Xubo Liu

Xubo Liu, Qiuqiang Kong, Yan Zhao et al. · bytedance

Language-queried audio source separation (LASS) is a new paradigm for computational auditory scene analysis (CASA). LASS aims to separate a target sound from an audio mixture given a natural language query, which provides a natural and scalable interface for digital audio applications. Recent works on LASS, despite attaining promising separation performance on specific sources (e.g., musical instruments, limited classes of audio events), are unable to separate audio concepts in the open domain. In this work, we introduce AudioSep, a foundation model for open-domain audio source separation with natural language queries. We train AudioSep on large-scale multimodal datasets and extensively evaluate its capabilities on numerous tasks including audio event separation, musical instrument separation, and speech enhancement. AudioSep demonstrates strong separation performance and impressive zero-shot generalization ability using audio captions or text labels as queries, substantially outperforming previous audio-queried and language-queried sound separation models. For reproducibility of this work, we will release the source code, evaluation benchmark and pre-trained model at: https://github.com/Audio-AGI/AudioSep.

Haohe Liu, Xubo Liu, Xinhao Mei et al.

Few-shot bioacoustic event detection is a task that detects the occurrence time of a novel sound given a few examples. Previous methods employ metric learning to build a latent space with the labeled part of different sound classes, also known as positive events. In this study, we propose a segment-level few-shot learning framework that utilizes both the positive and negative events during model optimization. Training with negative events, which are larger in volume than positive events, can increase the generalization ability of the model. In addition, we use transductive inference on the validation set during training for better adaptation to novel classes. We conduct ablation studies on our proposed method with different setups on input features, training data, and hyper-parameters. Our final system achieves an F-measure of 62.73 on the DCASE 2022 challenge task 5 (DCASE2022-T5) validation set, outperforming the performance of the baseline prototypical network 34.02 by a large margin. Using the proposed method, our submitted system ranks 2nd in DCASE2022-T5. The code of this paper is fully open-sourced at https://github.com/haoheliu/DCASE_2022_Task_5.

Haohe Liu, Zehua Chen, Yi Yuan et al.

Text-to-audio (TTA) system has recently gained attention for its ability to synthesize general audio based on text descriptions. However, previous studies in TTA have limited generation quality with high computational costs. In this study, we propose AudioLDM, a TTA system that is built on a latent space to learn the continuous audio representations from contrastive language-audio pretraining (CLAP) latents. The pretrained CLAP models enable us to train LDMs with audio embedding while providing text embedding as a condition during sampling. By learning the latent representations of audio signals and their compositions without modeling the cross-modal relationship, AudioLDM is advantageous in both generation quality and computational efficiency. Trained on AudioCaps with a single GPU, AudioLDM achieves state-of-the-art TTA performance measured by both objective and subjective metrics (e.g., frechet distance). Moreover, AudioLDM is the first TTA system that enables various text-guided audio manipulations (e.g., style transfer) in a zero-shot fashion. Our implementation and demos are available at https://audioldm.github.io.

Haohe Liu, Yi Yuan, Xubo Liu et al.

Although audio generation shares commonalities across different types of audio, such as speech, music, and sound effects, designing models for each type requires careful consideration of specific objectives and biases that can significantly differ from those of other types. To bring us closer to a unified perspective of audio generation, this paper proposes a framework that utilizes the same learning method for speech, music, and sound effect generation. Our framework introduces a general representation of audio, called "language of audio" (LOA). Any audio can be translated into LOA based on AudioMAE, a self-supervised pre-trained representation learning model. In the generation process, we translate any modalities into LOA by using a GPT-2 model, and we perform self-supervised audio generation learning with a latent diffusion model conditioned on LOA. The proposed framework naturally brings advantages such as in-context learning abilities and reusable self-supervised pretrained AudioMAE and latent diffusion models. Experiments on the major benchmarks of text-to-audio, text-to-music, and text-to-speech demonstrate state-of-the-art or competitive performance against previous approaches. Our code, pretrained model, and demo are available at https://audioldm.github.io/audioldm2.

Haohe Liu, Woosung Choi, Xubo Liu et al.

Speech super-resolution (SR) is a task to increase speech sampling rate by generating high-frequency components. Existing speech SR methods are trained in constrained experimental settings, such as a fixed upsampling ratio. These strong constraints can potentially lead to poor generalization ability in mismatched real-world cases. In this paper, we propose a neural vocoder based speech super-resolution method (NVSR) that can handle a variety of input resolution and upsampling ratios. NVSR consists of a mel-bandwidth extension module, a neural vocoder module, and a post-processing module. Our proposed system achieves state-of-the-art results on the VCTK multi-speaker benchmark. On 44.1 kHz target resolution, NVSR outperforms WSRGlow and Nu-wave by 8% and 37% respectively on log spectral distance and achieves a significantly better perceptual quality. We also demonstrate that prior knowledge in the pre-trained vocoder is crucial for speech SR by performing mel-bandwidth extension with a simple replication-padding method. Samples can be found in https://haoheliu.github.io/nvsr.

Xubo Liu, Haohe Liu, Qiuqiang Kong et al.

In this paper, we introduce the task of language-queried audio source separation (LASS), which aims to separate a target source from an audio mixture based on a natural language query of the target source (e.g., "a man tells a joke followed by people laughing"). A unique challenge in LASS is associated with the complexity of natural language description and its relation with the audio sources. To address this issue, we proposed LASS-Net, an end-to-end neural network that is learned to jointly process acoustic and linguistic information, and separate the target source that is consistent with the language query from an audio mixture. We evaluate the performance of our proposed system with a dataset created from the AudioCaps dataset. Experimental results show that LASS-Net achieves considerable improvements over baseline methods. Furthermore, we observe that LASS-Net achieves promising generalization results when using diverse human-annotated descriptions as queries, indicating its potential use in real-world scenarios. The separated audio samples and source code are available at https://liuxubo717.github.io/LASS-demopage.

Xubo Liu, Egor Lakomkin, Konstantinos Vougioukas et al.

Recently reported state-of-the-art results in visual speech recognition (VSR) often rely on increasingly large amounts of video data, while the publicly available transcribed video datasets are limited in size. In this paper, for the first time, we study the potential of leveraging synthetic visual data for VSR. Our method, termed SynthVSR, substantially improves the performance of VSR systems with synthetic lip movements. The key idea behind SynthVSR is to leverage a speech-driven lip animation model that generates lip movements conditioned on the input speech. The speech-driven lip animation model is trained on an unlabeled audio-visual dataset and could be further optimized towards a pre-trained VSR model when labeled videos are available. As plenty of transcribed acoustic data and face images are available, we are able to generate large-scale synthetic data using the proposed lip animation model for semi-supervised VSR training. We evaluate the performance of our approach on the largest public VSR benchmark - Lip Reading Sentences 3 (LRS3). SynthVSR achieves a WER of 43.3% with only 30 hours of real labeled data, outperforming off-the-shelf approaches using thousands of hours of video. The WER is further reduced to 27.9% when using all 438 hours of labeled data from LRS3, which is on par with the state-of-the-art self-supervised AV-HuBERT method. Furthermore, when combined with large-scale pseudo-labeled audio-visual data SynthVSR yields a new state-of-the-art VSR WER of 16.9% using publicly available data only, surpassing the recent state-of-the-art approaches trained with 29 times more non-public machine-transcribed video data (90,000 hours). Finally, we perform extensive ablation studies to understand the effect of each component in our proposed method.

Xinhao Mei, Xubo Liu, Mark D. Plumbley et al.

Automated audio captioning is a cross-modal translation task that aims to generate natural language descriptions for given audio clips. This task has received increasing attention with the release of freely available datasets in recent years. The problem has been addressed predominantly with deep learning techniques. Numerous approaches have been proposed, such as investigating different neural network architectures, exploiting auxiliary information such as keywords or sentence information to guide caption generation, and employing different training strategies, which have greatly facilitated the development of this field. In this paper, we present a comprehensive review of the published contributions in automated audio captioning, from a variety of existing approaches to evaluation metrics and datasets. We also discuss open challenges and envisage possible future research directions.

Yi Yuan, Haohe Liu, Xubo Liu et al.

Despite recent progress in text-to-audio (TTA) generation, we show that the state-of-the-art models, such as AudioLDM, trained on datasets with an imbalanced class distribution, such as AudioCaps, are biased in their generation performance. Specifically, they excel in generating common audio classes while underperforming in the rare ones, thus degrading the overall generation performance. We refer to this problem as long-tailed text-to-audio generation. To address this issue, we propose a simple retrieval-augmented approach for TTA models. Specifically, given an input text prompt, we first leverage a Contrastive Language Audio Pretraining (CLAP) model to retrieve relevant text-audio pairs. The features of the retrieved audio-text data are then used as additional conditions to guide the learning of TTA models. We enhance AudioLDM with our proposed approach and denote the resulting augmented system as Re-AudioLDM. On the AudioCaps dataset, Re-AudioLDM achieves a state-of-the-art Frechet Audio Distance (FAD) of 1.37, outperforming the existing approaches by a large margin. Furthermore, we show that Re-AudioLDM can generate realistic audio for complex scenes, rare audio classes, and even unseen audio types, indicating its potential in TTA tasks.

Xubo Liu, Zhongkai Zhu, Haohe Liu et al.

Despite breakthroughs in audio generation models, their capabilities are often confined to domain-specific conditions such as speech transcriptions and audio captions. However, real-world audio creation aims to generate harmonious audio containing various elements such as speech, music, and sound effects with controllable conditions, which is challenging to address using existing audio generation systems. We present WavJourney, a novel framework that leverages Large Language Models (LLMs) to connect various audio models for audio creation. WavJourney allows users to create storytelling audio content with diverse audio elements simply from textual descriptions. Specifically, given a text instruction, WavJourney first prompts LLMs to generate an audio script that serves as a structured semantic representation of audio elements. The audio script is then converted into a computer program, where each line of the program calls a task-specific audio generation model or computational operation function. The computer program is then executed to obtain a compositional and interpretable solution for audio creation. Experimental results suggest that WavJourney is capable of synthesizing realistic audio aligned with textually-described semantic, spatial and temporal conditions, achieving state-of-the-art results on text-to-audio generation benchmarks. Additionally, we introduce a new multi-genre story benchmark. Subjective evaluations demonstrate the potential of WavJourney in crafting engaging storytelling audio content from text. We further demonstrate that WavJourney can facilitate human-machine co-creation in multi-round dialogues. To foster future research, the code and synthesized audio are available at: https://audio-agi.github.io/WavJourney_demopage/.

Xubo Liu, Xinhao Mei, Qiushi Huang et al.

Audio captioning aims at using natural language to describe the content of an audio clip. Existing audio captioning systems are generally based on an encoder-decoder architecture, in which acoustic information is extracted by an audio encoder and then a language decoder is used to generate the captions. Training an audio captioning system often encounters the problem of data scarcity. Transferring knowledge from pre-trained audio models such as Pre-trained Audio Neural Networks (PANNs) have recently emerged as a useful method to mitigate this issue. However, there is less attention on exploiting pre-trained language models for the decoder, compared with the encoder. BERT is a pre-trained language model that has been extensively used in Natural Language Processing (NLP) tasks. Nevertheless, the potential of BERT as the language decoder for audio captioning has not been investigated. In this study, we demonstrate the efficacy of the pre-trained BERT model for audio captioning. Specifically, we apply PANNs as the encoder and initialize the decoder from the public pre-trained BERT models. We conduct an empirical study on the use of these BERT models for the decoder in the audio captioning model. Our models achieve competitive results with the existing audio captioning methods on the AudioCaps dataset.

Xubo Liu, Qiushi Huang, Xinhao Mei et al.

Audio captioning aims to generate text descriptions of audio clips. In the real world, many objects produce similar sounds. How to accurately recognize ambiguous sounds is a major challenge for audio captioning. In this work, inspired by inherent human multimodal perception, we propose visually-aware audio captioning, which makes use of visual information to help the description of ambiguous sounding objects. Specifically, we introduce an off-the-shelf visual encoder to extract video features and incorporate the visual features into an audio captioning system. Furthermore, to better exploit complementary audio-visual contexts, we propose an audio-visual attention mechanism that adaptively integrates audio and visual context and removes the redundant information in the latent space. Experimental results on AudioCaps, the largest audio captioning dataset, show that our proposed method achieves state-of-the-art results on machine translation metrics.

Xubo Liu, Haohe Liu, Qiuqiang Kong et al.

Recently, there has been increasing interest in building efficient audio neural networks for on-device scenarios. Most existing approaches are designed to reduce the size of audio neural networks using methods such as model pruning. In this work, we show that instead of reducing model size using complex methods, eliminating the temporal redundancy in the input audio features (e.g., mel-spectrogram) could be an effective approach for efficient audio classification. To do so, we proposed a family of simple pooling front-ends (SimPFs) which use simple non-parametric pooling operations to reduce the redundant information within the mel-spectrogram. We perform extensive experiments on four audio classification tasks to evaluate the performance of SimPFs. Experimental results show that SimPFs can achieve a reduction in more than half of the number of floating point operations (FLOPs) for off-the-shelf audio neural networks, with negligible degradation or even some improvements in audio classification performance.

Qiushi Huang, Yu Zhang, Tom Ko et al.

Persona-based dialogue systems aim to generate consistent responses based on historical context and predefined persona. Unlike conventional dialogue generation, the persona-based dialogue needs to consider both dialogue context and persona, posing a challenge for coherent training. Specifically, this requires a delicate weight balance between context and persona. To achieve that, in this paper, we propose an effective framework with Persona-Adaptive Attention (PAA), which adaptively integrates the weights from the persona and context information via our designed attention. In addition, a dynamic masking mechanism is applied to the PAA to not only drop redundant information in context and persona but also serve as a regularization mechanism to avoid overfitting. Experimental results demonstrate the superiority of the proposed PAA framework compared to the strong baselines in both automatic and human evaluation. Moreover, the proposed PAA approach can perform equivalently well in a low-resource regime compared to models trained in a full-data setting, which achieve a similar result with only 20% to 30% of data compared to the larger models trained in the full-data setting. To fully exploit the effectiveness of our design, we designed several variants for handling the weighted information in different ways, showing the necessity and sufficiency of our weighting and masking designs.

Haohe Liu, Xubo Liu, Qiuqiang Kong et al.

The audio spectrogram is a time-frequency representation that has been widely used for audio classification. One of the key attributes of the audio spectrogram is the temporal resolution, which depends on the hop size used in the Short-Time Fourier Transform (STFT). Previous works generally assume the hop size should be a constant value (e.g., 10 ms). However, a fixed temporal resolution is not always optimal for different types of sound. The temporal resolution affects not only classification accuracy but also computational cost. This paper proposes a novel method, DiffRes, that enables differentiable temporal resolution modeling for audio classification. Given a spectrogram calculated with a fixed hop size, DiffRes merges non-essential time frames while preserving important frames. DiffRes acts as a "drop-in" module between an audio spectrogram and a classifier and can be jointly optimized with the classification task. We evaluate DiffRes on five audio classification tasks, using mel-spectrograms as the acoustic features, followed by off-the-shelf classifier backbones. Compared with previous methods using the fixed temporal resolution, the DiffRes-based method can achieve the equivalent or better classification accuracy with at least 25% computational cost reduction. We further show that DiffRes can improve classification accuracy by increasing the temporal resolution of input acoustic features, without adding to the computational cost.

Yang Xiao, Xubo Liu, James King et al.

Continuously learning new classes without catastrophic forgetting is a challenging problem for on-device environmental sound classification given the restrictions on computation resources (e.g., model size, running memory). To address this issue, we propose a simple and efficient continual learning method. Our method selects the historical data for the training by measuring the per-sample classification uncertainty. Specifically, we measure the uncertainty by observing how the classification probability of data fluctuates against the parallel perturbations added to the classifier embedding. In this way, the computation cost can be significantly reduced compared with adding perturbation to the raw data. Experimental results on the DCASE 2019 Task 1 and ESC-50 dataset show that our proposed method outperforms baseline continual learning methods on classification accuracy and computational efficiency, indicating our method can efficiently and incrementally learn new classes without the catastrophic forgetting problem for on-device environmental sound classification.

Yi Yuan, Haohe Liu, Jinhua Liang et al.

Deep neural networks have recently achieved breakthroughs in sound generation. Despite the outstanding sample quality, current sound generation models face issues on small-scale datasets (e.g., overfitting), significantly limiting performance. In this paper, we make the first attempt to investigate the benefits of pre-training on sound generation with AudioLDM, the cutting-edge model for audio generation, as the backbone. Our study demonstrates the advantages of the pre-trained AudioLDM, especially in data-scarcity scenarios. In addition, the baselines and evaluation protocol for sound generation systems are not consistent enough to compare different studies directly. Aiming to facilitate further study on sound generation tasks, we benchmark the sound generation task on various frequently-used datasets. We hope our results on transfer learning and benchmarks can provide references for further research on conditional sound generation.

Jianyuan Sun, Xubo Liu, Xinhao Mei et al.

Acoustic scene classification (ASC) aims to classify an audio clip based on the characteristic of the recording environment. In this regard, deep learning based approaches have emerged as a useful tool for ASC problems. Conventional approaches to improving the classification accuracy include integrating auxiliary methods such as attention mechanism, pre-trained models and ensemble multiple sub-networks. However, due to the complexity of audio clips captured from different environments, it is difficult to distinguish their categories without using any auxiliary methods for existing deep learning models using only a single classifier. In this paper, we propose a novel approach for ASC using deep neural decision forest (DNDF). DNDF combines a fixed number of convolutional layers and a decision forest as the final classifier. The decision forest consists of a fixed number of decision tree classifiers, which have been shown to offer better classification performance than a single classifier in some datasets. In particular, the decision forest differs substantially from traditional random forests as it is stochastic, differentiable, and capable of using the back-propagation to update and learn feature representations in neural network. Experimental results on the DCASE2019 and ESC-50 datasets demonstrate that our proposed DNDF method improves the ASC performance in terms of classification accuracy and shows competitive performance as compared with state-of-the-art baselines.

Junqi Zhao, Xubo Liu, Jinzheng Zhao et al.

Universal sound separation (USS) is a task of separating mixtures of arbitrary sound sources. Typically, universal separation models are trained from scratch in a supervised manner, using labeled data. Self-supervised learning (SSL) is an emerging deep learning approach that leverages unlabeled data to obtain task-agnostic representations, which can benefit many downstream tasks. In this paper, we propose integrating a self-supervised pre-trained model, namely the audio masked autoencoder (A-MAE), into a universal sound separation system to enhance its separation performance. We employ two strategies to utilize SSL embeddings: freezing or updating the parameters of A-MAE during fine-tuning. The SSL embeddings are concatenated with the short-time Fourier transform (STFT) to serve as input features for the separation model. We evaluate our methods on the AudioSet dataset, and the experimental results indicate that the proposed methods successfully enhance the separation performance of a state-of-the-art ResUNet-based USS model.

Xinhao Mei, Xubo Liu, Jianyuan Sun et al.

Automated audio captioning is a cross-modal translation task for describing the content of audio clips with natural language sentences. This task has attracted increasing attention and substantial progress has been made in recent years. Captions generated by existing models are generally faithful to the content of audio clips, however, these machine-generated captions are often deterministic (e.g., generating a fixed caption for a given audio clip), simple (e.g., using common words and simple grammar), and generic (e.g., generating the same caption for similar audio clips). When people are asked to describe the content of an audio clip, different people tend to focus on different sound events and describe an audio clip diversely from various aspects using distinct words and grammar. We believe that an audio captioning system should have the ability to generate diverse captions, either for a fixed audio clip, or across similar audio clips. To this end, we propose an adversarial training framework based on a conditional generative adversarial network (C-GAN) to improve diversity of audio captioning systems. A caption generator and two hybrid discriminators compete and are learned jointly, where the caption generator can be any standard encoder-decoder captioning model used to generate captions, and the hybrid discriminators assess the generated captions from different criteria, such as their naturalness and semantics. We conduct experiments on the Clotho dataset. The results show that our proposed model can generate captions with better diversity as compared to state-of-the-art methods.

Haohe Liu, Qiuqiang Kong, Xubo Liu et al.

This study defines a new evaluation metric for audio tagging tasks to overcome the limitation of the conventional mean average precision (mAP) metric, which treats different kinds of sound as independent classes without considering their relations. Also, due to the ambiguities in sound labeling, the labels in the training and evaluation set are not guaranteed to be accurate and exhaustive, which poses challenges for robust evaluation with mAP. The proposed metric, ontology-aware mean average precision (OmAP) addresses the weaknesses of mAP by utilizing the AudioSet ontology information during the evaluation. Specifically, we reweight the false positive events in the model prediction based on the ontology graph distance to the target classes. The OmAP measure also provides more insights into model performance by evaluations with different coarse-grained levels in the ontology graph. We conduct human evaluations and demonstrate that OmAP is more consistent with human perception than mAP. To further verify the importance of utilizing the ontology information, we also propose a novel loss function (OBCE) that reweights binary cross entropy (BCE) loss based on the ontology distance. Our experiment shows that OBCE can improve both mAP and OmAP metrics on the AudioSet tagging task.

Yi Yuan, Haohe Liu, Xubo Liu et al.

Foley sound generation aims to synthesise the background sound for multimedia content. Previous models usually employ a large development set with labels as input (e.g., single numbers or one-hot vector). In this work, we propose a diffusion model based system for Foley sound generation with text conditions. To alleviate the data scarcity issue, our model is initially pre-trained with large-scale datasets and fine-tuned to this task via transfer learning using the contrastive language-audio pertaining (CLAP) technique. We have observed that the feature embedding extracted by the text encoder can significantly affect the performance of the generation model. Hence, we introduce a trainable layer after the encoder to improve the text embedding produced by the encoder. In addition, we further refine the generated waveform by generating multiple candidate audio clips simultaneously and selecting the best one, which is determined in terms of the similarity score between the embedding of the candidate clips and the embedding of the target text label. Using the proposed method, our system ranks ${1}^{st}$ among the systems submitted to DCASE Challenge 2023 Task 7. The results of the ablation studies illustrate that the proposed techniques significantly improve sound generation performance. The codes for implementing the proposed system are available online.

Arshdeep Singh, James A King, Xubo Liu et al.

This technical report describes the SurreyAudioTeam22s submission for DCASE 2022 ASC Task 1, Low-Complexity Acoustic Scene Classification (ASC). The task has two rules, (a) the ASC framework should have maximum 128K parameters, and (b) there should be a maximum of 30 millions multiply-accumulate operations (MACs) per inference. In this report, we present low-complexity systems for ASC that follow the rules intended for the task.

Jianyuan Sun, Xubo Liu, Xinhao Mei et al.

Automated audio captioning (AAC) aims to describe the content of an audio clip using simple sentences. Existing AAC methods are developed based on an encoder-decoder architecture that success is attributed to the use of a pre-trained CNN10 called PANNs as the encoder to learn rich audio representations. AAC is a highly challenging task due to its high-dimensional talent space involves audio of various scenarios. Existing methods only use the high-dimensional representation of the PANNs as the input of the decoder. However, the low-dimension representation may retain as much audio information as the high-dimensional representation may be neglected. In addition, although the high-dimensional approach may predict the audio captions by learning from existing audio captions, which lacks robustness and efficiency. To deal with these challenges, a fusion model which integrates low- and high-dimensional features AAC framework is proposed. In this paper, a new encoder-decoder framework is proposed called the Low- and High-Dimensional Feature Fusion (LHDFF) model for AAC. Moreover, in LHDFF, a new PANNs encoder is proposed called Residual PANNs (RPANNs) by fusing the low-dimensional feature from the intermediate convolution layer output and the high-dimensional feature from the final layer output of PANNs. To fully explore the information of the low- and high-dimensional fusion feature and high-dimensional feature respectively, we proposed dual transformer decoder structures to generate the captions in parallel. Especially, a probabilistic fusion approach is proposed that can ensure the overall performance of the system is improved by concentrating on the respective advantages of the two transformer decoders. Experimental results show that LHDFF achieves the best performance on the Clotho and AudioCaps datasets compared with other existing models

Yaru Chen, Ruohao Guo, Xubo Liu et al.

Audio-visual video parsing is the task of categorizing a video at the segment level with weak labels, and predicting them as audible or visible events. Recent methods for this task leverage the attention mechanism to capture the semantic correlations among the whole video across the audio-visual modalities. However, these approaches have overlooked the importance of individual segments within a video and the relationship among them, and tend to rely on a single modality when learning features. In this paper, we propose a novel interactive-enhanced cross-modal perception method~(CM-PIE), which can learn fine-grained features by applying a segment-based attention module. Furthermore, a cross-modal aggregation block is introduced to jointly optimize the semantic representation of audio and visual signals by enhancing inter-modal interactions. The experimental results show that our model offers improved parsing performance on the Look, Listen, and Parse dataset compared to other methods.

Umberto Cappellazzo, Xubo Liu, Pingchuan Ma et al.

Large language models (LLMs) have recently achieved impressive results in speech recognition across multiple modalities, including Auditory Speech Recognition (ASR), Visual Speech Recognition (VSR), and Audio-Visual Speech Recognition (AVSR). Despite this progress, current LLM-based approaches typically address each task independently, training separate models that raise computational and deployment resource use while missing potential cross-task synergies. They also rely on fixed-rate token compression, which restricts flexibility in balancing accuracy with efficiency. These limitations highlight the need for a unified framework that can support ASR, VSR, and AVSR while enabling elastic inference. To this end, we present Omni-AVSR, a unified audio-visual LLM that combines efficient multi-granularity training with parameter-efficient adaptation. Specifically, we adapt the matryoshka representation learning paradigm to efficiently train across multiple audio and visual granularities, reducing its inherent training resource use. Furthermore, we explore three LoRA-based strategies for adapting the backbone LLM, balancing shared and task-specific specialization. Experiments on LRS2 and LRS3 show that Omni-AVSR achieves comparable or superior accuracy to state-of-the-art baselines while training a single model at substantially lower training and deployment resource use. The model also remains robust under acoustic noise, and we analyze its scaling behavior as LLM size increases, providing insights into the trade-off between performance and efficiency.

Xiaoyu Bie, Xubo Liu, Gaël Richard

Neural audio codecs have significantly advanced audio compression by efficiently converting continuous audio signals into discrete tokens. These codecs preserve high-quality sound and enable sophisticated sound generation through generative models trained on these tokens. However, existing neural codec models are typically trained on large, undifferentiated audio datasets, neglecting the essential discrepancies between sound domains like speech, music, and environmental sound effects. This oversight complicates data modeling and poses additional challenges to the controllability of sound generation. To tackle these issues, we introduce the Source-Disentangled Neural Audio Codec (SD-Codec), a novel approach that combines audio coding and source separation. By jointly learning audio resynthesis and separation, SD-Codec explicitly assigns audio signals from different domains to distinct codebooks, sets of discrete representations. Experimental results indicate that SD-Codec not only maintains competitive resynthesis quality but also, supported by the separation results, demonstrates successful disentanglement of different sources in the latent space, thereby enhancing interpretability in audio codec and providing potential finer control over the audio generation process.

Yuancheng Wang, Zhenyu Tang, Yun Wang et al.

Speech tokenizers are foundational to speech language models, yet existing approaches face two major challenges: (1) balancing trade-offs between encoding semantics for understanding and acoustics for reconstruction, and (2) achieving low bit rates and low token rates. We propose Speech Diffusion Tokenizer (SiTok), a diffusion autoencoder that jointly learns semantic-rich representations through supervised learning and enables high-fidelity audio reconstruction with diffusion. We scale SiTok to 1.6B parameters and train it on 2 million hours of speech. Experiments show that SiTok outperforms strong baselines on understanding, reconstruction and generation tasks, at an extremely low token rate of $12.5$ Hz and a bit-rate of 200 bits-per-second.

Ruxue Yan, Xubo Liu, Wenya Guo et al.

Autoregressive image generation has seen recent improvements with the introduction of chain-of-thought and reinforcement learning. However, current methods merely specify "What" details to depict by rewriting the input prompt, yet fundamentally fail to reason about "How" to structure the overall image. This inherent limitation gives rise to persistent issues, such as spatial ambiguity directly causing unrealistic object overlaps. To bridge this gap, we propose CoR-Painter, a novel framework that pioneers a "How-to-What" paradigm by introducing Constrained Reasoning to guide the autoregressive generation. Specifically, it first deduces "How to draw" by deriving a set of visual constraints from the input prompt, which explicitly govern spatial relationships, key attributes, and compositional rules. These constraints steer the subsequent generation of a detailed description "What to draw", providing a structurally sound and coherent basis for accurate visual synthesis. Additionally, we introduce a Dual-Objective GRPO strategy that specifically optimizes the textual constrained reasoning and visual projection processes to ensure the coherence and quality of the entire generation pipeline. Extensive experiments on T2I-CompBench, GenEval, and WISE demonstrate that our method achieves state-of-the-art performance, with significant improvements in spatial metrics (e.g., +5.41% on T2I-CompBench).

Mengyu Zhang, Xubo Liu, Siyu Ding et al.

Reinforcement Learning with Verifiable Rewards (RLVR) has demonstrated great potential in enhancing the reasoning capabilities of large language models (LLMs), achieving remarkable progress in domains such as mathematics and programming where standard answers are available. However, for open-ended tasks lacking ground-truth solutions (e.g., creative writing and instruction following), existing studies typically regard them as non-reasoning scenarios, thereby overlooking the latent value of reasoning capabilities. This raises a key question: Can strengthening reasoning improve performance in open-ended tasks? To address this, we explore the transfer of the RLVR paradigm to the open domain. Yet, since RLVR fundamentally relies on verifiers that presuppose the existence of standard answers, it cannot be directly applied to open-ended tasks. To overcome this challenge, we introduce Verifiable Multiple-Choice Reformulation (VMR), a novel training strategy that restructures open-ended data into verifiable multiple-choice formats, enabling effective training even in the absence of explicit ground truth. Experimental results on multiple benchmarks validate the effectiveness of our method in improving LLM performance on open-ended tasks. Notably, across eight open-ended benchmarks, our VMR-based training delivers an average gain of 5.99 points over the baseline. Code will be released upon acceptance to facilitate reproducibility.

Yuhang He, Yash Jain, Xubo Liu et al.

Despite significant advancements in Text-to-Audio (TTA) generation models achieving high-fidelity audio with fine-grained context understanding, they struggle to model the relations between audio events described in the input text. However, previous TTA methods have not systematically explored audio event relation modeling, nor have they proposed frameworks to enhance this capability. In this work, we systematically study audio event relation modeling in TTA generation models. We first establish a benchmark for this task by: 1. proposing a comprehensive relation corpus covering all potential relations in real-world scenarios; 2. introducing a new audio event corpus encompassing commonly heard audios; and 3. proposing new evaluation metrics to assess audio event relation modeling from various perspectives. Furthermore, we propose a finetuning framework to enhance existing TTA models ability to model audio events relation. Code is available at: https://github.com/yuhanghe01/RiTTA

Qiushi Huang, Shuai Fu, Xubo Liu et al.

Personalized dialogue generation, focusing on generating highly tailored responses by leveraging persona profiles and dialogue context, has gained significant attention in conversational AI applications. However, persona profiles, a prevalent setting in current personalized dialogue datasets, typically composed of merely four to five sentences, may not offer comprehensive descriptions of the persona about the agent, posing a challenge to generate truly personalized dialogues. To handle this problem, we propose $\textbf{L}$earning Retrieval $\textbf{A}$ugmentation for $\textbf{P}$ersonalized $\textbf{D}$ial$\textbf{O}$gue $\textbf{G}$eneration ($\textbf{LAPDOG}$), which studies the potential of leveraging external knowledge for persona dialogue generation. Specifically, the proposed LAPDOG model consists of a story retriever and a dialogue generator. The story retriever uses a given persona profile as queries to retrieve relevant information from the story document, which serves as a supplementary context to augment the persona profile. The dialogue generator utilizes both the dialogue history and the augmented persona profile to generate personalized responses. For optimization, we adopt a joint training framework that collaboratively learns the story retriever and dialogue generator, where the story retriever is optimized towards desired ultimate metrics (e.g., BLEU) to retrieve content for the dialogue generator to generate personalized responses. Experiments conducted on the CONVAI2 dataset with ROCStory as a supplementary data source show that the proposed LAPDOG method substantially outperforms the baselines, indicating the effectiveness of the proposed method. The LAPDOG model code is publicly available for further exploration. https://github.com/hqsiswiliam/LAPDOG

Qiushi Huang, Xubo Liu, Tom Ko et al.

In conversational AI, personalizing dialogues with persona profiles and contextual understanding is essential. Despite large language models' (LLMs) improved response coherence, effective persona integration remains a challenge. In this work, we first study two common approaches for personalizing LLMs: textual prompting and direct fine-tuning. We observed that textual prompting often struggles to yield responses that are similar to the ground truths in datasets, while direct fine-tuning tends to produce repetitive or overly generic replies. To alleviate those issues, we propose \textbf{S}elective \textbf{P}rompt \textbf{T}uning (SPT), which softly prompts LLMs for personalized conversations in a selective way. Concretely, SPT initializes a set of soft prompts and uses a trainable dense retriever to adaptively select suitable soft prompts for LLMs according to different input contexts, where the prompt retriever is dynamically updated through feedback from the LLMs. Additionally, we propose context-prompt contrastive learning and prompt fusion learning to encourage the SPT to enhance the diversity of personalized conversations. Experiments on the CONVAI2 dataset demonstrate that SPT significantly enhances response diversity by up to 90\%, along with improvements in other critical performance indicators. Those results highlight the efficacy of SPT in fostering engaging and personalized dialogue generation. The SPT model code (https://github.com/hqsiswiliam/SPT) is publicly available for further exploration.

Qixin Deng, Qikai Yang, Ruibin Yuan et al.

Music composition represents the creative side of humanity, and itself is a complex task that requires abilities to understand and generate information with long dependency and harmony constraints. While demonstrating impressive capabilities in STEM subjects, current LLMs easily fail in this task, generating ill-written music even when equipped with modern techniques like In-Context-Learning and Chain-of-Thoughts. To further explore and enhance LLMs' potential in music composition by leveraging their reasoning ability and the large knowledge base in music history and theory, we propose ComposerX, an agent-based symbolic music generation framework. We find that applying a multi-agent approach significantly improves the music composition quality of GPT-4. The results demonstrate that ComposerX is capable of producing coherent polyphonic music compositions with captivating melodies, while adhering to user instructions.

Julian D Parker, Anton Smirnov, Jordi Pons et al.

The tokenization of speech with neural audio codec models is a vital part of modern AI pipelines for the generation or understanding of speech, alone or in a multimodal context. Traditionally such tokenization models have concentrated on low parameter-count architectures using only components with strong inductive biases. In this work we show that by scaling a transformer architecture with large parameter count to this problem, and applying a flexible Finite Scalar Quantization (FSQ) based bottleneck, it is possible to reach state-of-the-art speech quality at extremely low bit-rates of $400$ or $700$ bits-per-second. The trained models strongly out-perform existing baselines in both objective and subjective tests.

Yi Yuan, Zhuo Chen, Xubo Liu et al.

Contrastive language-audio pretraining~(CLAP) has been developed to align the representations of audio and language, achieving remarkable performance in retrieval and classification tasks. However, current CLAP struggles to capture temporal information within audio and text features, presenting substantial limitations for tasks such as audio retrieval and generation. To address this gap, we introduce T-CLAP, a temporal-enhanced CLAP model. We use Large Language Models~(LLMs) and mixed-up strategies to generate temporal-contrastive captions for audio clips from extensive audio-text datasets. Subsequently, a new temporal-focused contrastive loss is designed to fine-tune the CLAP model by incorporating these synthetic data. We conduct comprehensive experiments and analysis in multiple downstream tasks. T-CLAP shows improved capability in capturing the temporal relationship of sound events and outperforms state-of-the-art models by a significant margin.

Junqi Zhao, Jinzheng Zhao, Haohe Liu et al.

Diffusion models have significantly improved the quality and diversity of audio generation but are hindered by slow inference speed. Rectified flow enhances inference speed by learning straight-line ordinary differential equation (ODE) paths. However, this approach requires training a flow-matching model from scratch and tends to perform suboptimally, or even poorly, at low step counts. To address the limitations of rectified flow while leveraging the advantages of advanced pre-trained diffusion models, this study integrates pre-trained models with the rectified diffusion method to improve the efficiency of text-to-audio (TTA) generation. Specifically, we propose AudioTurbo, which learns first-order ODE paths from deterministic noise sample pairs generated by a pre-trained TTA model. Experiments on the AudioCaps dataset demonstrate that our model, with only 10 sampling steps, outperforms prior models and reduces inference to 3 steps compared to a flow-matching-based acceleration model.

Meng Cui, Xianghu Yue, Xinyuan Qian et al.

Fish Feeding Intensity Assessment (FFIA) is crucial in industrial aquaculture management. Recent multi-modal approaches have shown promise in improving FFIA robustness and efficiency. However, these methods face significant challenges when adapting to new fish species or environments due to catastrophic forgetting and the lack of suitable datasets. To address these limitations, we first introduce AV-CIL-FFIA, a new dataset comprising 81,932 labelled audio-visual clips capturing feeding intensities across six different fish species in real aquaculture environments. Then, we pioneer audio-visual class incremental learning (CIL) for FFIA and demonstrate through benchmarking on AV-CIL-FFIA that it significantly outperforms single-modality methods. Existing CIL methods rely heavily on historical data. Exemplar-based approaches store raw samples, creating storage challenges, while exemplar-free methods avoid data storage but struggle to distinguish subtle feeding intensity variations across different fish species. To overcome these limitations, we introduce HAIL-FFIA, a novel audio-visual class-incremental learning framework that bridges this gap with a prototype-based approach that achieves exemplar-free efficiency while preserving essential knowledge through compact feature representations. Specifically, HAIL-FFIA employs hierarchical representation learning with a dual-path knowledge preservation mechanism that separates general intensity knowledge from fish-specific characteristics. Additionally, it features a dynamic modality balancing system that adaptively adjusts the importance of audio versus visual information based on feeding behaviour stages. Experimental results show that HAIL-FFIA is superior to SOTA methods on AV-CIL-FFIA, achieving higher accuracy with lower storage needs while effectively mitigating catastrophic forgetting in incremental fish species learning.

Peng Gao, Yujian Lee, Zailong Chen et al.

Composed Image Retrieval (CIR) seeks to find a target image using a multi-modal query, which combines an image with modification text to pinpoint the target. While recent CIR methods have shown promise, they mainly focus on exploring relationships between the query pairs (image and text) through data augmentation or model design. These methods often assume perfect alignment between queries and target images, an idealized scenario rarely encountered in practice. In reality, pairs are often partially or completely mismatched due to issues like inaccurate modification texts, low-quality target images, and annotation errors. Ignoring these mismatches leads to numerous False Positive Pair (FFPs) denoted as noise pairs in the dataset, causing the model to overfit and ultimately reducing its performance. To address this problem, we propose the Noise-aware Contrastive Learning for CIR (NCL-CIR), comprising two key components: the Weight Compensation Block (WCB) and the Noise-pair Filter Block (NFB). The WCB coupled with diverse weight maps can ensure more stable token representations of multi-modal queries and target images. Meanwhile, the NFB, in conjunction with the Gaussian Mixture Model (GMM) predicts noise pairs by evaluating loss distributions, and generates soft labels correspondingly, allowing for the design of the soft-label based Noise Contrastive Estimation (NCE) loss function. Consequently, the overall architecture helps to mitigate the influence of mismatched and partially matched samples, with experimental results demonstrating that NCL-CIR achieves exceptional performance on the benchmark datasets.

Peng Gao, Yujian Lee, Hui Zhang et al.

Visible-infrared person re-identification (VI-ReID) aims to match people with the same identity between visible and infrared modalities. VI-ReID is a challenging task due to the large differences in individual appearance under different modalities. Existing methods generally try to bridge the cross-modal differences at image or feature level, which lacks exploring the discriminative embeddings. Effectively minimizing these cross-modal discrepancies relies on obtaining representations that are guided by identity and consistent across modalities, while also filtering out representations that are irrelevant to identity. To address these challenges, we introduce a dynamic identity-guided attention network (DIAN) to mine identity-guided and modality-consistent embeddings, facilitating effective bridging the gap between different modalities. Specifically, in DIAN, to pursue a semantically richer representation, we first use orthogonal projection to fuse the features from two connected coarse and fine layers. Furthermore, we first use dynamic convolution kernels to mine identity-guided and modality-consistent representations. More notably, a cross embedding balancing loss is introduced to effectively bridge cross-modal discrepancies by above embeddings. Experimental results on SYSU-MM01 and RegDB datasets show that DIAN achieves state-of-the-art performance. Specifically, for indoor search on SYSU-MM01, our method achieves 86.28% rank-1 accuracy and 87.41% mAP, respectively. Our code will be available soon.

Umberto Cappellazzo, Minsu Kim, Pingchuan Ma et al.

Large language models (LLMs) have recently shown strong potential in audio-visual speech recognition (AVSR), but their high computational demands and sensitivity to token granularity limit their practicality in resource-constrained settings. Token compression methods can reduce inference cost, but they require fixing a compression rate in advance and produce a single fixed-length output, offering no flexibility to balance information density and efficiency at inference time. Matryoshka representation learning (MRL) addresses this by enabling a single model to operate across multiple token granularities, allowing compression rates to be adjusted dynamically. However, current MRL-based methods treat each scale independently during training, limiting cross-scale generalization, robustness at high compression, and interpretability. To overcome these limitations, we propose MoME (Mixture of Matryoshka Experts), a novel framework that integrates sparse Mixture-of-Experts (MoE) into MRL-based LLMs for AVSR. MoME augments a frozen LLM with top-k routed and shared experts, allowing dynamic capacity allocation across scales and modalities. A shared router promotes consistent expert activation across granularities, enabling compressed sequences to benefit from representations learned at lower compression. Experiments on LRS2 and LRS3 demonstrate that MoME achieves state-of-the-art performance across AVSR, ASR, and VSR tasks, while requiring significantly fewer parameters and maintaining robustness under noise. MoME unifies the adaptability of MRL with the efficiency of MoE, offering a scalable and interpretable solution for resource-aware speech recognition.

Yi Yuan, Xubo Liu, Haohe Liu et al.

With the development of large-scale diffusion-based and language-modeling-based generative models, impressive progress has been achieved in text-to-audio generation. Despite producing high-quality outputs, existing text-to-audio models mainly aim to generate semantically aligned sound and fall short on precisely controlling fine-grained acoustic characteristics of specific sounds. As a result, users that need specific sound content may find it challenging to generate the desired audio clips. In this paper, we present DreamAudio for customized text-to-audio generation (CTTA). Specifically, we introduce a new framework that is designed to enable the model to identify auditory information from user-provided reference concepts for audio generation. Given a few reference audio samples containing personalized audio events, our system can generate new audio samples that include these specific events. In addition, two types of datasets are developed for training and testing the customized systems. The experiments show that the proposed model, DreamAudio, generates audio samples that are highly consistent with the customized audio features and aligned well with the input text prompts. Furthermore, DreamAudio offers comparable performance in general text-to-audio tasks. We also provide a human-involved dataset containing audio events from real-world CTTA cases as the benchmark for customized generation tasks.

Wenya Guo, Zhengkun Zhang, Xumeng Liu et al.

Instruction data selection aims to identify a high-quality subset from the training set that matches or exceeds the performance of the full dataset on target tasks. Existing methods focus on the instruction-to-response mapping, but neglect the human preference for diverse responses. In this paper, we propose Preference-oriented Data Selection method (ProDS) that scores training samples based on their alignment with preferences observed in the target set. Our key innovation lies in shifting the data selection criteria from merely estimating features for accurate response generation to explicitly aligning training samples with human preferences in target tasks. Specifically, direct preference optimization (DPO) is employed to estimate human preferences across diverse responses. Besides, a bidirectional preference synthesis strategy is designed to score training samples according to both positive preferences and negative preferences. Extensive experimental results demonstrate our superiority to existing task-agnostic and targeted methods.

Xinhao Mei, Xubo Liu, Jianyuan Sun et al.

Audio captioning aims at generating natural language descriptions for audio clips automatically. Existing audio captioning models have shown promising improvement in recent years. However, these models are mostly trained via maximum likelihood estimation (MLE),which tends to make captions generic, simple and deterministic. As different people may describe an audio clip from different aspects using distinct words and grammars, we argue that an audio captioning system should have the ability to generate diverse captions for a fixed audio clip and across similar audio clips. To address this problem, we propose an adversarial training framework for audio captioning based on a conditional generative adversarial network (C-GAN), which aims at improving the naturalness and diversity of generated captions. Unlike processing data of continuous values in a classical GAN, a sentence is composed of discrete tokens and the discrete sampling process is non-differentiable. To address this issue, policy gradient, a reinforcement learning technique, is used to back-propagate the reward to the generator. The results show that our proposed model can generate more diverse captions, as compared to state-of-the-art methods.

Xinhao Mei, Qiushi Huang, Xubo Liu et al.

Automated audio captioning aims to use natural language to describe the content of audio data. This paper presents an audio captioning system with an encoder-decoder architecture, where the decoder predicts words based on audio features extracted by the encoder. To improve the proposed system, transfer learning from either an upstream audio-related task or a large in-domain dataset is introduced to mitigate the problem induced by data scarcity. Besides, evaluation metrics are incorporated into the optimization of the model with reinforcement learning, which helps address the problem of ``exposure bias'' induced by ``teacher forcing'' training strategy and the mismatch between the evaluation metrics and the loss function. The resulting system was ranked 3rd in DCASE 2021 Task 6. Ablation studies are carried out to investigate how much each element in the proposed system can contribute to final performance. The results show that the proposed techniques significantly improve the scores of the evaluation metrics, however, reinforcement learning may impact adversely on the quality of the generated captions.

Xubo Liu, Turab Iqbal, Jinzheng Zhao et al.

Deep generative models have recently achieved impressive performance in speech and music synthesis. However, compared to the generation of those domain-specific sounds, generating general sounds (such as siren, gunshots) has received less attention, despite their wide applications. In previous work, the SampleRNN method was considered for sound generation in the time domain. However, SampleRNN is potentially limited in capturing long-range dependencies within sounds as it only back-propagates through a limited number of samples. In this work, we propose a method for generating sounds via neural discrete time-frequency representation learning, conditioned on sound classes. This offers an advantage in efficiently modelling long-range dependencies and retaining local fine-grained structures within sound clips. We evaluate our approach on the UrbanSound8K dataset, compared to SampleRNN, with the performance metrics measuring the quality and diversity of generated sounds. Experimental results show that our method offers comparable performance in quality and significantly better performance in diversity.

Xubo Liu, Qiushi Huang, Xinhao Mei et al.

Automated Audio captioning (AAC) is a cross-modal translation task that aims to use natural language to describe the content of an audio clip. As shown in the submissions received for Task 6 of the DCASE 2021 Challenges, this problem has received increasing interest in the community. The existing AAC systems are usually based on an encoder-decoder architecture, where the audio signal is encoded into a latent representation, and aligned with its corresponding text descriptions, then a decoder is used to generate the captions. However, training of an AAC system often encounters the problem of data scarcity, which may lead to inaccurate representation and audio-text alignment. To address this problem, we propose a novel encoder-decoder framework called Contrastive Loss for Audio Captioning (CL4AC). In CL4AC, the self-supervision signals derived from the original audio-text paired data are used to exploit the correspondences between audio and texts by contrasting samples, which can improve the quality of latent representation and the alignment between audio and texts, while trained with limited data. Experiments are performed on the Clotho dataset to show the effectiveness of our proposed approach.

Xinhao Mei, Xubo Liu, Qiushi Huang et al.

Audio captioning aims to automatically generate a natural language description of an audio clip. Most captioning models follow an encoder-decoder architecture, where the decoder predicts words based on the audio features extracted by the encoder. Convolutional neural networks (CNNs) and recurrent neural networks (RNNs) are often used as the audio encoder. However, CNNs can be limited in modelling temporal relationships among the time frames in an audio signal, while RNNs can be limited in modelling the long-range dependencies among the time frames. In this paper, we propose an Audio Captioning Transformer (ACT), which is a full Transformer network based on an encoder-decoder architecture and is totally convolution-free. The proposed method has a better ability to model the global information within an audio signal as well as capture temporal relationships between audio events. We evaluate our model on AudioCaps, which is the largest audio captioning dataset publicly available. Our model shows competitive performance compared to other state-of-the-art approaches.

Qiushi Huang, Tom Ko, H Lilian Tang et al.

Punctuation is critical in understanding natural language text. Currently, most automatic speech recognition (ASR) systems do not generate punctuation, which affects the performance of downstream tasks, such as intent detection and slot filling. This gives rise to the need for punctuation restoration. Recent work in punctuation restoration heavily utilizes pre-trained language models without considering data imbalance when predicting punctuation classes. In this work, we address this problem by proposing a token-level supervised contrastive learning method that aims at maximizing the distance of representation of different punctuation marks in the embedding space. The result shows that training with token-level supervised contrastive learning obtains up to 3.2% absolute F1 improvement on the test set.