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

Re-assign community