NexusSum: Hierarchical LLM Agents for Long-Form Narrative Summarization

Benchmark Model Rank Results
long-form-narrative-summarization-on-mensaNexusSum (Mistral Large)–BERTScore (F1): 65.73ROUGE-1: 44.91ROUGE-2: 11.43ROUGE-L: 19.23
long-form-narrative-summarization-on-mensaZero-Shot (Mistral Large)–BERTScore (F1): 54.80ROUGE-1: 37.43ROUGE-2: 10.52ROUGE-L: 21.52
long-form-narrative-summarization-on-moviesumNexusSum (Mistral Large)–BERTScore (F1): 63.53ROUGE-1: 44.91ROUGE-2: 11.43ROUGE-L: 19.23
long-form-narrative-summarization-on-moviesumZero-Shot (Mistral Large)–BERTScore (F1): 55.50ROUGE-1: 39.22ROUGE-2: 10.53ROUGE-L: 22.55