Yuchen Liu

Yuchen Liu, Natasha Ong, Kaiyan Peng et al. · meta-ai

We present Multiscale Multiview Vision Transformers (MMViT), which introduces multiscale feature maps and multiview encodings to transformer models. Our model encodes different views of the input signal and builds several channel-resolution feature stages to process the multiple views of the input at different resolutions in parallel. At each scale stage, we use a cross-attention block to fuse information across different views. This enables the MMViT model to acquire complex high-dimensional representations of the input at different resolutions. The proposed model can serve as a backbone model in multiple domains. We demonstrate the effectiveness of MMViT on audio and image classification tasks, achieving state-of-the-art results.

Chen Wang, Yuchen Liu, Boxing Chen et al.

End-to-end Speech Translation (ST) aims at translating the source language speech into target language text without generating the intermediate transcriptions. However, the training of end-to-end methods relies on parallel ST data, which are difficult and expensive to obtain. Fortunately, the supervised data for automatic speech recognition (ASR) and machine translation (MT) are usually more accessible, making zero-shot speech translation a potential direction. Existing zero-shot methods fail to align the two modalities of speech and text into a shared semantic space, resulting in much worse performance compared to the supervised ST methods. In order to enable zero-shot ST, we propose a novel Discrete Cross-Modal Alignment (DCMA) method that employs a shared discrete vocabulary space to accommodate and match both modalities of speech and text. Specifically, we introduce a vector quantization module to discretize the continuous representations of speech and text into a finite set of virtual tokens, and use ASR data to map corresponding speech and text to the same virtual token in a shared codebook. This way, source language speech can be embedded in the same semantic space as the source language text, which can be then transformed into target language text with an MT module. Experiments on multiple language pairs demonstrate that our zero-shot ST method significantly improves the SOTA, and even performers on par with the strong supervised ST baselines.

Huiping Zhuang, Yuchen Liu, Run He et al.

Online Class Incremental Learning (OCIL) aims to train models incrementally, where data arrive in mini-batches, and previous data are not accessible. A major challenge in OCIL is Catastrophic Forgetting, i.e., the loss of previously learned knowledge. Among existing baselines, replay-based methods show competitive results but requires extra memory for storing exemplars, while exemplar-free (i.e., data need not be stored for replay in production) methods are resource-friendly but often lack accuracy. In this paper, we propose an exemplar-free approach--Forward-only Online Analytic Learning (F-OAL). Unlike traditional methods, F-OAL does not rely on back-propagation and is forward-only, significantly reducing memory usage and computational time. Cooperating with a pre-trained frozen encoder with Feature Fusion, F-OAL only needs to update a linear classifier by recursive least square. This approach simultaneously achieves high accuracy and low resource consumption. Extensive experiments on benchmark datasets demonstrate F-OAL's robust performance in OCIL scenarios. Code is available at https://github.com/liuyuchen-cz/F-OAL.

Sicheng Mo, Thao Nguyen, Xun Huang et al.

We propose X-Fusion, a framework that extends pretrained Large Language Models (LLMs) for multimodal tasks while preserving their language capabilities. X-Fusion employs a dual-tower design with modality-specific weights, keeping the LLM's parameters frozen while integrating vision-specific information for both understanding and generation. Our experiments demonstrate that X-Fusion consistently outperforms alternative architectures on both image-to-text and text-to-image tasks. We find that incorporating understanding-focused data improves generation quality, reducing image data noise enhances overall performance, and feature alignment accelerates convergence for smaller models but has minimal impact on larger ones. Our findings provide valuable insights into building efficient unified multimodal models.

Donglei Yu, Xiaomian Kang, Yuchen Liu et al.

Simultaneous Machine Translation (SiMT) generates target outputs while receiving stream source inputs and requires a read/write policy to decide whether to wait for the next source token or generate a new target token, whose decisions form a \textit{decision path}. Existing SiMT methods, which learn the policy by exploring various decision paths in training, face inherent limitations. These methods not only fail to precisely optimize the policy due to the inability to accurately assess the individual impact of each decision on SiMT performance, but also cannot sufficiently explore all potential paths because of their vast number. Besides, building decision paths requires unidirectional encoders to simulate streaming source inputs, which impairs the translation quality of SiMT models. To solve these issues, we propose \textbf{S}elf-\textbf{M}odifying \textbf{S}tate \textbf{M}odeling (SM$^2$), a novel training paradigm for SiMT task. Without building decision paths, SM$^2$ individually optimizes decisions at each state during training. To precisely optimize the policy, SM$^2$ introduces Self-Modifying process to independently assess and adjust decisions at each state. For sufficient exploration, SM$^2$ proposes Prefix Sampling to efficiently traverse all potential states. Moreover, SM$^2$ ensures compatibility with bidirectional encoders, thus achieving higher translation quality. Experiments show that SM$^2$ outperforms strong baselines. Furthermore, SM$^2$ allows offline machine translation models to acquire SiMT ability with fine-tuning.

Chen Wang, Minpeng Liao, Zhongqiang Huang et al.

The emergence of large language models (LLMs) has sparked significant interest in extending their remarkable language capabilities to speech. However, modality alignment between speech and text still remains an open problem. Current solutions can be categorized into two strategies. One is a cascaded approach where outputs (tokens or states) of a separately trained speech recognition system are used as inputs for LLMs, which limits their potential in modeling alignment between speech and text. The other is an end-to-end approach that relies on speech instruction data, which is very difficult to collect in large quantities. In this paper, we address these issues and propose the BLSP approach that Bootstraps Language-Speech Pre-training via behavior alignment of continuation writing. We achieve this by learning a lightweight modality adapter between a frozen speech encoder and an LLM, ensuring that the LLM exhibits the same generation behavior regardless of the modality of input: a speech segment or its transcript. The training process can be divided into two steps. The first step prompts an LLM to generate texts with speech transcripts as prefixes, obtaining text continuations. In the second step, these continuations are used as supervised signals to train the modality adapter in an end-to-end manner. We demonstrate that this straightforward process can extend the capabilities of LLMs to speech, enabling speech recognition, speech translation, spoken language understanding, and speech conversation, even in zero-shot cross-lingual scenarios.

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.

Long Zhou, Yuchen Liu, Jiajun Zhang et al.

Current Neural Machine Translation (NMT) employs a language-specific encoder to represent the source sentence and adopts a language-specific decoder to generate target translation. This language-dependent design leads to large-scale network parameters and makes the duality of the parallel data underutilized. To address the problem, we propose in this paper a language-independent representor to replace the encoder and decoder by using weight sharing. This shared representor can not only reduce large portion of network parameters, but also facilitate us to fully explore the language duality by jointly training source-to-target, target-to-source, left-to-right and right-to-left translations within a multi-task learning framework. Experiments show that our proposed framework can obtain significant improvements over conventional NMT models on resource-rich and low-resource translation tasks with only a quarter of parameters.