Position-Aware Depth Decay Decoding ($D^3$): Boosting Large Language Model Inference Efficiency

Re-assign community