标签
This paper introduces NCP-Bench, a benchmark derived from 100 movie synopses for evaluating long-horizon narrative consistency in LLM-based interactive storytelling agents. Experiments show that even strong models like GPT-5.2 struggle to maintain logical consistency, with a 42% survival rate after 20 turns and high fact conflict rates.
NeuWorld 是一个新的交互式视频生成系统，它利用紧凑的神经隐式场景表示以及结合扩散变换器的变换器VAE，实现轨迹条件渲染，达到长时一致性。