Table 3 Schemes summarized.
From: An integrating RAG-LLM and deep Q-network framework for intelligent fish control systems
Fine tuning | RAG-LLM sensors/actuators | RAG-LLM DQN framework |
|---|---|---|
Sensor values and user comments are combined in a prompt text for the RAG-LLM API. | Water quality sensors read values via microcontrollers. | The initial model uses a pre-trained LLM to improve efficiency. |
Questions collected from sensors are stored in a NoSQL database. | Sensor values are formatted into prompts similar to fine-tuned data. | Majority voting selects the optimal policy from LLM and DQN outputs. |
Human experts provide answers via a connected website. | RAG-LLM provides recommendations based on sensor input. | Environmental feedback and recommendations from human experts inform policy adjustments. |
Q&A pairs are retrieved and used for fine-tuning. | Parsed text is converted into control commands. | Â |
