Nearly all recently-proposed large language models (LLMs) are based upon the decoder-only transformer architecture. But, is this always the best archi

Nearly all recently-proposed large language models (LLMs) are based upon the decoder-only transformer architecture. But, is this always the best archi