Make Some Noise: Unlocking Language Model Parallel Inference Capability through Noisy Training

Open in new window