# Dual-Stream Simultaneous Translation via 2D Grid Attention

arXiv: https://arxiv.org/abs/2609.34902 · HTML: https://mlsanity.com/paper/2609.34902/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 32 of its 57 references that are in the map.
- First posted on arXiv 28 Sep 2026.

## Builds on (most cited first, 30 of 32)

- [Attention Is All You Need](https://mlsanity.com/paper/1706.03762/) (2017)
- [Adam: A Method for Stochastic Optimization](https://mlsanity.com/paper/1412.6980/) (2014)
- [LoRA: Low-Rank Adaptation of Large Language Models](https://mlsanity.com/paper/2106.09685/) (2021)
- [Qwen3 Technical Report](https://mlsanity.com/paper/2505.09388/) (2025)
- [RoFormer: Enhanced Transformer with Rotary Position Embedding](https://mlsanity.com/paper/2104.09864/) (2021)
- [FlashAttention: Fast and Memory-Efficient Exact Attention with IO-Awareness](https://mlsanity.com/paper/2205.14135/) (2022)
- [COMET: A Neural Framework for MT Evaluation](https://mlsanity.com/paper/2009.09025/) (2020)
- [Massively Multilingual Neural Machine Translation in the Wild: Findings and Challenges](https://mlsanity.com/paper/1907.05019/) (2019)
- [WikiMatrix: Mining 135M Parallel Sentences in 1620 Language Pairs from Wikipedia](https://mlsanity.com/paper/1907.05791/) (2019)
- [Speculative Decoding: Exploiting Speculative Execution for Accelerating Seq2seq Generation](https://mlsanity.com/paper/2203.16487/) (2022)
- [CIF: Continuous Integrate-and-Fire for End-to-End Speech Recognition](https://mlsanity.com/paper/1905.11235/) (2019)
- [Efficient Wait-k Models for Simultaneous Machine Translation](https://mlsanity.com/paper/2005.08595/) (2020)
- [Curriculum Learning for Domain Adaptation in Neural Machine Translation](https://mlsanity.com/paper/1905.05816/) (2019)
- [Incremental Decoding and Training Methods for Simultaneous Translation in Neural Machine Translation](https://mlsanity.com/paper/1806.03661/) (2018)
- [Simultaneous Translation Policies: From Fixed to Adaptive](https://mlsanity.com/paper/2004.13169/) (2020)
- [Simpler and Faster Learning of Adaptive Policies for Simultaneous Translation](https://mlsanity.com/paper/1909.01559/) (2019)
- [Universal Simultaneous Machine Translation with Mixture-of-Experts Wait-k Policy](https://mlsanity.com/paper/2109.05238/) (2021)
- [Re-translation versus Streaming for Simultaneous Translation](https://mlsanity.com/paper/2004.03643/) (2020)
- [Future-Guided Incremental Transformer for Simultaneous Translation](https://mlsanity.com/paper/2012.12465/) (2020)
- [Learning When to Translate for Streaming Speech](https://mlsanity.com/paper/2109.07368/) (2021)
- [Does Simultaneous Speech Translation need Simultaneous Models?](https://mlsanity.com/paper/2204.03783/) (2022)
- [Translate-Distill: Learning Cross-Language Dense Retrieval by Translation and Distillation](https://mlsanity.com/paper/2401.04810/) (2024)
- [Improving Simultaneous Translation by Incorporating Pseudo-References with Fewer Reorderings](https://mlsanity.com/paper/2010.11247/) (2020)
- [Non-autoregressive Streaming Transformer for Simultaneous Translation](https://mlsanity.com/paper/2310.14883/) (2023)
- [Decoder-only Streaming Transformer for Simultaneous Translation](https://mlsanity.com/paper/2406.03878/) (2024)
- [Simultaneous Machine Translation with Large Language Models](https://mlsanity.com/paper/2309.06706/) (2023)
- [Adaptive Policy with Wait-k Model for Simultaneous Translation](https://mlsanity.com/paper/2310.14853/) (2023)
- [StreamUni: Achieving Streaming Speech Translation with a Unified Large Speech-Language Model](https://mlsanity.com/paper/2507.07803/) (2025)
- [LLMs Can Achieve High-quality Simultaneous Machine Translation as Efficiently as Offline](https://mlsanity.com/paper/2504.09570/) (2025)
- [Monotonic Simultaneous Translation with Chunk-wise Reordering and Refinement](https://mlsanity.com/paper/2110.09646/) (2021)

## Abstract (arXiv)

Simultaneous machine translation must generate target tokens before the source input is complete. Existing approaches address this through post-hoc read-write policies, leaving the attention mechanism unaware of bidirectional stream dependencies. We propose a dual-stream attention framework that represents source and target streams as a two-dimensional grid of hidden states and models their interaction through four structurally distinct attention types merged via joint QK Softmax normalization. Two approximations---broadcast and Hadamard---reduce the per-layer complexity from O(X^2Y+XY^2) to O(X^2+Y^2+XY) with provably decaying error. Training uses a self-guided loop: a per-cell loss heatmap drives dynamic-programming path recovery, which generates read/write decision supervision labels without external alignment. An incremental KV cache with anchored rotary position embeddings enables efficient streaming inference. On Chinese-to-English simultaneous translation, the proposed model outperforms the Wait-k baseline by +5.66 BLEURT and +10.36 COMET at comparable latency, and surpasses the non-streaming reference on COMET at a fraction of the response delay.