arXiv subjects

Baoxiang Li

Publications and source records attributed to Baoxiang Li.

13 recordsLinked to original sources

arXiv2026

SALMONN-2: Advancing General-Purpose Hearing Abilities with Self-Supervised Representations

Recent audio large language models (ALLMs) are typically built upon audio encoders trained with large amounts of supervised data. Since self-supervised learning (SSL) audio encoder models are known to learn general-purpose and transferable representations, we investigate whether general-purpose SSL audio representations can serve as an effective foundation for ALLMs. We present SALMONN-2, an ALLM built upon a unified SSL encoder. To better exploit the hierarchical representations learned by SSL encoders, we propose a multi-layer feature fusion (MLF) adapter that aggregates information from all encoder layers before projecting them into the language model. Beyond conventional audio understanding tasks, we further explore multimodal in-context learning (MICL) in ALLMs and study how this capability can be acquired through contextual biasing training. Experimental results show that a general-purpose SSL encoder achieves performance comparable to, or better than, specialised supervised audio encoders while providing a more balanced capability across speech, audio, music and paralinguistic tasks. SALMONN-2 further achieves state-of-the-art performance among comparable-scale open-weight models on ALLM understanding benchmarks, obtaining the best results on MMAU-Pro, MMAR and MMSU. We also show that MICL does not emerge naturally in ALLMs, but can be effectively acquired through targeted contextual biasing training.

eess.AS↗

arXiv2026

AuDirector: A Self-Reflective Closed-Loop Framework for Immersive Audio Storytelling

Despite advances in text and visual generation, creating coherent long-form audio narratives remains challenging. Existing frameworks often exhibit limitations such as mismatched character settings with voice performance, insufficient self-correction mechanisms, and limited human interactivity. To address these challenges, we propose AuDirector, a self-reflective closed-loop multi-agent framework. Specifically, it involves an Identity-Aware Pre-production mechanism that transforms narrative texts into character profiles and utterance-level emotional instructions to retrieve suitable voice candidates and guide expressive speech synthesis, thereby promoting context-aligned voice adaptation. To enhance quality, a Collaborative Synthesis and Correction module introduces a closed-loop self-correction mechanism to systematically audit and regenerate defective audio components. Furthermore, a Human-Guided Interactive Refinement module facilitates user control by interpreting natural language feedback to interactively refine the underlying scripts. Experiments demonstrate that AuDirector achieves superior performance compared to state-of-the-art baselines in structural coherence, emotional expressiveness, and acoustic fidelity. Audio samples can be found at https://anonymous-itsh.github.io/.

cs.SD↗

arXiv2026

HoliAntiSpoof: Audio LLM for Holistic Speech Anti-Spoofing

Recent advances in speech synthesis and editing have made speech spoofing increasingly challenging. However, most existing methods treat spoofing as binary classification, overlooking that diverse spoofing techniques manipulate multiple, coupled speech attributes and their semantic effects. In this paper, we introduce HoliAntiSpoof, the first audio large language model (ALLM) framework for holistic speech anti-spoofing analysis. HoliAntiSpoof reformulates spoofing analysis as a unified text generation task, enabling joint reasoning over spoofing methods, affected speech attributes, and their semantic impacts. To support semantic-level analysis, we introduce DailyTalkEdit, a new anti-spoofing benchmark that simulates realistic conversational manipulations and provides annotations of semantic influence. Extensive experiments demonstrate that HoliAntiSpoof outperforms conventional baselines across multiple settings, while preliminary results show that in-context learning further improves out-of-domain generalization. These findings indicate that ALLMs not only enhance speech spoofing detection performance but also enable interpretable analysis of spoofing behaviors and their semantic effects, pointing towards more trustworthy and explainable speech security. Data and code are publicly available.

cs.SD↗

arXiv2025

SPEAR: A Unified SSL Framework for Learning Speech and Audio Representations

Self-supervised learning (SSL) has significantly advanced acoustic representation learning. However, most existing models are optimised for either speech or audio event understanding, resulting in a persistent gap between these two domains. We address this gap with SPEAR (SPEech and Audio Representations), a self-supervised framework that distils complementary knowledge from a speech-focused SSL teacher and a general-audio SSL teacher into a single unified model. SPEAR applies multi-codebook vector quantisation to continuous teacher representations to produce fine-grained discrete tokens that capture both semantic and acoustic information. To effectively integrate these heterogeneous representations, SPEAR jointly predicts them given a masked input with an asymmetric pre-training loss. We further improve robustness in complex sound scenes through a novel token mixing mechanism. Extensive experiments demonstrate that SPEAR consistently outperforms existing unified speech and audio models. SPEAR establishes a new state-of-the-art on the SUPERB benchmark, surpassing WavLM Large on 12 of 15 tasks, while achieving competitive performance on the HEAR benchmark. These results position SPEAR as a versatile foundation for general-purpose speech and audio representation learning. The code and pre-trained models will be released.

eess.AS↗

arXiv2025

Bayesian Speech Synthesizers Can Learn from Multiple Teachers

Text-to-Speech (TTS) is inherently a "one-to-many" mapping characterized by intrinsic uncertainty, yet current paradigms often oversimplify it into a deterministic regression task. While continuous-valued autoregressive (AR) models have recently emerged as a promising alternative to discrete codec-based approaches, they typically rely on a fixed-variance prior, fundamentally constraining generation to a static point estimate that ignores the dynamic variability of natural speech. To bridge this gap, we propose BELLE (Bayesian evidential learning with language modelling), a framework that shifts from deterministic prediction to principled Bayesian inference without increasing model parameters or inference latency. By modeling the acoustic target as a Normal-Inverse-Gamma distribution, BELLE captures data-dependent aleatoric uncertainty. To enable accurate variance estimation on standard single-reference datasets, we introduce a "one-to-many" training strategy that leverages synthetic samples as a statistical support set, allowing the model to learn robust distributional properties rather than merely imitating teacher artifacts. Experiments demonstrate that BELLE, trained on only ~5k hours of data, outperforms leading open-source models trained on 50k hours (achieving a 25.8% relative WER reduction) and naturally supports high-quality streaming generation. Audio samples are available at https://belletts.github.io/Belle/.

cs.SD↗

arXiv2025

Can Audio Large Language Models Verify Speaker Identity?

This paper investigates adapting Audio Large Language Models (ALLMs) for speaker verification (SV). We reformulate SV as an audio question-answering task and conduct comprehensive zero-shot evaluations on public benchmarks, showing that current ALLMs have limited zero-shot SV capability and often struggle in diverse acoustic conditions. To address this challenge, we perform supervised fine-tuning on speaker verification data. A rule-based hard pair sampling strategy is proposed to construct more challenging training pairs. Lightweight fine-tuning substantially improves the performance, though there is still a gap between ALLMs and conventional models. Then, we extend to text-dependent SV by jointly querying ALLMs to verify speaker identity and spoken content, yielding results competitive with cascaded ASR-SV systems. Our findings demonstrate that with proper adaptation, ALLMs hold substantial potential as a unified model for robust speaker verification systems, while maintaining the general audio understanding capabilities.

cs.SD↗

arXiv2025

Robust Residual Finite Scalar Quantization for Neural Compression

Finite Scalar Quantization (FSQ) offers simplified training but suffers from residual magnitude decay in multi-stage settings, where subsequent stages receive exponentially weaker signals. We propose Robust Residual Finite Scalar Quantization (RFSQ), addressing this fundamental limitation through two novel conditioning strategies: learnable scaling factors and invertible layer normalization. Our experiments across audio and image modalities demonstrate RFSQ's effectiveness and generalizability. In audio reconstruction at 24 bits/frame, RFSQ-LayerNorm achieves 3.646 DNSMOS, a 3.6% improvement over state-of-the-art RVQ (3.518). On ImageNet, RFSQ achieves 0.102 L1 loss and 0.100 perceptual loss, with LayerNorm providing 9.7% L1 improvement and 17.4% perceptual improvement over unconditioned variants. The LayerNorm strategy consistently outperforms alternatives by maintaining normalized input statistics across stages, effectively preventing exponential magnitude decay that limits naive residual approaches. RFSQ combines FSQ's simplicity with multi-stage quantization's representational power, establishing a new standard for neural compression across diverse modalities.

eess.IV↗

arXiv2025

Programming Geotechnical Reliability Algorithms using Generative AI

Programming reliability algorithms is crucial for risk assessment in geotechnical engineering. This study explores the possibility of automating and accelerating this task using Generative AI based on Large Language Models (LLMs). Specifically, the most popular LLM, i.e., ChatGPT, is used to test the ability to generate MATLAB codes for four classical reliability algorithms. The four specific examples considered in this study are: (1) First Order Reliability Method (FORM); (2) Subset simulation; (3) Random field simulation; and (4) Bayesian update using Gibbs sampling. The results obtained using the generated codes are compared with benchmark methods. It is found that the use of LLMs can be promising for generating reliability codes. Failure, limitations, and challenges of adopting LLMs are also discussed. Overall, this study demonstrates that existing LLMs can be leveraged powerfully and can contribute toward accelerating the adoption of reliability techniques in routine geotechnical engineering.

stat.AP↗

arXiv2024

CIF-T: A Novel CIF-based Transducer Architecture for Automatic Speech Recognition

RNN-T models are widely used in ASR, which rely on the RNN-T loss to achieve length alignment between input audio and target sequence. However, the implementation complexity and the alignment-based optimization target of RNN-T loss lead to computational redundancy and a reduced role for predictor network, respectively. In this paper, we propose a novel model named CIF-Transducer (CIF-T) which incorporates the Continuous Integrate-and-Fire (CIF) mechanism with the RNN-T model to achieve efficient alignment. In this way, the RNN-T loss is abandoned, thus bringing a computational reduction and allowing the predictor network a more significant role. We also introduce Funnel-CIF, Context Blocks, Unified Gating and Bilinear Pooling joint network, and auxiliary training strategy to further improve performance. Experiments on the 178-hour AISHELL-1 and 10000-hour WenetSpeech datasets show that CIF-T achieves state-of-the-art results with lower computational overhead compared to RNN-T models.

cs.SD↗

arXiv2023

Dual-path Transformer Based Neural Beamformer for Target Speech Extraction

Neural beamformers, which integrate both pre-separation and beamforming modules, have demonstrated impressive effectiveness in target speech extraction. Nevertheless, the performance of these beamformers is inherently limited by the predictive accuracy of the pre-separation module. In this paper, we introduce a neural beamformer supported by a dual-path transformer. Initially, we employ the cross-attention mechanism in the time domain to extract crucial spatial information related to beamforming from the noisy covariance matrix. Subsequently, in the frequency domain, the self-attention mechanism is employed to enhance the model's ability to process frequency-specific details. By design, our model circumvents the influence of pre-separation modules, delivering performance in a more comprehensive end-to-end manner. Experimental results reveal that our model not only outperforms contemporary leading neural beamforming algorithms in separation performance but also achieves this with a significant reduction in parameter count.

cs.SD↗

arXiv2022

A Polyphone BERT for Polyphone Disambiguation in Mandarin Chinese

Grapheme-to-phoneme (G2P) conversion is an indispensable part of the Chinese Mandarin text-to-speech (TTS) system, and the core of G2P conversion is to solve the problem of polyphone disambiguation, which is to pick up the correct pronunciation for several candidates for a Chinese polyphonic character. In this paper, we propose a Chinese polyphone BERT model to predict the pronunciations of Chinese polyphonic characters. Firstly, we create 741 new Chinese monophonic characters from 354 source Chinese polyphonic characters by pronunciation. Then we get a Chinese polyphone BERT by extending a pre-trained Chinese BERT with 741 new Chinese monophonic characters and adding a corresponding embedding layer for new tokens, which is initialized by the embeddings of source Chinese polyphonic characters. In this way, we can turn the polyphone disambiguation task into a pre-training task of the Chinese polyphone BERT. Experimental results demonstrate the effectiveness of the proposed model, and the polyphone BERT model obtain 2% (from 92.1% to 94.1%) improvement of average accuracy compared with the BERT-based classifier model, which is the prior state-of-the-art in polyphone disambiguation.

eess.AS↗

arXiv2022

Locality Matters: A Locality-Biased Linear Attention for Automatic Speech Recognition

Conformer has shown a great success in automatic speech recognition (ASR) on many public benchmarks. One of its crucial drawbacks is the quadratic time-space complexity with respect to the input sequence length, which prohibits the model to scale-up as well as process longer input audio sequences. To solve this issue, numerous linear attention methods have been proposed. However, these methods often have limited performance on ASR as they treat tokens equally in modeling, neglecting the fact that the neighbouring tokens are often more connected than the distanced tokens. In this paper, we take this fact into account and propose a new locality-biased linear attention for Conformer. It not only achieves higher accuracy than the vanilla Conformer, but also enjoys linear space-time computational complexity. To be specific, we replace the softmax attention with a locality-biased linear attention (LBLA) mechanism in Conformer blocks. The LBLA contains a kernel function to ensure the linear complexities and a cosine reweighing matrix to impose more weights on neighbouring tokens. Extensive experiments on the LibriSpeech corpus show that by introducing this locality bias to the Conformer, our method achieves a lower word error rate with more than 22% inference speed.

cs.SD↗

arXiv2022

Dynamic Latency for CTC-Based Streaming Automatic Speech Recognition With Emformer

An inferior performance of the streaming automatic speech recognition models versus non-streaming model is frequently seen due to the absence of future context. In order to improve the performance of the streaming model and reduce the computational complexity, a frame-level model using efficient augment memory transformer block and dynamic latency training method is employed for streaming automatic speech recognition in this paper. The long-range history context is stored into the augment memory bank as a complement to the limited history context used in the encoder. Key and value are cached by a cache mechanism and reused for next chunk to reduce computation. Afterwards, a dynamic latency training method is proposed to obtain better performance and support low and high latency inference simultaneously. Our experiments are conducted on benchmark 960h LibriSpeech data set. With an average latency of 640ms, our model achieves a relative WER reduction of 6.0% on test-clean and 3.0% on test-other versus the truncate chunk-wise Transformer.

cs.SD↗