Unified segment-to-segment framework for simultaneous sequence generation

S Zhang, Y Feng - Advances in Neural Information …, 2024 - proceedings.neurips.cc
Simultaneous sequence generation is a pivotal task for real-time scenarios, such as
streaming speech recognition, simultaneous machine translation and simultaneous speech …

Glancing future for simultaneous machine translation

S Guo, S Zhang, Y Feng - ICASSP 2024-2024 IEEE …, 2024 - ieeexplore.ieee.org
Simultaneous machine translation (SiMT) outputs translation while reading the source
sentence. Unlike conventional sequence-to-sequence (seq2seq) training, existing SiMT …

SimulPL: Aligning Human Preferences in Simultaneous Machine Translation

D Yu, Y Zhao, J Zhu, Y Xu, Y Zhou, C Zong - arxiv preprint arxiv …, 2025 - arxiv.org
Simultaneous Machine Translation (SiMT) generates translations while receiving streaming
source inputs. This requires the SiMT model to learn a read/write policy, deciding when to …

Large Language Models Are Read/Write Policy-Makers for Simultaneous Generation

S Guo, S Zhang, Z Ma, Y Feng - arxiv preprint arxiv:2501.00868, 2025 - arxiv.org
Simultaneous generation models write generation results while reading streaming inputs,
necessitating a policy-maker to determine the appropriate output timing. Existing …

Decoder-only Streaming Transformer for Simultaneous Translation

S Guo, S Zhang, Y Feng - arxiv preprint arxiv:2406.03878, 2024 - arxiv.org
Simultaneous Machine Translation (SiMT) generates translation while reading source
tokens, essentially producing the target prefix based on the source prefix. To achieve good …

Agent-SiMT: Agent-assisted Simultaneous Machine Translation with Large Language Models

S Guo, S Zhang, Z Ma, M Zhang, Y Feng - arxiv preprint arxiv:2406.06910, 2024 - arxiv.org
Simultaneous Machine Translation (SiMT) generates target translations while reading the
source sentence. It relies on a policy to determine the optimal timing for reading sentences …

PsFuture: A Pseudo-Future-based Zero-Shot Adaptive Policy for Simultaneous Machine Translation

L Zhao, J Li, Z Zeng - arxiv preprint arxiv:2410.04075, 2024 - arxiv.org
Simultaneous Machine Translation (SiMT) requires target tokens to be generated in real-
time as streaming source tokens are consumed. Traditional approaches to SiMT typically …