arXiv subjects

Duc Le

Publications and source records attributed to Duc Le.

At least 19 recordsLinked to original sources

arXiv2026

MELD: Mel-Spectrogram-Based Speech Language Modeling with Discrete Latent Variables

Recent speech language models rely on encoders that are optimized separately from autoregressive models. Since these encoders are unaware of the downstream objectives, the extracted representations may not be optimal for downstream tasks. To address this limitation, we introduce a discrete latent variable model on mel spectrograms that jointly optimizes the encoder and the speech language model. Joint optimization not only brings improvements over codec-based and other mel-spectrogram-based baselines on zero-shot Text-to-Speech (TTS) and Speech-to-Text (STT) tasks, but also effectively alleviates common issues in autoregressive mel spectrogram modeling, such as prolonged silence generation and word omissions.

eess.AS↗

arXiv2025

Quasiparticle Dynamics in the 4d-4f Ising-like Double Perovskite Ba2DyRuO6 studied using Neutron Scattering and Machine-Learning Framework

Double perovskites containing 4d--4f interactions provide a platform to study complex magnetic phenomena in correlated systems. Here, we investigate the magnetic ground state and quasiparticle excitations of the fascinating double perovskite system, Ba$_2$DyRuO$_6$, through Time of flight (TOF) neutron diffraction (TOF), inelastic neutron scattering (INS), and theoretical modelling. The compound Ba$_2$DyRuO$_6$ is reported to exhibit a single magnetic transition, in sharp contrast to most of the other rare-earth (R) members in this family, A$_2$RRuO$_6$ (A = Ca/Sr/Ba), which typically show magnetic ordering of the Ru ions, followed by R-ion ordering. Our neutron diffraction results confirm that long-range antiferromagnetic order emerges at $T_\mathrm{N} \approx 47$~K, primarily driven by 4d--4f Ru$^{5+}$--Dy$^{3+}$ exchange interactions, where both Dy and Ru moments start to order simultaneously. The ordered ground state is a collinear antiferromagnet with Ising character, carrying ordered moments of $\mu_{\mathrm{Ru}} = 1.6(1)~\mu_\mathrm{B}$ and $\mu_{\mathrm{Dy}} = 5.1(1)~\mu_\mathrm{B}$ at 1.5~K. Low-temperature INS reveals well-defined magnon excitations below 10~meV. SpinW modelling of the INS spectra evidences complex exchange interactions and the presence of magnetic anisotropy, which governs the Ising ground state and accounts for the observed magnon spectrum. Combined INS and Raman spectroscopy reveal crystal-electric-field (CEF) excitations of Dy$^{3+}$ at 46.5 and 71.8~meV in the paramagnetic region. The observed CEF levels are reproduced by point-charge calculations consistent with the $O_h$ symmetry of Dy$^{3+}$. A complementary machine-learning approach is used to analyse the phonon spectrum and compare with INS data. Together, these results clarify the origin of phonon and magnon excitations and their role in the ground-state magnetism of Ba$_2$DyRuO$_6$.

cond-mat.str-el↗

arXiv2025

Latent Speech-Text Transformer

Auto-regressive speech-text models pre-trained on interleaved text tokens and discretized speech tokens demonstrate strong speech understanding and generation, yet remain substantially less compute-efficient than text LLMs, partly due to the much longer sequences of speech tokens relative to text. This modality imbalance disproportionately allocates pre-training and inference compute to speech, potentially hindering effective cross-modal alignment and slowing performance scaling by orders of magnitude. We introduce the Latent Speech-Text Transformer (LST), which aggregates speech tokens into latent speech patches that serve as higher-level autoregressive units. This design aligns the sequence-modeling granularity between speech and text while improving computational efficiency. The resulting patches can align with textual units to facilitate cross-modal knowledge transfer and compactly capture recurring acoustic patterns such as silence. Across story-completion benchmarks under both compute-controlled and data-controlled settings, LST consistently improves speech accuracy while also improving text performance, achieving up to +6.5% absolute gain on speech HellaSwag in compute-controlled training (+5.3% in data-controlled training). Under compute-controlled scaling from 420M to 1.8B parameters in a near compute-optimal regime, gains grow with scale, and improvements persist up to 7B parameters under fixed-token budgets. These benefits extend to downstream tasks: LST stabilizes ASR adaptation and reduces the effective autoregressive sequence length during ASR and TTS inference, lowering computational cost without degrading reconstruction quality. The code is available at https://github.com/facebookresearch/lst.

cs.CL↗

arXiv2025

Physics-informed Ground Reaction Dynamics from Human Motion Capture

Body dynamics are crucial information for the analysis of human motions in important research fields, ranging from biomechanics, sports science to computer vision and graphics. Modern approaches collect the body dynamics, external reactive force specifically, via force plates, synchronizing with human motion capture data, and learn to estimate the dynamics from a black-box deep learning model. Being specialized devices, force plates can only be installed in laboratory setups, imposing a significant limitation on the learning of human dynamics. To this end, we propose a novel method for estimating human ground reaction dynamics directly from the more reliable motion capture data with physics laws and computational simulation as constrains. We introduce a highly accurate and robust method for computing ground reaction forces from motion capture data using Euler's integration scheme and PD algorithm. The physics-based reactive forces are used to inform the learning model about the physics-informed motion dynamics thus improving the estimation accuracy. The proposed approach was tested on the GroundLink dataset, outperforming the baseline model on: 1) the ground reaction force estimation accuracy compared to the force plates measurement; and 2) our simulated root trajectory precision. The implementation code is available at https://github.com/cuongle1206/Phys-GRD

cs.CV↗

arXiv2025

Blended parameterization in an atmospheric model: Improving severestorm ensemble prediction by considering uncertainties in model physics

Physics parameterizations are often needed for numerical weather prediction (NWP) of precipitation forecast. This is mainly because the resolutions of most computational atmospheric models are not fine enough to explicitly resolve sub-grid scale processes associated with precipitation systems. The various options in each physical parameterization scheme introduce model physics uncertainty, leading to variations in simulated precipitation due to differing representations of physical processes. We aim to quantify and reduce uncertainties in severe storm prediction arising from selecting physics parameterization schemes. In this study, we introduced a method called "blended parameterization" in an atmospheric model. This method parameterizes the selection of physical parameterization schemes using weighting parameters. This approach reduces the model selection problem to the optimization of these weighting parameters. The Markov Chain Monte Carlo (MCMC) method was used to estimate the posterior probability distribution of the weighting parameters. The large computational cost of MCMC was resolved by a surrogate model that efficiently mimics the relationship between weighting parameters and likelihood. Subsequently, the weighting parameters were sampled from the posterior distribution, followed by conducting an ensemble simulation to predict rainfall in Vietnam. Our optimized "blended parameterized" ensemble prediction system outperformed a conventional physical ensemble prediction system similar to that operationally used in Vietnam.

physics.geo-ph↗

arXiv2025

From Visual Explanations to Counterfactual Explanations with Latent Diffusion

Visual counterfactual explanations are ideal hypothetical images that change the decision-making of the classifier with high confidence toward the desired class while remaining visually plausible and close to the initial image. In this paper, we propose a new approach to tackle two key challenges in recent prominent works: i) determining which specific counterfactual features are crucial for distinguishing the "concept" of the target class from the original class, and ii) supplying valuable explanations for the non-robust classifier without relying on the support of an adversarially robust model. Our method identifies the essential region for modification through algorithms that provide visual explanations, and then our framework generates realistic counterfactual explanations by combining adversarial attacks based on pruning the adversarial gradient of the target classifier and the latent diffusion model. The proposed method outperforms previous state-of-the-art results on various evaluation criteria on ImageNet and CelebA-HQ datasets. In general, our method can be applied to arbitrary classifiers, highlight the strong association between visual and counterfactual explanations, make semantically meaningful changes from the target classifier, and provide observers with subtle counterfactual images.

cs.CV↗

arXiv2024

Textless Streaming Speech-to-Speech Translation using Semantic Speech Tokens

Cascaded speech-to-speech translation systems often suffer from the error accumulation problem and high latency, which is a result of cascaded modules whose inference delays accumulate. In this paper, we propose a transducer-based speech translation model that outputs discrete speech tokens in a low-latency streaming fashion. This approach eliminates the need for generating text output first, followed by machine translation (MT) and text-to-speech (TTS) systems. The produced speech tokens can be directly used to generate a speech signal with low latency by utilizing an acoustic language model (LM) to obtain acoustic tokens and an audio codec model to retrieve the waveform. Experimental results show that the proposed method outperforms other existing approaches and achieves state-of-the-art results for streaming translation in terms of BLEU, average latency, and BLASER 2.0 scores for multiple language pairs using the CVSS-C dataset as a benchmark.

eess.AS↗

arXiv2024

Seed-Music: A Unified Framework for High Quality and Controlled Music Generation

We introduce Seed-Music, a suite of music generation systems capable of producing high-quality music with fine-grained style control. Our unified framework leverages both auto-regressive language modeling and diffusion approaches to support two key music creation workflows: controlled music generation and post-production editing. For controlled music generation, our system enables vocal music generation with performance controls from multi-modal inputs, including style descriptions, audio references, musical scores, and voice prompts. For post-production editing, it offers interactive tools for editing lyrics and vocal melodies directly in the generated audio. We encourage readers to listen to demo audio examples at https://team.doubao.com/seed-music "https://team.doubao.com/seed-music".

cs.SD↗

arXiv2024

PRoDeliberation: Parallel Robust Deliberation for End-to-End Spoken Language Understanding

Spoken Language Understanding (SLU) is a critical component of voice assistants; it consists of converting speech to semantic parses for task execution. Previous works have explored end-to-end models to improve the quality and robustness of SLU models with Deliberation, however these models have remained autoregressive, resulting in higher latencies. In this work we introduce PRoDeliberation, a novel method leveraging a Connectionist Temporal Classification-based decoding strategy as well as a denoising objective to train robust non-autoregressive deliberation models. We show that PRoDeliberation achieves the latency reduction of parallel decoding (2-10x improvement over autoregressive models) while retaining the ability to correct Automatic Speech Recognition (ASR) mistranscriptions of autoregressive deliberation systems. We further show that the design of the denoising training allows PRoDeliberation to overcome the limitations of small ASR devices, and we provide analysis on the necessity of each component of the system.

cs.CL↗

arXiv2023

Seq2seq for Automatic Paraphasia Detection in Aphasic Speech

Paraphasias are speech errors that are often characteristic of aphasia and they represent an important signal in assessing disease severity and subtype. Traditionally, clinicians manually identify paraphasias by transcribing and analyzing speech-language samples, which can be a time-consuming and burdensome process. Identifying paraphasias automatically can greatly help clinicians with the transcription process and ultimately facilitate more efficient and consistent aphasia assessment. Previous research has demonstrated the feasibility of automatic paraphasia detection by training an automatic speech recognition (ASR) model to extract transcripts and then training a separate paraphasia detection model on a set of hand-engineered features. In this paper, we propose a novel, sequence-to-sequence (seq2seq) model that is trained end-to-end (E2E) to perform both ASR and paraphasia detection tasks. We show that the proposed model outperforms the previous state-of-the-art approach for both word-level and utterance-level paraphasia detection tasks and provide additional follow-up evaluations to further understand the proposed model behavior.

cs.SD↗

arXiv2023

StemGen: A music generation model that listens

End-to-end generation of musical audio using deep learning techniques has seen an explosion of activity recently. However, most models concentrate on generating fully mixed music in response to abstract conditioning information. In this work, we present an alternative paradigm for producing music generation models that can listen and respond to musical context. We describe how such a model can be constructed using a non-autoregressive, transformer-based model architecture and present a number of novel architectural and sampling improvements. We train the described architecture on both an open-source and a proprietary dataset. We evaluate the produced models using standard quality metrics and a new approach based on music information retrieval descriptors. The resulting model reaches the audio quality of state-of-the-art text-conditioned models, as well as exhibiting strong musical coherence with its context.

cs.SD↗

arXiv2023

A Foundation Model for Music Informatics

This paper investigates foundation models tailored for music informatics, a domain currently challenged by the scarcity of labeled data and generalization issues. To this end, we conduct an in-depth comparative study among various foundation model variants, examining key determinants such as model architectures, tokenization methods, temporal resolution, data, and model scalability. This research aims to bridge the existing knowledge gap by elucidating how these individual factors contribute to the success of foundation models in music informatics. Employing a careful evaluation framework, we assess the performance of these models across diverse downstream tasks in music information retrieval, with a particular focus on token-level and sequence-level classification. Our results reveal that our model demonstrates robust performance, surpassing existing models in specific key metrics. These findings contribute to the understanding of self-supervised learning in music informatics and pave the way for developing more effective and versatile foundation models in the field. A pretrained version of our model is publicly available to foster reproducibility and future research.

cs.SD↗

arXiv2023

Scaling Up Music Information Retrieval Training with Semi-Supervised Learning

In the era of data-driven Music Information Retrieval (MIR), the scarcity of labeled data has been one of the major concerns to the success of an MIR task. In this work, we leverage the semi-supervised teacher-student training approach to improve MIR tasks. For training, we scale up the unlabeled music data to 240k hours, which is much larger than any public MIR datasets. We iteratively create and refine the pseudo-labels in the noisy teacher-student training process. Knowledge expansion is also explored to iteratively scale up the model sizes from as small as less than 3M to almost 100M parameters. We study the performance correlation between data size and model size in the experiments. By scaling up both model size and training data, our models achieve state-of-the-art results on several MIR tasks compared to models that are either trained in a supervised manner or based on a self-supervised pretrained model. To our knowledge, this is the first attempt to study the effects of scaling up both model and training data for a variety of MIR tasks.

eess.AS↗

arXiv2023

Modality Confidence Aware Training for Robust End-to-End Spoken Language Understanding

End-to-end (E2E) spoken language understanding (SLU) systems that generate a semantic parse from speech have become more promising recently. This approach uses a single model that utilizes audio and text representations from pre-trained speech recognition models (ASR), and outperforms traditional pipeline SLU systems in on-device streaming scenarios. However, E2E SLU systems still show weakness when text representation quality is low due to ASR transcription errors. To overcome this issue, we propose a novel E2E SLU system that enhances robustness to ASR errors by fusing audio and text representations based on the estimated modality confidence of ASR hypotheses. We introduce two novel techniques: 1) an effective method to encode the quality of ASR hypotheses and 2) an effective approach to integrate them into E2E SLU models. We show accuracy improvements on STOP dataset and share the analysis to demonstrate the effectiveness of our approach.

cs.CL↗

arXiv2023

Text Generation with Speech Synthesis for ASR Data Augmentation

Aiming at reducing the reliance on expensive human annotations, data synthesis for Automatic Speech Recognition (ASR) has remained an active area of research. While prior work mainly focuses on synthetic speech generation for ASR data augmentation, its combination with text generation methods is considerably less explored. In this work, we explore text augmentation for ASR using large-scale pre-trained neural networks, and systematically compare those to traditional text augmentation methods. The generated synthetic texts are then converted to synthetic speech using a text-to-speech (TTS) system and added to the ASR training data. In experiments conducted on three datasets, we find that neural models achieve 9%-15% relative WER improvement and outperform traditional methods. We conclude that text augmentation, particularly through modern neural approaches, is a viable tool for improving the accuracy of ASR systems.

cs.CL↗

arXiv2022

Improving Fast-slow Encoder based Transducer with Streaming Deliberation

This paper introduces a fast-slow encoder based transducer with streaming deliberation for end-to-end automatic speech recognition. We aim to improve the recognition accuracy of the fast-slow encoder based transducer while keeping its latency low by integrating a streaming deliberation model. Specifically, the deliberation model leverages partial hypotheses from the streaming fast encoder and implicitly learns to correct recognition errors. We modify the parallel beam search algorithm for fast-slow encoder based transducer to be efficient and compatible with the deliberation model. In addition, the deliberation model is designed to process streaming data. To further improve the deliberation performance, a simple text augmentation approach is explored. We also compare LSTM and Conformer models for encoding partial hypotheses. Experiments on Librispeech and in-house data show relative WER reductions (WERRs) from 3% to 5% with a slight increase in model size and negligible extra token emission latency compared with fast-slow encoder based transducer. Compared with vanilla neural transducers, the proposed deliberation model together with fast-slow encoder based transducer obtains relative 10-11% WERRs on Librispeech and around relative 6% WERR on in-house data with smaller emission delays.

eess.AS↗

arXiv2022

Massively Multilingual ASR on 70 Languages: Tokenization, Architecture, and Generalization Capabilities

End-to-end multilingual ASR has become more appealing because of several reasons such as simplifying the training and deployment process and positive performance transfer from high-resource to low-resource languages. However, scaling up the number of languages, total hours, and number of unique tokens is not a trivial task. This paper explores large-scale multilingual ASR models on 70 languages. We inspect two architectures: (1) Shared embedding and output and (2) Multiple embedding and output model. In the shared model experiments, we show the importance of tokenization strategy across different languages. Later, we use our optimal tokenization strategy to train multiple embedding and output model to further improve our result. Our multilingual ASR achieves 13.9%-15.6% average WER relative improvement compared to monolingual models. We show that our multilingual ASR generalizes well on an unseen dataset and domain, achieving 9.5% and 7.5% WER on Multilingual Librispeech (MLS) with zero-shot and finetuning, respectively.

cs.CL↗

arXiv2022

Factorized Blank Thresholding for Improved Runtime Efficiency of Neural Transducers

We show how factoring the RNN-T's output distribution can significantly reduce the computation cost and power consumption for on-device ASR inference with no loss in accuracy. With the rise in popularity of neural-transducer type models like the RNN-T for on-device ASR, optimizing RNN-T's runtime efficiency is of great interest. While previous work has primarily focused on the optimization of RNN-T's acoustic encoder and predictor, this paper focuses the attention on the joiner. We show that despite being only a small part of RNN-T, the joiner has a large impact on the overall model's runtime efficiency. We propose to utilize HAT-style joiner factorization for the purpose of skipping the more expensive non-blank computation when the blank probability exceeds a certain threshold. Since the blank probability can be computed very efficiently and the RNN-T output is dominated by blanks, our proposed method leads to a 26-30% decoding speed-up and 43-53% reduction in on-device power consumption, all the while incurring no accuracy degradation and being relatively simple to implement.

eess.AS↗