Speculate, then Collaborate: Fusing Knowledge of Language Models during Decoding

Devs

Speculate, then Collaborate: Fusing Knowledge of Language Models during Decoding | Read Paper on Bytez