NPC decision-making with LLMs enables dynamic, context-aware responses in games and simulations by leveraging natural language understanding and generation. By integrating reinforcement learning, retrieval-augmented generation (RAG), or fine-tuned models, NPCs can adapt to player actions, maintain coherent narratives, and exhibit more human-like behavior.
Focuses on efficiently modifying LLMs to update or correct specific facts while preserving overall model performance. By leveraging targeted fine-tuning, memory-efficient methods, or intervention-based techniques, the project aims to enable real-time updates without requiring full retraining.
Transfer the color style of one image to another one.
