AEQA-NAT : Adaptive End-to-end Quantization Alignment Training Framework for Non-autoregressive Machine Translation

Xiangyu Qu, Guojing Liu, Liang Li
Proceedings of the 42nd International Conference on Machine Learning, PMLR 267:50848-50864, 2025.

Abstract

Non-autoregressive Transformers (NATs) have garnered significant attention due to their efficient decoding compared to autoregressive methods. However, existing conditional dependency modeling schemes based on masked language modeling introduce a training-inference gap in NATs. For instance, while NATs sample target words during training to enhance input, this condition cannot be met during inference, and simply annealing the sampling rate to zero during training leads to model performance degradation. We demonstrate that this training-inference gap prevents NATs from fully realizing their potential. To address this, we propose an adaptive end-to-end quantization alignment training framework, which introduces a semantic consistency space to adaptively align NAT training, eliminating the need for target information and thereby bridging the training-inference gap.Experimental results demonstrate that our method outperforms most existing fully NAT models, delivering performance on par with Autoregressive Transformer (AT) while being 17.0 times more efficient in inference.

Cite this Paper

BibTeX
@InProceedings{pmlr-v267-qu25e, title = {{AEQA}-{NAT} : Adaptive End-to-end Quantization Alignment Training Framework for Non-autoregressive Machine Translation}, author = {Qu, Xiangyu and Liu, Guojing and Li, Liang}, booktitle = {Proceedings of the 42nd International Conference on Machine Learning}, pages = {50848--50864}, year = {2025}, editor = {Singh, Aarti and Fazel, Maryam and Hsu, Daniel and Lacoste-Julien, Simon and Berkenkamp, Felix and Maharaj, Tegan and Wagstaff, Kiri and Zhu, Jerry}, volume = {267}, series = {Proceedings of Machine Learning Research}, month = {13--19 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v267/main/assets/qu25e/qu25e.pdf}, url = {https://proceedings.mlr.press/v267/qu25e.html}, abstract = {Non-autoregressive Transformers (NATs) have garnered significant attention due to their efficient decoding compared to autoregressive methods. However, existing conditional dependency modeling schemes based on masked language modeling introduce a training-inference gap in NATs. For instance, while NATs sample target words during training to enhance input, this condition cannot be met during inference, and simply annealing the sampling rate to zero during training leads to model performance degradation. We demonstrate that this training-inference gap prevents NATs from fully realizing their potential. To address this, we propose an adaptive end-to-end quantization alignment training framework, which introduces a semantic consistency space to adaptively align NAT training, eliminating the need for target information and thereby bridging the training-inference gap.Experimental results demonstrate that our method outperforms most existing fully NAT models, delivering performance on par with Autoregressive Transformer (AT) while being 17.0 times more efficient in inference.} }
Endnote
%0 Conference Paper %T AEQA-NAT : Adaptive End-to-end Quantization Alignment Training Framework for Non-autoregressive Machine Translation %A Xiangyu Qu %A Guojing Liu %A Liang Li %B Proceedings of the 42nd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2025 %E Aarti Singh %E Maryam Fazel %E Daniel Hsu %E Simon Lacoste-Julien %E Felix Berkenkamp %E Tegan Maharaj %E Kiri Wagstaff %E Jerry Zhu %F pmlr-v267-qu25e %I PMLR %P 50848--50864 %U https://proceedings.mlr.press/v267/qu25e.html %V 267 %X Non-autoregressive Transformers (NATs) have garnered significant attention due to their efficient decoding compared to autoregressive methods. However, existing conditional dependency modeling schemes based on masked language modeling introduce a training-inference gap in NATs. For instance, while NATs sample target words during training to enhance input, this condition cannot be met during inference, and simply annealing the sampling rate to zero during training leads to model performance degradation. We demonstrate that this training-inference gap prevents NATs from fully realizing their potential. To address this, we propose an adaptive end-to-end quantization alignment training framework, which introduces a semantic consistency space to adaptively align NAT training, eliminating the need for target information and thereby bridging the training-inference gap.Experimental results demonstrate that our method outperforms most existing fully NAT models, delivering performance on par with Autoregressive Transformer (AT) while being 17.0 times more efficient in inference.
APA
Qu, X., Liu, G. & Li, L.. (2025). AEQA-NAT : Adaptive End-to-end Quantization Alignment Training Framework for Non-autoregressive Machine Translation. Proceedings of the 42nd International Conference on Machine Learning, in Proceedings of Machine Learning Research 267:50848-50864 Available from https://proceedings.mlr.press/v267/qu25e.html.

Related Material