LitVISTA Benchmark Evaluates Narrative Orchestration in LLMs

Researchers have introduced LitVISTA, a structurally annotated benchmark for evaluating narrative orchestration in large language models. The benchmark is grounded in the VISTA Space framework, which unifies human and model perspectives on narrative function and structure. LitVISTA operationalizes this framework to assess how well models capture rhythm, tension, and emotional dynamics in literary texts, addressing a structural misalignment where current LLMs focus excessively on causal coherence. In an oracle setting with gold event anchors, frontier LLMs were evaluated. The work is published on arXiv.