# Interesting papers ## Context as Memory: Scene-Consistent Interactive Long Video Generation with Memory Retrieval - https://context-as-memory.github.io - https://arxiv.org/pdf/2506.03141 - Genie3와 같은 Video generation 모델들이 3D consistent하게 geometry 생성을 어떻게 하는가? - 바로 Keyframe을 만들어서, context로서 집어넣는 것... 오호 마치 SLAM 같구나 <img width="1388" height="791" alt="Image" src="https://github.com/user-attachments/assets/3f70f150-28cc-4bd1-ba94-64fb99ac9fd8" />  <img width="682" height="591" alt="Image" src="https://github.com/user-attachments/assets/012ed4d4-c563-4e06-8c31-9871c698444e" /> ## GameFactory - Creating New Games with Generative Interactive Videos - https://yujiwen.github.io/gamefactory/ - Genie3와 굉장히 유사함 - 얘는 깊게 얘기하지 않을겁니다. CaM이 잘 적용된 예시를 보여주기 위함임 - 관심있으신 분들께서는 아래 논문도 보세요 - Position: Interactive Generative Video as Next-Generation Game Engine (https://arxiv.org/abs/2503.17359) - A Survey of Interactive Generative Video (https://arxiv.org/abs/2504.21853) <img width="1169" height="1547" alt="Image" src="https://github.com/user-attachments/assets/c3b55f31-86bd-4fca-ad73-46982d8f5cb6" /> ## ReMEmbR: Building and Reasoning Over Long-Horizon Spatio-Temporal Memory for Robot Navigation - https://nvidia-ai-iot.github.io/remembr/ - Visual-language embedding을 사용해서 RAG DB를 구축한 후, LLM을 이용해서 query를 해 navigation을 하는 방식  