@inproceedings{goyal-etal-2017-differentiable,
    title = "Differentiable Scheduled Sampling for Credit Assignment",
    author = "Goyal, Kartik  and
      Dyer, Chris  and
      Berg-Kirkpatrick, Taylor",
    editor = "Barzilay, Regina  and
      Kan, Min-Yen",
    booktitle = "Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics (Volume 2: Short Papers)",
    month = jul,
    year = "2017",
    address = "Vancouver, Canada",
    publisher = "Association for Computational Linguistics",
    url = "https://preview.aclanthology.org/add-emnlp-2024-awards/P17-2058/",
    doi = "10.18653/v1/P17-2058",
    pages = "366--371",
    abstract = "We demonstrate that a continuous relaxation of the argmax operation can be used to create a differentiable approximation to greedy decoding in sequence-to-sequence (seq2seq) models. By incorporating this approximation into the scheduled sampling training procedure{--}a well-known technique for correcting exposure bias{--}we introduce a new training objective that is continuous and differentiable everywhere and can provide informative gradients near points where previous decoding decisions change their value. By using a related approximation, we also demonstrate a similar approach to sampled-based training. We show that our approach outperforms both standard cross-entropy training and scheduled sampling procedures in two sequence prediction tasks: named entity recognition and machine translation."
}