Tag
The paper proposes NarraLite, an efficient multimodal generative recommendation framework that uses latent narrative reasoning to improve episodic content prediction with better accuracy and efficiency.
ClueWeaver is a reward-guided dual-agent framework that improves long-narrative question answering with compact local language models by decomposing evidence selection and reasoning into separate agents optimized via reinforcement learning.
The paper distinguishes between revision-driven updates (non-monotonic) and delayed elaboration (monotonic) in incremental narrative interpretation, demonstrating their application through visual narratives.