"Transformers rely on self-attention and it evaluates how each word relates to the rest of the sequence, not just its neighbouring words."
A to the point line by @arpit_bhayani in his article "How LLM inference works".
"Small LLM don't understand the instructions as literally as big model"
Tried using a small model locally and understood this point very clearly.
#ai#ml