Dong Zhang

Xin Zhang, Dong Zhang, Shimin Li et al.

Current speech large language models build upon discrete speech representations, which can be categorized into semantic tokens and acoustic tokens. However, existing speech tokens are not specifically designed for speech language modeling. To assess the suitability of speech tokens for building speech language models, we established the first benchmark, SLMTokBench. Our results indicate that neither semantic nor acoustic tokens are ideal for this purpose. Therefore, we propose SpeechTokenizer, a unified speech tokenizer for speech large language models. SpeechTokenizer adopts the Encoder-Decoder architecture with residual vector quantization (RVQ). Unifying semantic and acoustic tokens, SpeechTokenizer disentangles different aspects of speech information hierarchically across different RVQ layers. Furthermore, We construct a Unified Speech Language Model (USLM) leveraging SpeechTokenizer. Experiments show that SpeechTokenizer performs comparably to EnCodec in speech reconstruction and demonstrates strong performance on the SLMTokBench benchmark. Also, USLM outperforms VALL-E in zero-shot Text-to-Speech tasks. Code and models are available at https://github.com/ZhangXInFD/SpeechTokenizer/.

Zhaowei Li, Wei Wang, YiQing Cai et al.

Significant advancements has recently been achieved in the field of multi-modal large language models (MLLMs), demonstrating their remarkable capabilities in understanding and reasoning across diverse tasks. However, these models are often trained for specific tasks and rely on task-specific input-output formats, limiting their applicability to a broader range of tasks. This raises a fundamental question: Can we develop a unified approach to represent and handle different multi-modal tasks to maximize the generalizability of MLLMs? In this paper, we propose UnifiedMLLM, a comprehensive model designed to represent various tasks using a unified representation. Our model exhibits strong capabilities in comprehending the implicit intent of user instructions and preforming reasoning. In addition to generating textual responses, our model also outputs task tokens and grounding tokens, serving as indicators of task types and task granularity. These outputs are subsequently routed through the task router and directed to specific expert models for task completion. To train our model, we construct a task-specific dataset and an 100k multi-task dataset encompassing complex scenarios. Employing a three-stage training strategy, we equip our model with robust reasoning and task processing capabilities while preserving its generalization capacity and knowledge reservoir. Extensive experiments showcase the impressive performance of our unified representation approach across various tasks, surpassing existing methodologies. Furthermore, our approach exhibits exceptional scalability and generality. Our code, model, and dataset will be available at \url{https://github.com/lzw-lzw/UnifiedMLLM}.

Jeffry Wicaksana, Zengqiang Yan, Dong Zhang et al.

The purpose of federated learning is to enable multiple clients to jointly train a machine learning model without sharing data. However, the existing methods for training an image segmentation model have been based on an unrealistic assumption that the training set for each local client is annotated in a similar fashion and thus follows the same image supervision level. To relax this assumption, in this work, we propose a label-agnostic unified federated learning framework, named FedMix, for medical image segmentation based on mixed image labels. In FedMix, each client updates the federated model by integrating and effectively making use of all available labeled data ranging from strong pixel-level labels, weak bounding box labels, to weakest image-level class labels. Based on these local models, we further propose an adaptive weight assignment procedure across local clients, where each client learns an aggregation weight during the global model update. Compared to the existing methods, FedMix not only breaks through the constraint of a single level of image supervision, but also can dynamically adjust the aggregation weight of each local client, achieving rich yet discriminative feature representations. To evaluate its effectiveness, experiments have been carried out on two challenging medical image segmentation tasks, i.e., breast tumor segmentation and skin lesion segmentation. The results validate that our proposed FedMix outperforms the state-of-the-art method by a large margin.

Zhaowei Li, Qi Xu, Dong Zhang et al.

Multi-modal large language models have demonstrated impressive performance across various tasks in different modalities. However, existing multi-modal models primarily emphasize capturing global information within each modality while neglecting the importance of perceiving local information across modalities. Consequently, these models lack the ability to effectively understand the fine-grained details of input data, limiting their performance in tasks that require a more nuanced understanding. To address this limitation, there is a compelling need to develop models that enable fine-grained understanding across multiple modalities, thereby enhancing their applicability to a wide range of tasks. In this paper, we propose GroundingGPT, a language enhanced multi-modal grounding model. Beyond capturing global information like other multi-modal models, our proposed model excels at tasks demanding a detailed understanding of local information within the input. It demonstrates precise identification and localization of specific regions in images or moments in videos. To achieve this objective, we design a diversified dataset construction pipeline, resulting in a multi-modal, multi-granularity dataset for model training. The code, dataset, and demo of our model can be found at https: //github.com/lzw-lzw/GroundingGPT.

Dong Zhang, Zhaowei Li, Shimin Li et al.

Speech language models have significantly advanced in generating realistic speech, with neural codec language models standing out. However, the integration of human feedback to align speech outputs to human preferences is often neglected. This paper addresses this gap by first analyzing the distribution gap in codec language models, highlighting how it leads to discrepancies between the training and inference phases, which negatively affects performance. Then we explore leveraging learning from human feedback to bridge the distribution gap. We introduce SpeechAlign, an iterative self-improvement strategy that aligns speech language models to human preferences. SpeechAlign involves constructing a preference codec dataset contrasting golden codec tokens against synthetic tokens, followed by preference optimization to improve the codec language model. This cycle of improvement is carried out iteratively to steadily convert weak models to strong ones. Through both subjective and objective evaluations, we show that SpeechAlign can bridge the distribution gap and facilitating continuous self-improvement of the speech language model. Moreover, SpeechAlign exhibits robust generalization capabilities and works for smaller models. Code and models will be available at https://github.com/0nutation/SpeechGPT.

Pingcheng Dong, Yonghao Tan, Dong Zhang et al.

Non-linear functions are prevalent in Transformers and their lightweight variants, incurring substantial and frequently underestimated hardware costs. Previous state-of-the-art works optimize these operations by piece-wise linear approximation and store the parameters in look-up tables (LUT), but most of them require unfriendly high-precision arithmetics such as FP/INT 32 and lack consideration of integer-only INT quantization. This paper proposed a genetic LUT-Approximation algorithm namely GQA-LUT that can automatically determine the parameters with quantization awareness. The results demonstrate that GQA-LUT achieves negligible degradation on the challenging semantic segmentation task for both vanilla and linear Transformer models. Besides, proposed GQA-LUT enables the employment of INT8-based LUT-Approximation that achieves an area savings of 81.3~81.7% and a power reduction of 79.3~80.2% compared to the high-precision FP/INT 32 alternatives. Code is available at https:// github.com/PingchengDong/GQA-LUT.

Han Zhu, Wei Kang, Liyong Guo et al.

Generating spoken dialogue is more challenging than monologue text-to-speech (TTS) due to the need for realistic turn-taking and distinct speaker timbres. Existing spoken dialogue generation models, being auto-regressive, suffer from slow and unstable inference. To overcome these limitations, we introduce ZipVoice-Dialog, a non-autoregressive zero-shot spoken dialogue generation model built upon flow matching. Key designs include: 1) speaker-turn embeddings for precise speaker turn-taking; 2) a curriculum learning strategy for stable speech-text alignment; 3) specialized strategies to enable stereo dialogue generation. Additionally, recognizing the lack of open-source large-scale spoken dialogue datasets, we curated OpenDialog, a 6.8k-hour spoken dialogue dataset from in-the-wild speech data. Furthermore, we established a benchmark to comprehensively evaluate various models. Experimental results demonstrate that ZipVoice-Dialog achieves superior performance in intelligibility, speaker turn-taking accuracy, speaker similarity, and inference speed. Our codes, model checkpoints, demo samples, and the OpenDialog dataset are all publicly available at https://github.com/k2-fsa/ZipVoice.

Yonghao Tan, Pingcheng Dong, Yongkun Wu et al.

DNN accelerators, significantly advanced by model compression and specialized dataflow techniques, have marked considerable progress. However, the frequent access of high-precision partial sums (PSUMs) leads to excessive memory demands in architectures utilizing input/weight stationary dataflows. Traditional compression strategies have typically overlooked PSUM quantization, which may account for 69% of power consumption. This study introduces a novel Additive Partial Sum Quantization (APSQ) method, seamlessly integrating PSUM accumulation into the quantization framework. A grouping strategy that combines APSQ with PSUM quantization enhanced by a reconfigurable architecture is further proposed. The APSQ performs nearly lossless on NLP and CV tasks across BERT, Segformer, and EfficientViT models while compressing PSUMs to INT8. This leads to a notable reduction in energy costs by 28-87%. Extended experiments on LLaMA2-7B demonstrate the potential of APSQ for large language models. Code is available at https://github.com/Yonghao-Tan/APSQ.

Dong Zhang, Xin Zhang, Jun Zhan et al.

Benefiting from effective speech modeling, current Speech Large Language Models (SLLMs) have demonstrated exceptional capabilities in in-context speech generation and efficient generalization to unseen speakers. However, the prevailing information modeling process is encumbered by certain redundancies, leading to inefficiencies in speech generation. We propose Chain-of-Information Generation (CoIG), a method for decoupling semantic and perceptual information in large-scale speech generation. Building on this, we develop SpeechGPT-Gen, an 8-billion-parameter SLLM efficient in semantic and perceptual information modeling. It comprises an autoregressive model based on LLM for semantic information modeling and a non-autoregressive model employing flow matching for perceptual information modeling. Additionally, we introduce the novel approach of infusing semantic information into the prior distribution to enhance the efficiency of flow matching. Extensive experimental results demonstrate that SpeechGPT-Gen markedly excels in zero-shot text-to-speech, zero-shot voice conversion, and speech-to-speech dialogue, underscoring CoIG's remarkable proficiency in capturing and modeling speech's semantic and perceptual dimensions. Code and models are available at https://github.com/0nutation/SpeechGPT.

Fengyun Wang, Qianru Sun, Dong Zhang et al.

Semantic scene completion (SSC) aims to predict complete 3D voxel occupancy and semantics from a single-view RGB-D image, and recent SSC methods commonly adopt multi-modal inputs. However, our investigation reveals two limitations: ineffective feature learning from single modalities and overfitting to limited datasets. To address these issues, this paper proposes a novel SSC framework - Adversarial Modality Modulation Network (AMMNet) - with a fresh perspective of optimizing gradient updates. The proposed AMMNet introduces two core modules: a cross-modal modulation enabling the interdependence of gradient flows between modalities, and a customized adversarial training scheme leveraging dynamic gradient competition. Specifically, the cross-modal modulation adaptively re-calibrates the features to better excite representation potentials from each single modality. The adversarial training employs a minimax game of evolving gradients, with customized guidance to strengthen the generator's perception of visual fidelity from both geometric completeness and semantic correctness. Extensive experimental results demonstrate that AMMNet outperforms state-of-the-art SSC methods by a large margin, providing a promising direction for improving the effectiveness and generalization of SSC methods.

Dong Zhang

We develop a thermodynamic theory for machine learning (ML) systems. Similar to physical thermodynamic systems which are characterized by energy and entropy, ML systems possess these characteristics as well. This comparison inspire us to integrate the concept of temperature into ML systems grounded in the fundamental principles of thermodynamics, and establish a basic thermodynamic framework for machine learning systems with non-Boltzmann distributions. We introduce the concept of states within a ML system, identify two typical types of state, and interpret model training and refresh as a process of state phase transition. We consider that the initial potential energy of a ML system is described by the model's loss functions, and the energy adheres to the principle of minimum potential energy. For a variety of energy forms and parameter initialization methods, we derive the temperature of systems during the phase transition both analytically and asymptotically, highlighting temperature as a vital indicator of system data distribution and ML training complexity. Moreover, we perceive deep neural networks as complex heat engines with both global temperature and local temperatures in each layer. The concept of work efficiency is introduced within neural networks, which mainly depends on the neural activation functions. We then classify neural networks based on their work efficiency, and describe neural networks as two types of heat engines.

Zhang Dong, Songhang deng, Mingbang Wang et al.

Large-scale coreference resolution presents a significant challenge in natural language processing, necessitating a balance between efficiency and accuracy. In response to this challenge, we introduce an End-to-End Neural Coreference Resolution system tailored for large-scale applications. Our system efficiently identifies and resolves coreference links in text, ensuring minimal computational overhead without compromising on performance. By utilizing advanced neural network architectures, we incorporate various contextual embeddings and attention mechanisms, which enhance the quality of predictions for coreference pairs. Furthermore, we apply optimization strategies to accelerate processing speeds, making the system suitable for real-world deployment. Extensive evaluations conducted on benchmark datasets demonstrate that our model achieves improved accuracy compared to existing approaches, while effectively maintaining rapid inference times. Rigorous testing confirms the ability of our system to deliver precise coreference resolutions efficiently, thereby establishing a benchmark for future advancements in this field.

Dong Zhang, Shimin Li, Xin Zhang et al.

Multi-modal large language models are regarded as a crucial step towards Artificial General Intelligence (AGI) and have garnered significant interest with the emergence of ChatGPT. However, current speech-language models typically adopt the cascade paradigm, preventing inter-modal knowledge transfer. In this paper, we propose SpeechGPT, a large language model with intrinsic cross-modal conversational abilities, capable of perceiving and generating multi-model content. With discrete speech representations, we first construct SpeechInstruct, a large-scale cross-modal speech instruction dataset. Additionally, we employ a three-stage training strategy that includes modality-adaptation pre-training, cross-modal instruction fine-tuning, and chain-of-modality instruction fine-tuning. The experimental results demonstrate that SpeechGPT has an impressive capacity to follow multi-modal human instructions and highlight the potential of handling multiple modalities with one model. Demos are shown in https://0nutation.github.io/SpeechGPT.github.io/.

Huajian Huang, Wen-Yan Lin, Siying Liu et al.

SLAM (Simultaneous Localization And Mapping) seeks to provide a moving agent with real-time self-localization. To achieve real-time speed, SLAM incrementally propagates position estimates. This makes SLAM fast but also makes it vulnerable to local pose estimation failures. As local pose estimation is ill-conditioned, local pose estimation failures happen regularly, making the overall SLAM system brittle. This paper attempts to correct this problem. We note that while local pose estimation is ill-conditioned, pose estimation over longer sequences is well-conditioned. Thus, local pose estimation errors eventually manifest themselves as mapping inconsistencies. When this occurs, we save the current map and activate two new SLAM threads. One processes incoming frames to create a new map and the other, recovery thread, backtracks to link new and old maps together. This creates a Dual-SLAM framework that maintains real-time performance while being robust to local pose estimation failures. Evaluation on benchmark datasets shows Dual-SLAM can reduce failures by a dramatic $88\%$.