Boosting RAG Performance with Expert Human Feedback
Human feedback (HF) is not just useful for generative AI—it’s essential, especially when it comes to models using RAG. A generative AI model uses information from datasets with various documents during training. The data that trained the AI model is set in stone in the model’s parameters; we can’t change it unless we train it again. However, in the world of retrieval-based text and multimodal datasets, there is information we can see and tweak. That is where HF comes in. By providing feedback on what the AI model pulls from its datasets, HF can directly influence the quality of its future responses. Engaging with this process makes humans an active player in the RAG’s development. It adds a new dimension to AI projects: adaptive RAG.
We have explored and implemented naïve, advanced, and modular RAG so far. Now, we will add adaptive RAG to our generative AI toolbox. We know that even the...
