The Augmentation is actually related to generation part, not augmented retrieval. Otherwise it would be called "Augmented Retrieval Generation" . I checked the original paper from Meta and they also say "Our work aims to expand the space of possible tasks with a single, unified architecture, by learning a retrieval module to augment pre-trained, generative language models". Though I agree the prompt is also augmented with the retrieved information to get an Augmented response
@AIBites6 ай бұрын
ah I see.. :) thanks for the pointer. I will check the paper again. You guys are awesome and this community is great! 🙂
@B_knows_A_R_D-xh5lo3 ай бұрын
😊😊😊😊
@InquilineKea6 ай бұрын
Is it a form of continual/lifelong learning
@AIBites6 ай бұрын
We may say so. But note quite continual learning. With RAG you are just addinga. memory module which can be refreshed with up-to-date data. So, the LLM always gets up-to-date information or context. You do NOT train the model. In continual learning, on the other hand, you train the model.