Dataset for Dynamic Context Tuning in Retrieval-Augmented Generation

Retrieval-Augmented Generation (RAG) has significantly enhanced large language models (LLMs) by grounding their outputs in external tools and knowledge sources. However, existing RAG systems are typically limited to static, single-turn interactions with fixed toolsets, making them less effective in dynamic domains such as healthcare and smart homes, where user intent, available tools, and context continuously evolve. We introduce Dynamic Context Tuning (DCT), a lightweight framework that extends RAG to support multi-turn dialogues and evolving tool environments without requiring retraining. DCT integrates an attention-based context cache to track relevant past information, LoRA-based retrieval to dynamically select domain-specific tools, and efficient context compression to keep inputs within LLM context limits. Experiments on both synthetic and real-world benchmarks demonstrate that DCT improves plan accuracy by 14% and reduces hallucinations by 37%, while matching GPT-4\u2019s performance at a significantly lower cost. Furthermore, DCT generalizes effectively to previously unseen tools, enabling scalable and adaptable AI assistants across diverse dynamic environments.