Tag
This paper introduces NarrativeWorldBench, a benchmark for evaluating long-horizon narrative consistency in audio dramas, and N-VSSM, a latent state-space model that outperforms frontier LLMs across multiple horizons and languages.