Narrative World Model: Writer Memory for Long-Form Fiction Outperforms Graphiti/Zep on Multi-Hop QA

It's a writer memory system grounded in narratology, using a typed temporal-state graph to represent story evolution, outperforming current systems on complex plot and character questions.

NWM significantly outperforms Graphiti/Zep on multi-hop narrative QA, due to its narratology-grounded structure and query-conditioned retrieval, not extraction quality.

Yes, the system can be adapted for other languages including Arabic, enhancing AI creative writing tools in the MENA region.

It provides an advanced framework for building contextual memory systems, enabling better and more accurate writing tools in the region.