Ana Kavramlar
Knowledge Cutoff
The date beyond which a model has no training data and is unaware of events.
Because LLMs are trained on static snapshots of data, they have a training cutoff date — after which they know nothing about new events, models, or research. For time-sensitive applications, grounding via RAG or tool use is required. Knowledge cutoffs typically lag behind model release by 6–18 months.
İlgili Terimler
Grounding
Connecting model responses to verified, real-world information sources.
RAG (Retrieval-Augmented Generation)
Enhancing model responses by fetching relevant documents from an external knowledge base at query time.
Hallucination
When a model generates plausible-sounding but factually incorrect information.
Pre-Training
The initial large-scale training phase where a model learns language from massive text corpora.