Knowledge Infused Decoding

Knowledge Infused Decoding (KID) is a decoding algorithm that infuses knowledge (from Wikipedia) into each step decoding of text generation.

KID has been described in the following paper:

@inproceedings{liu2022knowledge,
    author = {Liu, Ruibo and Zheng, Guoqing and Gupta, Shashank and Gaonkar, Radhika and Gao, Chongyang and Vosoughi, Soroush and Shokouhi, Milad and Awadallah, Ahmed H.},
    title = {Knowledge Infused Decoding},
    booktitle = {International Conference on Learning Representations (ICLR)},
    year = {2022},
    month = {April},
    publisher = {ICLR},
    url = {https://www.microsoft.com/en-us/research/publication/knowledge-infused-decoding/},
}