A clerk-bot must cite this morning’s notice, which did not exist when the LLM was trained. The system first finds relevant passages, then conditions generation on them. What is that pattern?
Select an answer to reveal the explanation.
Short Explanation
This morning’s notice did not exist when the model was trained. The system finds the relevant passages, then writes from them. That is RAG: ground generation in an updatable source, not in the weights alone. Daily pretraining, a BERT blank, and a dock cluster are not that pattern.
Full Explanation
RAG retrieves current passages and then generates from them, so facts that post-date training can still be cited. Daily pretraining, MLM, and graph communities are not that pattern.