LLMs don't have "perception" at all, they only ever output a likely text complet...

		otabdeveloper4 9 months ago \| parent \| context \| favorite \| on: Meta's Llama 3.1 can recall 42 percent of the firs... LLMs don't have "perception" at all, they only ever output a likely text completion token.