Re:Training and Aligning LLMs with RLHF and RLHF Alternatives
by westurner on Monday September 11, 2023 @02:36AM
(#37463686)
When the next token is a URL, and the URL does not match the preceding anchor text.
Additional layers of these 'LLMs' could read the responses and determine whether their premises are valid and their logic is sound as necessary to support the presented conclusion(s), and then just suggest a different citation URL for the preceding text.
Additional layers of these 'LLMs' could read the responses and determine whether their premises are valid and their logic is sound as necessary to support the presented conclusion(s), and then just suggest a different citation URL for the preceding text.
"#StructuredPremises"