Qiquan Zhang

Chen Zhang, Luis Fernando D'Haro, Qiquan Zhang et al.

Recent model-based reference-free metrics for open-domain dialogue evaluation exhibit promising correlations with human judgment. However, they either perform turn-level evaluation or look at a single dialogue quality dimension. One would expect a good evaluation metric to assess multiple quality dimensions at the dialogue level. To this end, we are motivated to propose a multi-dimensional dialogue-level metric, which consists of three sub-metrics with each targeting a specific dimension. The sub-metrics are trained with novel self-supervised objectives and exhibit strong correlations with human judgment for their respective dimensions. Moreover, we explore two approaches to combine the sub-metrics: metric ensemble and multitask learning. Both approaches yield a holistic metric that significantly outperforms individual sub-metrics. Compared to the existing state-of-the-art metric, the combined metrics achieve around 16% relative improvement on average across three high-quality dialogue-level evaluation benchmarks.

Qian Chen, Luyao Cheng, Chong Deng et al.

Recent advancements in joint speech-text models show great potential for seamless voice interactions. However, existing models face critical challenges: temporal resolution mismatch between speech tokens (25Hz) and text tokens (~3Hz) dilutes semantic information, incurs high computational costs, and causes catastrophic forgetting of text LLM knowledge. We introduce Fun-Audio-Chat, a Large Audio Language Model addressing these limitations via two innovations from our previous work DrVoice. First, Dual-Resolution Speech Representations (DRSR): the Shared LLM processes audio at efficient 5Hz (via token grouping), while the Speech Refined Head generates high-quality tokens at 25Hz, balancing efficiency (~50% GPU reduction) and quality. Second, Core-Cocktail Training, a two-stage fine-tuning with intermediate merging that mitigates catastrophic forgetting. We then apply Multi-Task DPO Training to enhance robustness, audio understanding, instruction-following and voice empathy. This multi-stage post-training enables Fun-Audio-Chat to retain text LLM knowledge while gaining powerful audio understanding, reasoning, and generation. Unlike recent LALMs requiring large-scale audio-text pre-training, Fun-Audio-Chat leverages pre-trained models and extensive post-training. Fun-Audio-Chat 8B and MoE 30B-A3B achieve competitive performance on Speech-to-Text and Speech-to-Speech tasks, ranking top among similar-scale models on Spoken QA benchmarks. They also achieve competitive to superior performance on Audio Understanding, Speech Function Calling, Instruction-Following and Voice Empathy. We develop Fun-Audio-Chat-Duplex, a full-duplex variant with strong performance on Spoken QA and full-duplex interactions. We open-source Fun-Audio-Chat-8B with training and inference code, and provide an interactive demo.

Chen Zhang, Luis Fernando D'Haro, Qiquan Zhang et al.

Chatbots are expected to be knowledgeable across multiple domains, e.g. for daily chit-chat, exchange of information, and grounding in emotional situations. To effectively measure the quality of such conversational agents, a model-based automatic dialogue evaluation metric (ADEM) is expected to perform well across multiple domains. Despite significant progress, an ADEM that works well in one domain does not necessarily generalize to another. This calls for a dedicated network architecture for domain generalization. To tackle the multi-domain dialogue evaluation task, we propose a Panel of Experts (PoE), a multitask network that consists of a shared transformer encoder and a collection of lightweight adapters. The shared encoder captures the general knowledge of dialogues across domains, while each adapter specializes in one specific domain and serves as a domain expert. To validate the idea, we construct a high-quality multi-domain dialogue dataset leveraging data augmentation and pseudo-labeling. The PoE network is comprehensively assessed on 16 dialogue evaluation datasets spanning a wide range of dialogue domains. It achieves state-of-the-art performance in terms of mean Spearman correlation over all the evaluation datasets. It exhibits better zero-shot generalization than existing state-of-the-art ADEMs and the ability to easily adapt to new domains with few-shot transfer learning.

Jinyang Wu, Zihan Pan, Qiquan Zhang et al.

Neural audio codecs discretize speech via residual vector quantization (RVQ), forming a coarse-to-fine hierarchy across quantizers. While codec models have been explored for representation learning, their discrete structure remains underutilized in speech deepfake detection. In particular, different quantization levels capture complementary acoustic cues, where early quantizers encode coarse structure and later quantizers refine residual details that reveal synthesis artifacts. Existing systems either rely on continuous encoder features or ignore this quantizer-level hierarchy. We propose a hierarchy-aware representation learning framework that models quantizer-level contributions through learnable global weighting, enabling structured codec representations aligned with forensic cues. Keeping the speech encoder backbone frozen and updating only 4.4% additional parameters, our method achieves relative EER reductions of 46.2% on ASVspoof 2019 and 13.9% on ASVspoof5 over strong baselines.

Xiangyu Zhang, Hexin Liu, Kaishuai Xu et al.

Depression is a critical concern in global mental health, prompting extensive research into AI-based detection methods. Among various AI technologies, Large Language Models (LLMs) stand out for their versatility in mental healthcare applications. However, their primary limitation arises from their exclusive dependence on textual input, which constrains their overall capabilities. Furthermore, the utilization of LLMs in identifying and analyzing depressive states is still relatively untapped. In this paper, we present an innovative approach to integrating acoustic speech information into the LLMs framework for multimodal depression detection. We investigate an efficient method for depression detection by integrating speech signals into LLMs utilizing Acoustic Landmarks. By incorporating acoustic landmarks, which are specific to the pronunciation of spoken words, our method adds critical dimensions to text transcripts. This integration also provides insights into the unique speech patterns of individuals, revealing the potential mental states of individuals. Evaluations of the proposed approach on the DAIC-WOZ dataset reveal state-of-the-art results when compared with existing Audio-Text baselines. In addition, this approach is not only valuable for the detection of depression but also represents a new perspective in enhancing the ability of LLMs to comprehend and process speech signals.

Zexia Fan, Yu Chen, Qiquan Zhang et al.

Sound source localization (SSL) demonstrates remarkable results in controlled settings but struggles in real-world deployment due to dual imbalance challenges: intra-task imbalance arising from long-tailed direction-of-arrival (DoA) distributions, and inter-task imbalance induced by cross-task skews and overlaps. These often lead to catastrophic forgetting, significantly degrading the localization accuracy. To mitigate these issues, we propose a unified framework with two key innovations. Specifically, we design a GCC-PHAT-based data augmentation (GDA) method that leverages peak characteristics to alleviate intra-task distribution skews. We also propose an Analytic dynamic imbalance rectifier (ADIR) with task-adaption regularization, which enables analytic updates that adapt to inter-task dynamics. On the SSLR benchmark, our proposal achieves state-of-the-art (SoTA) results of 89.0% accuracy, 5.3° mean absolute error, and 1.6 backward transfer, demonstrating robustness to evolving imbalances without exemplar storage.

Xiangyu Zhang, Daijiao Liu, Hexin Liu et al.

Recently, Denoising Diffusion Probabilistic Models (DDPMs) have attained leading performances across a diverse range of generative tasks. However, in the field of speech synthesis, although DDPMs exhibit impressive performance, their long training duration and substantial inference costs hinder practical deployment. Existing approaches primarily focus on enhancing inference speed, while approaches to accelerate training a key factor in the costs associated with adding or customizing voices often necessitate complex modifications to the model, compromising their universal applicability. To address the aforementioned challenges, we propose an inquiry: is it possible to enhance the training/inference speed and performance of DDPMs by modifying the speech signal itself? In this paper, we double the training and inference speed of Speech DDPMs by simply redirecting the generative target to the wavelet domain. This method not only achieves comparable or superior performance to the original model in speech synthesis tasks but also demonstrates its versatility. By investigating and utilizing different wavelet bases, our approach proves effective not just in speech synthesis, but also in speech enhancement.

Xinyuan Qian, Jiaran Gao, Yaodan Zhang et al.

Speech enhancement plays an essential role in various applications, and the integration of visual information has been demonstrated to bring substantial advantages. However, the majority of current research concentrates on the examination of facial and lip movements, which can be compromised or entirely inaccessible in scenarios where occlusions occur or when the camera view is distant. Whereas contextual visual cues from the surrounding environment have been overlooked: for example, when we see a dog bark, our brain has the innate ability to discern and filter out the barking noise. To this end, in this paper, we introduce a novel task, i.e. SAV-SE. To our best knowledge, this is the first proposal to use rich contextual information from synchronized video as auxiliary cues to indicate the type of noise, which eventually improves the speech enhancement performance. Specifically, we propose the VC-S$^2$E method, which incorporates the Conformer and Mamba modules for their complementary strengths. Extensive experiments are conducted on public MUSIC, AVSpeech and AudioSet datasets, where the results demonstrate the superiority of VC-S$^2$E over other competitive methods. We will make the source code publicly available. Project demo page: https://AVSEPage.github.io/

Jinyang Wu, Bin Zhu, Xiandong Zou et al.

As Speech Large Language Models (Speech LLMs) become increasingly integrated into voice-based applications, ensuring their robustness against manipulative or adversarial input becomes critical. Although prior work has studied adversarial attacks in text-based LLMs and vision-language models, the unique cognitive and perceptual challenges of speech-based interaction remain underexplored. In contrast, speech presents inherent ambiguity, continuity, and perceptual diversity, which make adversarial attacks more difficult to detect. In this paper, we introduce gaslighting attacks, strategically crafted prompts designed to mislead, override, or distort model reasoning as a means to evaluate the vulnerability of Speech LLMs. Specifically, we construct five manipulation strategies: Anger, Cognitive Disruption, Sarcasm, Implicit, and Professional Negation, designed to test model robustness across varied tasks. It is worth noting that our framework captures both performance degradation and behavioral responses, including unsolicited apologies and refusals, to diagnose different dimensions of susceptibility. Moreover, acoustic perturbation experiments are conducted to assess multi-modal robustness. To quantify model vulnerability, comprehensive evaluation across 5 Speech and multi-modal LLMs on over 10,000 test samples from 5 diverse datasets reveals an average accuracy drop of 24.3% under the five gaslighting attacks, indicating significant behavioral vulnerability. These findings highlight the need for more resilient and trustworthy speech-based AI systems.

Haoyang Zhang, Hexin Liu, Xiangyu Zhang et al.

The speech tokenizer plays a crucial role in recent speech tasks, generally serving as a bridge between speech signals and language models. While low-frame-rate codecs are widely employed as speech tokenizers, the impact of frame rates on speech tokens remains underexplored. In this study, we investigate how varying frame rates affect speech tokenization by examining Mandarin and English, two typologically distinct languages. We encode speech at different frame rates and evaluate the resulting semantic tokens in the speech recognition task. Our findings reveal that frame rate variations influence speech tokenization differently for each language, highlighting the interplay between frame rates, phonetic density, and language-specific acoustic features. The results provide insights into optimizing frame rate selection for speech tokenizers, with implications for automatic speech recognition, text-to-speech, and other speech-related applications.

Qiquan Zhang, Qi Song, Zhaoheng Ni et al.

Most studies on speech enhancement generally don't consider the energy distribution of speech in time-frequency (T-F) representation, which is important for accurate prediction of mask or spectra. In this paper, we present a simple yet effective T-F attention (TFA) module, where a 2-D attention map is produced to provide differentiated weights to the spectral components of T-F representation. To validate the effectiveness of our proposed TFA module, we use the residual temporal convolution network (ResTCN) as the backbone network and conduct extensive experiments on two commonly used training targets. Our experiments demonstrate that applying our TFA module significantly improves the performance in terms of five objective evaluation metrics with negligible parameter overhead. The evaluation results show that the proposed ResTCN with the TFA module (ResTCN+TFA) consistently outperforms other baselines by a large margin.

Peng Gao, Qiquan Zhang, Fei Wang et al.

Although numerous recent tracking approaches have made tremendous advances in the last decade, achieving high-performance visual tracking remains a challenge. In this paper, we propose an end-to-end network model to learn reinforced attentional representation for accurate target object discrimination and localization. We utilize a novel hierarchical attentional module with long short-term memory and multi-layer perceptrons to leverage both inter- and intra-frame attention to effectively facilitate visual pattern emphasis. Moreover, we incorporate a contextual attentional correlation filter into the backbone network to make our model trainable in an end-to-end fashion. Our proposed approach not only takes full advantage of informative geometries and semantics but also updates correlation filters online without fine-tuning the backbone network to enable the adaptation of variations in the target object's appearance. Extensive experiments conducted on several popular benchmark datasets demonstrate that our proposed approach is effective and computationally efficient.