Yifu Chen, Shengpeng Ji, Zhengqing Liu, Qian Chen, Wen Wang, Ziqing Wang, Yangzhuo Li, Tianle Liang, Zhou Zhao. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Tianle Liang, Yifu Chen, Shengpeng Ji, Yijun Chen, Zhiyang Jia, Jingyu Lu, Fan Zhuo, Xueyi Pu, Yangzhuo Li, Zhou Zhao. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Speech watermarking techniques can proactively mitigate the potential harmful consequences of instant voice cloning techniques. These techniques involve the insertion of signals into speech that are imperceptible to humans but can be detected by algorithms. Previous approaches typically embed watermark messages into continuous space. However, intuitively, embedding watermark information into robust discrete latent space can significantly improve the robustness of watermarking systems. In this paper, we propose DiscreteWM, a novel speech watermarking framework that injects watermarks into the discrete intermediate representations of speech. Specifically, we map speech into discrete latent space with a vector-quantized autoencoder and inject watermarks by changing the modular arithmetic relation of discrete IDs. To ensure the imperceptibility of watermarks, we also propose a manipulator model to select the candidate tokens for watermark embedding. Experimental results demonstrate that our framework achieves state-of-the-art performance in robustness and imperceptibility, simultaneously. Moreover, our flexible frame-wise approach can serve as an efficient solution for both voice cloning detection and information hiding. Additionally, DiscreteWM can encode 1 to 150 bits of watermark information within a 1-second speech clip, indicating its encoding capacity.

Zehan Wang, Ke Lei, Chen Zhu, Jiawei Huang, Sashuai Zhou, Luping Liu, Xize Cheng, Shengpeng Ji, Zhenhui Ye, Tao Jin, Zhou Zhao. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2025.

Minghui Fang, Shengpeng Ji, Jialong Zuo, Hai Huang, Yan Xia, Jieming Zhu, Xize Cheng, Xiaoda Yang, Wenrui Liu, Gang Wang, Zhenhua Dong, Zhou Zhao. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2025.

Yifu Chen, Shengpeng Ji, Haoxiao Wang, Ziqing Wang, Siyu Chen, Jinzheng He, Jin Xu, Zhou Zhao. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2025.

In recent years, large language models have achieved significant success in generative tasks (e.g., speech cloning and audio generation) related to speech, audio, music, and other signal domains. A crucial element of these models is the discrete acoustic codecs, which serve as an intermediate representation replacing the mel-spectrogram. However, there exist several gaps between discrete codecs and downstream speech language models. Specifically, 1) Due to the reconstruction paradigm of the Codec model and the structure of residual vector quantization, the initial channel of the codebooks contains excessive information, making it challenging to directly generate acoustic tokens from weakly supervised signals such as text in downstream tasks. 2) Achieving good reconstruction performance requires the utilization of numerous codebooks, which increases the burden on downstream speech language models. Consequently, leveraging the characteristics of speech language models, we propose Language-Codec. In the Language-Codec, we introduce a Masked Channel Residual Vector Quantization (MCRVQ) mechanism along with improved fourier transform structures, refined discriminator design to address the aforementioned gaps. We compare our method with competing audio compression algorithms and observe significant outperformance across extensive evaluations. Furthermore, we also validate the efficiency of the Language-Codec on downstream speech language models. The source code and pretrained models will be open-sourced after the paper is accepted. Codes are available at https://github.com/jishengpeng/Languagecodec.

In this paper, we present ControlSpeech, a text-to-speech (TTS) system capable of fully cloning the speaker’s voice and enabling arbitrary control and adjustment of speaking style. Prior zero-shot TTS models only mimic the speaker’s voice without further control and adjustment capabilities while prior controllable TTS models cannot perform speaker-specific voice generation. Therefore, ControlSpeech focuses on a more challenging task—a TTS system with controllable timbre, content, and style at the same time. ControlSpeech takes speech prompts, content prompts, and style prompts as inputs and utilizes bidirectional attention and mask-based parallel decoding to capture codec representations corresponding to timbre, content, and style in a discrete decoupling codec space. Moreover, we analyze the many-to-many issue in textual style control and propose the Style Mixture Semantic Density (SMSD) module, which is based on Gaussian mixture density networks, to resolve this problem. To facilitate empirical validations, we make available a new style controllable dataset called VccmDataset. Our experimental results demonstrate that ControlSpeech exhibits comparable or state-of-the-art (SOTA) performance in terms of controllability, timbre similarity, audio quality, robustness, and generalizability. Codes are available at https://github.com/jishengpeng/ControlSpeech.

Yidi Jiang, Qian Chen, Shengpeng Ji, Yu Xi, Wen Wang, Chong Zhang, Xianghu Yue, ShiLiang Zhang, Haizhou Li. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2025.

Recent advancements in speech large language models (SpeechLLMs) have attracted considerable attention. Nonetheless, current methods exhibit suboptimal performance in adhering to speech instructions. Notably, the intelligence of models significantly diminishes when processing speech-form input as compared to direct text-form input. Prior work has attempted to mitigate this semantic inconsistency between speech and text representations through techniques such as representation and behavior alignment, which involve the meticulous design of data pairs during the post-training phase. In this paper, we introduce a simple and scalable training method called InSerter, which stands for Interleaved Speech-Text Representation Pre-training. InSerter is designed to pre-train large-scale unsupervised speech-text sequences, where the speech is synthesized from randomly selected segments of an extensive text corpus using text-to-speech conversion. Consequently, the model acquires the ability to generate textual continuations corresponding to the provided speech segments, obviating the need for intensive data design endeavors. To systematically evaluate speech instruction-following capabilities, we introduce SpeechInstructBench, the first comprehensive benchmark specifically designed for speech-oriented instruction-following tasks. Our proposed model InSerter achieves SOTA performance in SpeechInstructBench and demonstrates superior or competitive results across diverse speech processing tasks.

Jialong Zuo, Shengpeng Ji, Minghui Fang, Mingze Li, Ziyue Jiang, Xize Cheng, Xiaoda Yang, Chen Feiyang, Xinyu Duan, Zhou Zhao. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2025.

In recent years, speech generation fields have achieved significant advancements, primarily due to improvements in large TTS (text-to-speech) systems and scalable TTS datasets. However, there is still a lack of large-scale multilingual TTS datasets, which limits the development of cross-language and multilingual TTS systems. Hence, we refine Voxpopuli dataset and propose VoxpopuliTTS dataset. This dataset comprises 30,000 hours of high-quality speech data, across 3 languages with multiple speakers and styles, suitable for various speech tasks such as TTS and ASR. To enhance the quality of speech data from Voxpopuli, we improve the existing processing pipeline by: 1) filtering out low-quality speech-text pairs based on ASR confidence scores, and 2) concatenating short transcripts by checking semantic information completeness to generate the long transcript. Experimental results demonstrate the effectiveness of the VoxpopuliTTS dataset and the proposed processing pipeline.

Contrastive Language-Image Pre-training (CLIP) learns robust visual models through language supervision, making it a crucial visual encoding technique for various applications. However, CLIP struggles with comprehending spatial concepts in images, potentially restricting the spatial intelligence of CLIP-based AI systems. In this work, we propose SpatialCLIP, an enhanced version of CLIP with better spatial understanding capabilities. To capture the intricate 3D spatial relationships in images, we improve both "visual model" and "language supervision" of CLIP. Specifically, we design 3D-inspired ViT to replace the standard ViT in CLIP. By lifting 2D image tokens into 3D space and incorporating design insights from point cloud networks, our visual model gains greater potential for spatial perception. Meanwhile, captions with accurate and detailed spatial information are very rare. To explore better language supervision for spatial understanding, we re-caption images and perturb their spatial phrases as negative descriptions, which compels the visual model to seek spatial cues to distinguish these hard negative captions. With the enhanced visual model, we introduce SpatialLLaVA, following the same LLaVA-1.5 training protocol, to investigate the importance of visual representations for MLLM’s spatial intelligence. Furthermore, we create SpatialBench, a benchmark specifically designed to evaluate CLIP and MLLM in spatial reasoning. Spatial-CLIP and SpatialLLaVA achieve substantial performance improvements, demonstrating stronger capabilities in spatial perception and reasoning, while maintaining comparable results on general-purpose benchmarks.

This paper introduces PFlow-VC, a conditional flow matching voice conversion model that leverages fine-grained discrete pitch tokens and target speaker prompt information for expressive voice conversion (VC). Previous VC works primarily focus on speaker conversion, with further exploration needed in enhancing expressiveness (such as prosody and emotion) for timbre conversion. Unlike previous methods, we adopt a simple and efficient approach to enhance the style expressiveness of voice conversion models. Specifically, we pretrain a self-supervised pitch VQVAE model to discretize speaker-irrelevant pitch information and leverage a masked pitch-conditioned flow matching model for Mel-spectrogram synthesis, which provides in-context pitch modeling capabilities for the speaker conversion model, effectively improving the voice style transfer capacity. Additionally, we improve timbre similarity by combining global timbre embeddings with time-varying timbre tokens. Experiments on unseen LibriTTS test-clean and emotional speech dataset ESD show the superiority of the PFlow-VC model in both timbre conversion and style transfer. Audio samples are available on the demo page https://speechai-demo.github.io/PFlow-VC/.

This paper extends Cross Modal Generalization (CMG) to open-set environments by proposing the more challenging Open-set Cross Modal Generalization (OSCMG) task. This task evaluates multimodal unified representations in open-set conditions, addressing the limitations of prior closed-set cross-modal evaluations. OSCMG requires not only cross-modal knowledge transfer but also robust generalization to unseen classes within new modalities, a scenario frequently encountered in real-world applications. Existing multimodal unified representation work lacks consideration for open-set environments. To tackle this, we propose MICU, comprising two key components: Fine-Coarse Masked multimodal InfoNCE (FCMI) and Cross modal Unified Jigsaw Puzzles (CUJP). FCMI enhances multimodal alignment by applying contrastive learning at both holistic semantic and temporal levels, incorporating masking to enhance generalization. CUJP enhances feature diversity and model uncertainty by integrating modality-agnostic feature selection with self-supervised learning, thereby strengthening the model's ability to handle unknown categories in open-set tasks. Extensive experiments on CMG and the newly proposed OSCMG validate the effectiveness of our approach. The code is available at https://github.com/haihuangcode/CMG.

Recently, human-computer interaction with various modalities has shown promising applications, like GPT-4o and Gemini. Meanwhile, multimodal representation models have emerged as the foundation for these versatile multimodal understanding and generation pipeline. Models like CLIP, CLAP and ImageBind can map their specialized modalities into respective joint spaces. To construct a high-quality omni representation space that can be shared and expert in any modality, we propose to merge these advanced models into a unified space in scale. With this insight, we present \textbf{OmniBind}, advanced multimodal joint representation models via fusing knowledge of 14 pre-trained spaces, which support 3D, audio, image, video and language inputs. To alleviate the interference between different knowledge sources in integrated space, we dynamically assign weights to different spaces by learning routers with two objectives: cross-modal overall alignment and language representation decoupling. Notably, since binding and routing spaces only require lightweight networks, OmniBind is extremely training-efficient. Extensive experiments demonstrate the versatility and superiority of OmniBind as an omni representation model, highlighting its great potential for diverse applications, such as any-query and composable multimodal understanding.

With the rapid advancement of large models, voice assistants are gradually acquiring the ability to engage in open-ended daily conversations with humans. However, current spoken dialogue systems often overlook multi-modal information in audio beyond text, such as speech rate, volume, emphasis, and background sounds. Relying solely on Automatic Speech Recognition (ASR) can lead to the loss of valuable auditory cues, thereby weakening the system’s ability to generate contextually appropriate responses. To address this limitation, we propose \textbf{VoxDialogue}, a comprehensive benchmark for evaluating the ability of spoken dialogue systems to understand multi-modal information beyond text. Specifically, we have identified 12 attributes highly correlated with acoustic information beyond words and have meticulously designed corresponding spoken dialogue test sets for each attribute, encompassing a total of 4.5K multi-turn spoken dialogue samples. Finally, we evaluated several existing spoken dialogue models, analyzing their performance on the 12 attribute subsets of VoxDialogue. Experiments have shown that in spoken dialogue scenarios, many acoustic cues cannot be conveyed through textual information and must be directly interpreted from the audio input. In contrast, while direct spoken dialogue systems excel at processing acoustic signals, they still face limitations in handling complex dialogue tasks due to their restricted context understanding capabilities. All data and code will be open source at \url{https://voxdialogue.github.io/}.

Query-based sound separation (QSS) effectively isolate sound signals that match the content of a given query, enhancing the understanding of audio data. However, most existing QSS methods rely on a single modality for separation, lacking the ability to fully leverage homologous but heterogeneous information across multiple modalities for the same sound signal. To address this limitation, we introduce Omni-modal Sound Separation (**OmniSep**), a novel framework capable of isolating clean soundtracks based on omni-modal queries, encompassing both single-modal and multi-modal composed queries. Specifically, we introduce the **Query-Mixup** strategy, which blends query features from different modalities during training. This enables OmniSep to optimize multiple modalities concurrently, effectively bringing all modalities under a unified framework for sound separation. We further enhance this flexibility by allowing queries to influence sound separation positively or negatively, facilitating the retention or removal of specific sounds as desired. Finally, OmniSep employs a retrieval-augmented approach known as **Query-Aug**, which enables open-vocabulary sound separation. Experimental evaluations on MUSIC, VGGSOUND-CLEAN+, and MUSIC-CLEAN+ datasets demonstrate effectiveness of OmniSep, achieving state-of-the-art performance in text-, image-, and audio-queried sound separation tasks. For samples and further information, please visit the demo page at \url{https://omnisep.github.io/}.

Language models have been effectively applied to modeling natural signals, such as images, video, speech, and audio. A crucial component of these models is the codec tokenizer, which compresses high-dimensional natural signals into lower-dimensional discrete tokens. In this paper, we introduce WavTokenizer, which offers several advantages over previous SOTA acoustic codec models in the audio domain: 1) extreme compression. By compressing the layers of quantizers and the temporal dimension of the discrete codec, one-second audio of 24kHz sampling rate requires only a single quantizer with 40 or 75 tokens. 2) improved subjective quality. Despite the reduced number of tokens, WavTokenizer achieves state-of-the-art reconstruction quality with outstanding UTMOS scores and inherently contains richer semantic information. Specifically, we achieve these results by designing a broader VQ space, extended contextual windows, and improved attention networks, as well as introducing a powerful multi-scale discriminator and an inverse Fourier transform structure. We conducted extensive reconstruction experiments in the domains of speech, audio, and music. WavTokenizer exhibited strong performance across various objective and subjective metrics compared to state-of-the-art models. We also tested semantic information, VQ utilization, and adaptability to generative models. Comprehensive ablation studies confirm the necessity of each module in WavTokenizer. The code is available at https://github.com/jishengpeng/WavTokenizer.

Bridge models in image restoration construct a diffusion process from degraded to clear images. However, existing methods typically require training a bridge model from scratch for each specific type of degradation, resulting in high computational costs and limited performance. This work aims to efficiently leverage pretrained generative priors within existing image restoration bridges to eliminate this requirement. The main challenge is that standard generative models are typically designed for a diffusion process that starts from pure noise, while restoration tasks begin with a low-quality image, resulting in a mismatch in the state distributions between the two processes. To address this challenge, we propose a transition equation that bridges two diffusion processes with the same endpoint distribution. Based on this, we introduce the IRBridge framework, which enables the direct utilization of generative models within image restoration bridges, offering a more flexible and adaptable approach to image restoration. Extensive experiments on six image restoration tasks demonstrate that IRBridge efficiently integrates generative priors, resulting in improved robustness and generalization performance. Code will be available at GitHub.

Image restoration under adverse weather conditions has been extensively explored, leading to numerous high-performance methods. In particular, recent advances in All-in-One approaches have shown impressive results by training on multi-task image restoration datasets. However, most of these methods rely on dedicated network modules or parameters for each specific degradation type, resulting in a significant parameter overhead. Moreover, the relatedness across different restoration tasks is often overlooked. In light of these issues, we propose a parameter-efficient All-in-One image restoration framework that leverages task-aware enhanced prompts to tackle various adverse weather degradations. Specifically, we adopt a two-stage training paradigm consisting of a pretraining phase and a prompt-tuning phase to mitigate parameter conflicts across tasks. We first employ supervised learning to acquire general restoration knowledge, and then adapt the model to handle specific degradation via trainable soft prompts. Crucially, we enhance these task-specific prompts in a task-aware manner. We apply low-rank decomposition to these prompts to capture both task-general and task-specific characteristics, and impose contrastive constraints to better align them with the actual inter-task relatedness. These enhanced prompts not only improve the parameter efficiency of the restoration model but also enable more accurate task modeling, as evidenced by t-SNE analysis. Experimental results on different restoration tasks demonstrate that the proposed method achieves superior performance with only 2.75M parameters.

Hallucinations present a significant challenge in the development and evaluation of large language models (LLMs), directly affecting their reliability and accuracy. While notable advancements have been made in research on textual and visual hallucinations, there is still a lack of a comprehensive benchmark for evaluating auditory hallucinations in large audio language models (LALMs). To fill this gap, we introduce AHa-Bench, a systematic and comprehensive benchmark for audio hallucinations. Audio data, in particular, uniquely combines the multi-attribute complexity of visual data with the semantic richness of textual data, leading to auditory hallucinations that share characteristics with both visual and textual hallucinations. Based on the source of these hallucinations, AHa-Bench categorizes them into semantic hallucinations, acoustic hallucinations, and semantic-acoustic confusion hallucinations. In addition, we systematically evaluate seven open-source local perception language models (LALMs), demonstrating the challenges these models face in audio understanding, especially when it comes to jointly understanding semantic and acoustic information. Through the development of a comprehensive evaluation framework, AHa-Bench aims to enhance the robustness and stability of LALMs, fostering more reliable and nuanced audio understanding in LALMs. The benchmark dataset is available at \url{https://huggingface.co/datasets/ahabench/AHa-Bench}.

Zero-shot text-to-speech (TTS) has gained significant attention due to its powerful voice cloning capabilities, requiring only a few seconds of unseen speaker voice prompts.However, all previous work has been developed for cloud-based systems.Taking autoregressive models as an example, although these approaches achieve high-fidelity voice cloning, they fall short in terms of inference speed, model size, and robustness.Therefore, we propose MobileSpeech, which is a fast, lightweight, and robust zero-shot text-tospeech system based on mobile devices for the first time.Specifically: 1) leveraging discrete codec, we design a parallel speech mask decoder module called SMD, which incorporates hierarchical information from the speech codec and weight mechanisms across different codec layers during the generation process.Moreover, to bridge the gap between text and speech, we introduce a high-level probabilistic mask that simulates the progression of information flow from less to more during speech generation.2) For speaker prompts, we extract fine-grained prompt duration from the prompt speech and incorporate text, prompt speech by cross attention in SMD.We demonstrate the effectiveness of MobileSpeech on multilingual datasets at different levels, achieving state-ofthe-art results in terms of generating speed and speech quality.MobileSpeech achieves RTF of 0.09 on a single A100 GPU and we have successfully deployed MobileSpeech on mobile devices.Audio samples are available at https://mobilespeech.github.io/ .

Xiaoda Yang, Xize Cheng, Jiaqi Duan, Hongshun Qiu, Minjie Hong, Minghui Fang, Shengpeng Ji, Jialong Zuo, Zhiqing Hong, Zhimeng Zhang, Tao Jin. Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing. 2024.

Recently, there has been a growing interest in the field of controllable Text-to-Speech (TTS). While previous studies have relied on users providing specific style factor values based on acoustic knowledge or selecting reference speeches that meet certain requirements, generating speech solely from natural text prompts has emerged as a new challenge for researchers. This challenge arises due to the scarcity of high-quality speech datasets with natural text style prompt and the absence of advanced text-controllable TTS models. In light of this, 1) we propose TextrolSpeech, which is the first large-scale speech emotion dataset annotated with rich text attributes. The dataset comprises 236,203 pairs of style prompt in natural text descriptions with five style factors and corresponding speech samples. Through iterative experimentation, we introduce a multi-stage prompt programming approach that effectively utilizes the GPT model for generating natural style descriptions in large volumes. 2) Furthermore, to address the need for generating audio with greater style diversity, we propose an efficient architecture called Salle. This architecture treats text controllable TTS as a language model task, utilizing audio codec codes as an intermediate representation to replace the conventional mel-spectrogram. Finally, we successfully demonstrate the ability of the proposed model by showing a comparable performance in the controllable TTS task. Audio samples are available on the demo page https://sall-e.github.io/.

Zero-shot text-to-speech (TTS) aims to synthesize voices with unseen speech prompts, which significantly reduces the data and computation requirements for voice cloning by skipping the fine-tuning process. However, the prompting mechanisms of zero-shot TTS still face challenges in the following aspects: 1) previous works of zero-shot TTS are typically trained with single-sentence prompts, which significantly restricts their performance when the data is relatively sufficient during the inference stage. 2) The prosodic information in prompts is highly coupled with timbre, making it untransferable to each other. This paper introduces Mega-TTS 2, a generic prompting mechanism for zero-shot TTS, to tackle the aforementioned challenges. Specifically, we design a powerful acoustic autoencoder that separately encodes the prosody and timbre information into the compressed latent space while providing high-quality reconstructions. Then, we propose a multi-reference timbre encoder and a prosody latent language model (P-LLM) to extract useful information from multi-sentence prompts. We further leverage the probabilities derived from multiple P-LLM outputs to produce transferable and controllable prosody. Experimental results demonstrate that Mega-TTS 2 could not only synthesize identity-preserving speech with a short prompt of an unseen speaker from arbitrary sources but consistently outperform the fine-tuning method when the volume of data ranges from 10 seconds to 5 minutes. Furthermore, our method enables to transfer various speaking styles to the target timbre in a fine-grained and controlled manner. Audio samples can be found in https://boostprompt.github.io/boostprompt/.

Talking Face Generation (TFG) reconstructs facial motions concerning lips given speech input, which aims to generate highquality, synchronized, and lip-readable videos. Previous efforts have achieved success in generating quality and synchronization, and recently, there has been an increasing focus on the importance of intelligibility. Despite these efforts, there remains a challenge in achieving a balance among quality, synchronization, and intelligibility, often resulting in trade-offs that compromise one aspect in favor of another. In light of this, we propose SyncTalklip, a novel dual-tower framework designed to overcome the challenges of synchronization while improving lip-reading performance. To enhance the performance of SyncTalklip in both synchronization and intelligibility, we design AV-SyncNet, a pre-trained multi-task model, aiming to achieve a dual-focus on synchronization and intelligibility. Moreover, we propose a novel cross-modal contrastive learning bringing audio and video closer to enhance synchronization. Experimental results demonstrate that SyncTalklip achieves state-of-the-art performance in quality, intelligibility, and synchronization. Furthermore, extensive experiments have demonstrated our model's generalizability across domains. The code and demo is available at https://sync-talklip.github.io.

Neural networks (NNs) are widely used in classification-based networking analysis to help traffic transmission and system security. However, there are heterogeneous network devices (e.g., switches and routers) in a network. Manually customizing NNs with specific device requirements (e.g., max allowed running latency) can be time-consuming and labor-intensive. Furthermore, the diverse data characteristics of different networking classification tasks add to the burden of NN customization. This paper introduces Loong, a neural architecture search (NAS) based system that automatically generates NNs for various networking tasks and devices. Loong includes a neural operation embedding module, which embeds candidate neural operations into the layer to be designed. Then, the layer-wise training is used to generate a task-specific NN layer by layer. This layer-wise scheme simultaneously trains and selects candidate neural operations using gradient feedback. Finally, only the important operations are selected to form the layer, maximizing accuracy. By incorporating multiple objectives, including deployment memory and running latency of devices, into the training and selection of NNs, Loong is able to customize NNs for heterogeneous network devices. Experiments show that Loong's NNs outperform 13 manual-designed and NAS-based NNs, with a 4.11% improvement in F1-score. Additionally, Loong's NNs achieve faster (7.92X) speeds on commodity devices.