Zhongjun He

Pengzhi Gao, Zhongjun He, Hua Wu et al.

We introduce Bi-SimCut: a simple but effective training strategy to boost neural machine translation (NMT) performance. It consists of two procedures: bidirectional pretraining and unidirectional finetuning. Both procedures utilize SimCut, a simple regularization method that forces the consistency between the output distributions of the original and the cutoff sentence pairs. Without leveraging extra dataset via back-translation or integrating large-scale pretrained model, Bi-SimCut achieves strong translation performance across five translation benchmarks (data sizes range from 160K to 20.2M): BLEU scores of 31.16 for en -> de and 38.37 for de -> en on the IWSLT14 dataset, 30.78 for en -> de and 35.15 for de -> en on the WMT14 dataset, and 27.17 for zh -> en on the WMT17 dataset. SimCut is not a new method, but a version of Cutoff (Shen et al., 2020) simplified and adapted for NMT, and it could be considered as a perturbation-based method. Given the universality and simplicity of SimCut and Bi-SimCut, we believe they can serve as strong baselines for future NMT research.

Pengzhi Gao, Liwen Zhang, Zhongjun He et al.

Multilingual sentence representations are the foundation for similarity-based bitext mining, which is crucial for scaling multilingual neural machine translation (NMT) system to more languages. In this paper, we introduce MuSR: a one-for-all Multilingual Sentence Representation model that supports more than 220 languages. Leveraging billions of English-centric parallel corpora, we train a multilingual Transformer encoder, coupled with an auxiliary Transformer decoder, by adopting a multilingual NMT framework with CrossConST, a cross-lingual consistency regularization technique proposed in Gao et al. (2023). Experimental results on multilingual similarity search and bitext mining tasks show the effectiveness of our approach. Specifically, MuSR achieves superior performance over LASER3 (Heffernan et al., 2022) which consists of 148 independent multilingual sentence encoders.

Pengzhi Gao, Ruiqing Zhang, Zhongjun He et al.

Consistency regularization methods, such as R-Drop (Liang et al., 2021) and CrossConST (Gao et al., 2023), have achieved impressive supervised and zero-shot performance in the neural machine translation (NMT) field. Can we also boost end-to-end (E2E) speech-to-text translation (ST) by leveraging consistency regularization? In this paper, we conduct empirical studies on intra-modal and cross-modal consistency and propose two training strategies, SimRegCR and SimZeroCR, for E2E ST in regular and zero-shot scenarios. Experiments on the MuST-C benchmark show that our approaches achieve state-of-the-art (SOTA) performance in most translation directions. The analyses prove that regularization brought by the intra-modal consistency, instead of modality gap, is crucial for the regular E2E ST, and the cross-modal consistency could close the modality gap and boost the zero-shot E2E ST performance.

Haifeng Wang, Hua Wu, Tian Wu et al.

In this report, we introduce ERNIE 5.0, a natively autoregressive foundation model desinged for unified multimodal understanding and generation across text, image, video, and audio. All modalities are trained from scratch under a unified next-group-of-tokens prediction objective, based on an ultra-sparse mixture-of-experts (MoE) architecture with modality-agnostic expert routing. To address practical challenges in large-scale deployment under diverse resource constraints, ERNIE 5.0 adopts a novel elastic training paradigm. Within a single pre-training run, the model learns a family of sub-models with varying depths, expert capacities, and routing sparsity, enabling flexible trade-offs among performance, model size, and inference latency in memory- or time-constrained scenarios. Moreover, we systematically address the challenges of scaling reinforcement learning to unified foundation models, thereby guaranteeing efficient and stable post-training under ultra-sparse MoE architectures and diverse multimodal settings. Extensive experiments demonstrate that ERNIE 5.0 achieves strong and balanced performance across multiple modalities. To the best of our knowledge, among publicly disclosed models, ERNIE 5.0 represents the first production-scale realization of a trillion-parameter unified autoregressive model that supports both multimodal understanding and generation. To facilitate further research, we present detailed visualizations of modality-agnostic expert routing in the unified model, alongside comprehensive empirical analysis of elastic training, aiming to offer profound insights to the community.

Pengzhi Gao, Zhongjun He, Hua Wu et al.

The training paradigm for machine translation has gradually shifted, from learning neural machine translation (NMT) models with extensive parallel corpora to instruction finetuning on multilingual large language models (LLMs) with high-quality translation pairs. In this paper, we focus on boosting many-to-many multilingual translation of LLMs with an emphasis on zero-shot translation directions. We demonstrate that prompt strategies adopted during finetuning are crucial to zero-shot translation and introduce a cross-lingual consistency regularization, XConST, to bridge the representation gap among different languages and improve zero-shot translation performance. XConST is not a new method, but a version of CrossConST (Gao et al., 2023a) adapted for translation instruction finetuning with LLMs. Experimental results on ALMA (Xu et al., 2023), Tower (Team, 2024), and LLaMA-2 (Touvron et al., 2023) show that our approach consistently improves translation performance. Our implementations are available at https://github.com/gpengzhi/CrossConST-LLM.

Hao Xiong, Zhongjun He, Xiaoguang Hu et al.

Attention-based Encoder-Decoder has the effective architecture for neural machine translation (NMT), which typically relies on recurrent neural networks (RNN) to build the blocks that will be lately called by attentive reader during the decoding process. This design of encoder yields relatively uniform composition on source sentence, despite the gating mechanism employed in encoding RNN. On the other hand, we often hope the decoder to take pieces of source sentence at varying levels suiting its own linguistic structure: for example, we may want to take the entity name in its raw form while taking an idiom as a perfectly composed unit. Motivated by this demand, we propose Multi-channel Encoder (MCE), which enhances encoding components with different levels of composition. More specifically, in addition to the hidden state of encoding RNN, MCE takes 1) the original word embedding for raw encoding with no composition, and 2) a particular design of external memory in Neural Turing Machine (NTM) for more complex composition, while all three encoding strategies are properly blended during decoding. Empirical study on Chinese-English translation shows that our model can improve by 6.52 BLEU points upon a strong open source NMT system: DL4MT1. On the WMT14 English- French task, our single shallow system achieves BLEU=38.8, comparable with the state-of-the-art deep models.

Mengyu Bu, Shaolei Zhang, Zhongjun He et al.

Multilingual large language models (LLMs) possess impressive multilingual understanding and generation capabilities. However, their performance and cross-lingual alignment often lag for non-dominant languages. A common solution is to fine-tune LLMs on large-scale and more balanced multilingual corpus, but such approaches often lead to imprecise alignment and suboptimal knowledge transfer, struggling with limited improvements across languages. In this paper, we propose AlignX to bridge the multilingual performance gap, which is a two-stage representation-level framework for enhancing multilingual performance of pre-trained LLMs. In the first stage, we align multilingual representations with multilingual semantic alignment and language feature integration. In the second stage, we stimulate the multilingual capability of LLMs via multilingual instruction fine-tuning. Experimental results on several pre-trained LLMs demonstrate that our approach enhances LLMs' multilingual general and cross-lingual generation capability. Further analysis indicates that AlignX brings the multilingual representations closer and improves the cross-lingual alignment.

Minlong Peng, Jingyi Yang, Zhongjun He et al.

This work presents a swift method to assess the efficacy of particular types of instruction-tuning data, utilizing just a handful of probe examples and eliminating the need for model retraining. This method employs the idea of gradient-based data influence estimation, analyzing the gradient projections of probe examples from the chosen strategy onto evaluation examples to assess its advantages. Building upon this method, we conducted three swift studies to investigate the potential of Chain-of-thought (CoT) data, query clarification data, and response evaluation data in enhancing model generalization. Subsequently, we embarked on a validation study to corroborate the findings of these swift studies. In this validation study, we developed training datasets tailored to each studied strategy and compared model performance with and without the use of these datasets. The results of the validation study aligned with the findings of the swift studies, validating the efficacy of our proposed method.

Pengzhi Gao, Liwen Zhang, Zhongjun He et al.

The multilingual neural machine translation (NMT) model has a promising capability of zero-shot translation, where it could directly translate between language pairs unseen during training. For good transfer performance from supervised directions to zero-shot directions, the multilingual NMT model is expected to learn universal representations across different languages. This paper introduces a cross-lingual consistency regularization, CrossConST, to bridge the representation gap among different languages and boost zero-shot translation performance. The theoretical analysis shows that CrossConST implicitly maximizes the probability distribution for zero-shot translation, and the experimental results on both low-resource and high-resource benchmarks show that CrossConST consistently improves the translation performance. The experimental analysis also proves that CrossConST could close the sentence representation gap and better align the representation space. Given the universality and simplicity of CrossConST, we believe it can serve as a strong baseline for future multilingual NMT research.

Jicheng Li, Pengzhi Gao, Xuanfu Wu et al.

Diverse machine translation aims at generating various target language translations for a given source language sentence. Leveraging the linear relationship in the sentence latent space introduced by the mixup training, we propose a novel method, MixDiversity, to generate different translations for the input sentence by linearly interpolating it with different sentence pairs sampled from the training corpus when decoding. To further improve the faithfulness and diversity of the translations, we propose two simple but effective approaches to select diverse sentence pairs in the training corpus and adjust the interpolation weight for each pair correspondingly. Moreover, by controlling the interpolation weight, our method can achieve the trade-off between faithfulness and diversity without any additional training, which is required in most of the previous methods. Experiments on WMT'16 en-ro, WMT'14 en-de, and WMT'17 zh-en are conducted to show that our method substantially outperforms all previous diverse machine translation methods.

Ruiqing Zhang, Xiyang Wang, Chuanqiang Zhang et al.

This paper presents BSTC (Baidu Speech Translation Corpus), a large-scale Chinese-English speech translation dataset. This dataset is constructed based on a collection of licensed videos of talks or lectures, including about 68 hours of Mandarin data, their manual transcripts and translations into English, as well as automated transcripts by an automatic speech recognition (ASR) model. We have further asked three experienced interpreters to simultaneously interpret the testing talks in a mock conference setting. This corpus is expected to promote the research of automatic simultaneous translation as well as the development of practical systems. We have organized simultaneous translation tasks and used this corpus to evaluate automatic simultaneous translation systems.

Yuchen Liu, Jiajun Zhang, Hao Xiong et al.

Speech-to-text translation (ST), which translates source language speech into target language text, has attracted intensive attention in recent years. Compared to the traditional pipeline system, the end-to-end ST model has potential benefits of lower latency, smaller model size, and less error propagation. However, it is notoriously difficult to implement such a model without transcriptions as intermediate. Existing works generally apply multi-task learning to improve translation quality by jointly training end-to-end ST along with automatic speech recognition (ASR). However, different tasks in this method cannot utilize information from each other, which limits the improvement. Other works propose a two-stage model where the second model can use the hidden state from the first one, but its cascade manner greatly affects the efficiency of training and inference process. In this paper, we propose a novel interactive attention mechanism which enables ASR and ST to perform synchronously and interactively in a single model. Specifically, the generation of transcriptions and translations not only relies on its previous outputs but also the outputs predicted in the other task. Experiments on TED speech translation corpora have shown that our proposed model can outperform strong baselines on the quality of speech translation and achieve better speech recognition performances as well.

Tianchi Bi, Hao Xiong, Zhongjun He et al.

Conventional Neural Machine Translation (NMT) models benefit from the training with an additional agent, e.g., dual learning, and bidirectional decoding with one agent decoding from left to right and the other decoding in the opposite direction. In this paper, we extend the training framework to the multi-agent scenario by introducing diverse agents in an interactive updating process. At training time, each agent learns advanced knowledge from others, and they work together to improve translation quality. Experimental results on NIST Chinese-English, IWSLT 2014 German-English, WMT 2014 English-German and large-scale Chinese-English translation tasks indicate that our approach achieves absolute improvements over the strong baseline systems and shows competitive performance on all tasks.

Hao Xiong, Ruiqing Zhang, Chuanqiang Zhang et al.

In this paper, we present DuTongChuan, a novel context-aware translation model for simultaneous interpreting. This model allows to constantly read streaming text from the Automatic Speech Recognition (ASR) model and simultaneously determine the boundaries of Information Units (IUs) one after another. The detected IU is then translated into a fluent translation with two simple yet effective decoding strategies: partial decoding and context-aware decoding. In practice, by controlling the granularity of IUs and the size of the context, we can get a good trade-off between latency and translation quality easily. Elaborate evaluation from human translators reveals that our system achieves promising translation quality (85.71% for Chinese-English, and 86.36% for English-Chinese), specially in the sense of surprisingly good discourse coherence. According to an End-to-End (speech-to-speech simultaneous interpreting) evaluation, this model presents impressive performance in reducing latency (to less than 3 seconds at most times). Furthermore, we successfully deploy this model in a variety of Baidu's products which have hundreds of millions of users, and we release it as a service in our AI platform.

Yuchen Liu, Hao Xiong, Zhongjun He et al.

End-to-end speech translation (ST), which directly translates from source language speech into target language text, has attracted intensive attentions in recent years. Compared to conventional pipeline systems, end-to-end ST models have advantages of lower latency, smaller model size and less error propagation. However, the combination of speech recognition and text translation in one model is more difficult than each of these two tasks. In this paper, we propose a knowledge distillation approach to improve ST model by transferring the knowledge from text translation model. Specifically, we first train a text translation model, regarded as a teacher model, and then ST model is trained to learn output probabilities from teacher model through knowledge distillation. Experiments on English- French Augmented LibriSpeech and English-Chinese TED corpus show that end-to-end ST is possible to implement on both similar and dissimilar language pairs. In addition, with the instruction of teacher model, end-to-end ST model can gain significant improvements by over 3.5 BLEU points.

Hao Xiong, Zhongjun He, Hua Wu et al.

Discourse coherence plays an important role in the translation of one text. However, the previous reported models most focus on improving performance over individual sentence while ignoring cross-sentence links and dependencies, which affects the coherence of the text. In this paper, we propose to use discourse context and reward to refine the translation quality from the discourse perspective. In particular, we generate the translation of individual sentences at first. Next, we deliberate the preliminary produced translations, and train the model to learn the policy that produces discourse coherent text by a reward teacher. Practical results on multiple discourse test datasets indicate that our model significantly improves the translation quality over the state-of-the-art baseline system by +1.23 BLEU score. Moreover, our model generates more discourse coherent text and obtains +2.2 BLEU improvements when evaluated by discourse metrics.

Mingbo Ma, Liang Huang, Hao Xiong et al.

Simultaneous translation, which translates sentences before they are finished, is useful in many scenarios but is notoriously difficult due to word-order differences. While the conventional seq-to-seq framework is only suitable for full-sentence translation, we propose a novel prefix-to-prefix framework for simultaneous translation that implicitly learns to anticipate in a single translation model. Within this framework, we present a very simple yet surprisingly effective wait-k policy trained to generate the target sentence concurrently with the source sentence, but always k words behind. Experiments show our strategy achieves low latency and reasonable quality (compared to full-sentence translation) on 4 directions: zh<->en and de<->en.

Hairong Liu, Mingbo Ma, Liang Huang et al.

Neural machine translation (NMT) is notoriously sensitive to noises, but noises are almost inevitable in practice. One special kind of noise is the homophone noise, where words are replaced by other words with similar pronunciations. We propose to improve the robustness of NMT to homophone noises by 1) jointly embedding both textual and phonetic information of source sentences, and 2) augmenting the training dataset with homophone noises. Interestingly, to achieve better translation quality and more robustness, we found that most (though not all) weights should be put on the phonetic rather than textual information. Experiments show that our method not only significantly improves the robustness of NMT to homophone noises, but also surprisingly improves the translation quality on some clean test sets.

Yong Cheng, Wei Xu, Zhongjun He et al.

While end-to-end neural machine translation (NMT) has made remarkable progress recently, NMT systems only rely on parallel corpora for parameter estimation. Since parallel corpora are usually limited in quantity, quality, and coverage, especially for low-resource languages, it is appealing to exploit monolingual corpora to improve NMT. We propose a semi-supervised approach for training NMT models on the concatenation of labeled (parallel corpora) and unlabeled (monolingual corpora) data. The central idea is to reconstruct the monolingual corpora using an autoencoder, in which the source-to-target and target-to-source translation models serve as the encoder and decoder, respectively. Our approach can not only exploit the monolingual corpora of the target language, but also of the source language. Experiments on the Chinese-English dataset show that our approach achieves significant improvements over state-of-the-art SMT and NMT systems.

Yong Cheng, Shiqi Shen, Zhongjun He et al.

The attentional mechanism has proven to be effective in improving end-to-end neural machine translation. However, due to the intricate structural divergence between natural languages, unidirectional attention-based models might only capture partial aspects of attentional regularities. We propose agreement-based joint training for bidirectional attention-based end-to-end neural machine translation. Instead of training source-to-target and target-to-source translation models independently,our approach encourages the two complementary models to agree on word alignment matrices on the same training data. Experiments on Chinese-English and English-French translation tasks show that agreement-based joint training significantly improves both alignment and translation quality over independent training.

Shiqi Shen, Yong Cheng, Zhongjun He et al.

We propose minimum risk training for end-to-end neural machine translation. Unlike conventional maximum likelihood estimation, minimum risk training is capable of optimizing model parameters directly with respect to arbitrary evaluation metrics, which are not necessarily differentiable. Experiments show that our approach achieves significant improvements over maximum likelihood estimation on a state-of-the-art neural machine translation system across various languages pairs. Transparent to architectures, our approach can be applied to more neural networks and potentially benefit more NLP tasks.