral Large) ROUGE-1 44.91 #1 of 5 Archive leaderboard report
Long-Form Narrative Summarization MovieSum NexusSum (Mistral Large) ROUGE-2 11.43 #1 of 5 Archive leaderboard report
Long-Form Narrative Summarization MovieSum NexusSum (Mistral Large) ROUGE-L 19.23 #1 of 5 Archive leaderboard report
Long-Form Narrative Summarization MovieSum Zero-Shot (Mistral Large) BERTScore (F1) 55.50 #5 of 5 Archive leaderboard report
Long-Form Narrative Summarization MovieSum Zero-Shot (Mistral Large) ROUGE-1 39.22 #5 of 5 Archive leaderboard report
Long-Form Narrative Summarization MovieSum Zero-Shot (Mistral Large) ROUGE-2 10.53 #5 of 5 Archive leaderboard report
Long-Form Narrative Summarization MovieSum Zero-Shot (Mistral Large) ROUGE-L 22.55 #5 of 5 Archive leaderboard report
Long-Form Narrative Summarization SummScreen NexusSum (Mistral Large) BERTScore (F1) 61.59 #1 of 5 Archive leaderboard report
Long-Form Narrative Summarization SummScreen NexusSum (Mistral Large) ROUGE-1 30.44 #1 of 5 Archive leaderboard report
Long-Form Narrative Summarization SummScreen NexusSum (Mistral Large) ROUGE-2 6.40 #1 of 5 Archive leaderboard report
Long-Form Narrative Summarization SummScreen NexusSum (Mistral Large) ROUGE-L 17.95 #1 of 5 Archive leaderboard report
Long-Form Narrative Summarization SummScreen Zero-Shot (Mistral Large) BERTScore (F1) 57.23 #5 of 5 Archive leaderboard report
Long-Form Narrative Summarization SummScreen Zero-Shot (Mistral Large) ROUGE-1 29.18 #5 of 5 Archive leaderboard report
Long-Form Narrative Summarization SummScreen Zero-Shot (Mistral Large) ROUGE-2 7.43 #5 of 5 Archive leaderboard report
Long-Form Narrative Summarization SummScreen Zero-Shot (Mistral Large) ROUGE-L 19.06 #5 of 5 Archive leaderboard report

Ranks are positions in the archive's leaderboards as they stood at the 2025-07-28 snapshot. Results published since then are not among these rows, so a rank here is not a current standing.

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections