Search

In recent years, we have observed a rapid advancement in speech language models (SpeechLLMs), catching up with humans' listening and reasoning abilities. Remarkably, SpeechLLMs have demonstrated impressive spoken dialogue question-answering (SQA) performance in benchmarks like Gaokao, the English listening test of the college entrance exam in China, which seemingly requires understanding both the spoken content and voice characteristics of speakers in a conversation. However, after carefully examining Gaokao's questions, we find the correct answers to many questions can be inferred from the conversation context alone without identifying the speaker asked in the question. Our evaluation of state-of-the-art models Qwen-Audio and WavLLM in both Gaokao and our proposed "What Do You Like?" dataset shows a significantly higher accuracy in these context-based questions than in identity-critical questions, which can only be answered correctly with correct speaker identification. Our results and analysis suggest that when solving SQA, the current SpeechLLMs exhibit limited speaker awareness from the audio and behave similarly to an LLM reasoning from the conversation transcription without sound. We propose that our definitions and automated classification of context-based and identity-critical questions could offer a more accurate evaluation framework of SpeechLLMs in SQA tasks., Comment: Accepted to IEEE SLT 2024

This work reimplements a recent semantic bootstrapping child-language acquisition model, which was originally designed for English, and trains it to learn a new language: Hebrew. The model learns from pairs of utterances and logical forms as meaning representations, and acquires both syntax and word meanings simultaneously. The results show that the model mostly transfers to Hebrew, but that a number of factors, including the richer morphology in Hebrew, makes the learning slower and less robust. This suggests that a clear direction for future work is to enable the model to leverage the similarities between different word forms.

Test-Time Adaptation (TTA) has emerged as a crucial solution to the domain shift challenge, wherein the target environment diverges from the original training environment. A prime exemplification is TTA for Automatic Speech Recognition (ASR), which enhances model performance by leveraging output prediction entropy minimization as a self-supervision signal. However, a key limitation of this self-supervision lies in its primary focus on acoustic features, with minimal attention to the linguistic properties of the input. To address this gap, we propose Language Informed Test-Time Adaptation (LI-TTA), which incorporates linguistic insights during TTA for ASR. LI-TTA integrates corrections from an external language model to merge linguistic with acoustic information by minimizing the CTC loss from the correction alongside the standard TTA loss. With extensive experiments, we show that LI-TTA effectively improves the performance of TTA for ASR in various distribution shift situations., Comment: INTERSPEECH 2024

Reinforcement Learning from Human Feedback (RLHF) leverages human preference data to train language models to align more closely with human essence. These human preference data, however, are labeled at the sequence level, creating a mismatch between sequence-level preference labels and tokens, which are autoregressively generated from the language model. Although several recent approaches have tried to provide token-level (i.e., dense) rewards for each individual token, these typically rely on predefined discrete reward values (e.g., positive: +1, negative: -1, neutral: 0), failing to account for varying degrees of preference inherent to each token. To address this limitation, we introduce TLCR (Token-Level Continuous Reward) for RLHF, which incorporates a discriminator trained to distinguish positive and negative tokens, and the confidence of the discriminator is used to assign continuous rewards to each token considering the context. Extensive experiments show that our proposed TLCR leads to consistent performance improvements over previous sequence-level or token-level discrete rewards on open-ended generation benchmarks., Comment: ACL2024 Findings

Certain environmental noises have been associated with negative developmental outcomes for infants and young children. Though classifying or tagging sound events in a domestic environment is an active research area, previous studies focused on data collected from a non-stationary microphone placed in the environment or from the perspective of adults. Further, many of these works ignore infants or young children in the environment or have data collected from only a single family where noise from the fixed sound source can be moderate at the infant's position or vice versa. Thus, despite the recent success of large pre-trained models for noise event detection, the performance of these models on infant-centric noise soundscapes in the home is yet to be explored. To bridge this gap, we have collected and labeled noises in home soundscapes from 22 families in an unobtrusive manner, where the data are collected through an infant-worn recording device. In this paper, we explore the performance of a large pre-trained model (Audio Spectrogram Transformer [AST]) on our noise-conditioned infant-centric environmental data as well as publicly available home environmental datasets. Utilizing different training strategies such as resampling, utilizing public datasets, mixing public and infant-centric training sets, and data augmentation using noise and masking, we evaluate the performance of a large pre-trained model on sparse and imbalanced infant-centric data. Our results show that fine-tuning the large pre-trained model by combining our collected dataset with public datasets increases the F1-score from 0.11 (public datasets) and 0.76 (collected datasets) to 0.84 (combined datasets) and Cohen's Kappa from 0.013 (public datasets) and 0.77 (collected datasets) to 0.83 (combined datasets) compared to only training with public or collected datasets, respectively., Comment: Accepted in IEEE/ACM CHASE 2024

Recent advancements in supervised automatic speech recognition (ASR) have achieved remarkable performance, largely due to the growing availability of large transcribed speech corpora. However, most languages lack sufficient paired speech and text data to effectively train these systems. In this article, we tackle the challenge of developing ASR systems without paired speech and text corpora by proposing the removal of reliance on a phoneme lexicon. We explore a new research direction: word-level unsupervised ASR. Using a curated speech corpus containing only high-frequency English words, our system achieves a word error rate of nearly 20% without parallel transcripts or oracle word boundaries. Furthermore, we experimentally demonstrate that an unsupervised speech recognizer can emerge from joint speech-to-speech and text-to-text masked token-infilling. This innovative model surpasses the performance of previous unsupervised ASR models trained with direct distribution matching., Comment: This work has been submitted to the IEEE for possible publication

In deep learning, test-time adaptation has gained attention as a method for model fine-tuning without the need for labeled data. A prime exemplification is the recently proposed test-time prompt tuning for large-scale vision-language models such as CLIP. Unfortunately, these prompts have been mainly developed to improve accuracy, overlooking the importance of calibration, which is a crucial aspect for quantifying prediction uncertainty. However, traditional calibration methods rely on substantial amounts of labeled data, making them impractical for test-time scenarios. To this end, this paper explores calibration during test-time prompt tuning by leveraging the inherent properties of CLIP. Through a series of observations, we find that the prompt choice significantly affects the calibration in CLIP, where the prompts leading to higher text feature dispersion result in better-calibrated predictions. Introducing the Average Text Feature Dispersion (ATFD), we establish its relationship with calibration error and present a novel method, Calibrated Test-time Prompt Tuning (C-TPT), for optimizing prompts during test-time with enhanced calibration. Through extensive experiments on different CLIP architectures and datasets, we show that C-TPT can effectively improve the calibration of test-time prompt tuning without needing labeled data. The code is publicly accessible at https://github.com/hee-suk-yoon/C-TPT., Comment: ICLR 2024

In Automatic Speech Recognition (ASR) systems, a recurring obstacle is the generation of narrowly focused output distributions. This phenomenon emerges as a side effect of Connectionist Temporal Classification (CTC), a robust sequence learning tool that utilizes dynamic programming for sequence mapping. While earlier efforts have tried to combine the CTC loss with an entropy maximization regularization term to mitigate this issue, they employed a constant weighting term on the regularization during the training, which we find may not be optimal. In this work, we introduce Adaptive Maximum Entropy Regularization (AdaMER), a technique that can modulate the impact of entropy regularization throughout the training process. This approach not only refines ASR model training but ensures that as training proceeds, predictions display the desired model confidence.

Quantum computers hold the promise of solving certain problems that lie beyond the reach of conventional computers. Establishing this capability, especially for impactful and meaningful problems, remains a central challenge. One such problem is the simulation of nonequilibrium dynamics of a magnetic spin system quenched through a quantum phase transition. State-of-the-art classical simulations demand resources that grow exponentially with system size. Here we show that superconducting quantum annealing processors can rapidly generate samples in close agreement with solutions of the Schr\"odinger equation. We demonstrate area-law scaling of entanglement in the model quench in two-, three- and infinite-dimensional spin glasses, supporting the observed stretched-exponential scaling of effort for classical approaches. We assess approximate methods based on tensor networks and neural networks and conclude that no known approach can achieve the same accuracy as the quantum annealer within a reasonable timeframe. Thus quantum annealers can answer questions of practical importance that classical computers cannot.

To understand why self-supervised learning (SSL) models have empirically achieved strong performances on several speech-processing downstream tasks, numerous studies have focused on analyzing the encoded information of the SSL layer representations in adult speech. Limited work has investigated how pre-training and fine-tuning affect SSL models encoding children's speech and vocalizations. In this study, we aim to bridge this gap by probing SSL models on two relevant downstream tasks: (1) phoneme recognition (PR) on the speech of adults, older children (8-10 years old), and younger children (1-4 years old), and (2) vocalization classification (VC) distinguishing cry, fuss, and babble for infants under 14 months old. For younger children's PR, the superiority of fine-tuned SSL models is largely due to their ability to learn features that represent older children's speech and then adapt those features to the speech of younger children. For infant VC, SSL models pre-trained on large-scale home recordings learn to leverage phonetic representations at middle layers, and thereby enhance the performance of this task., Comment: Accepted to 2024 ICASSP Workshop of Self-supervision in Audio, Speech and Beyond (SASB)

This paper explores advancements in high-fidelity personalized image generation through the utilization of pre-trained text-to-image diffusion models. While previous approaches have made significant strides in generating versatile scenes based on text descriptions and a few input images, challenges persist in maintaining the subject fidelity within the generated images. In this work, we introduce an innovative algorithm named HiFi Tuner to enhance the appearance preservation of objects during personalized image generation. Our proposed method employs a parameter-efficient fine-tuning framework, comprising a denoising process and a pivotal inversion process. Key enhancements include the utilization of mask guidance, a novel parameter regularization technique, and the incorporation of step-wise subject representations to elevate the sample fidelity. Additionally, we propose a reference-guided generation approach that leverages the pivotal inversion of a reference image to mitigate unwanted subject variations and artifacts. We further extend our method to a novel image editing task: substituting the subject in an image through textual manipulations. Experimental evaluations conducted on the DreamBooth dataset using the Stable Diffusion model showcase promising results. Fine-tuning solely on textual embeddings improves CLIP-T score by 3.6 points and improves DINO score by 9.6 points over Textual Inversion. When fine-tuning all parameters, HiFi Tuner improves CLIP-T score by 1.2 points and improves DINO score by 1.2 points over DreamBooth, establishing a new state of the art.

Quantum Error Mitigation (QEM) presents a promising near-term approach to reduce error when estimating expectation values in quantum computing. Here, we introduce QEM techniques tailored for quantum annealing, using Zero-Noise Extrapolation (ZNE). We implement ZNE through zero-temperature extrapolation as well as energy-time rescaling. We conduct experimental investigations into the quantum critical dynamics of a transverse-field Ising spin chain, demonstrating the successful mitigation of thermal noise through both of these techniques. Moreover, we show that energy-time rescaling effectively mitigates control errors in the coherent regime where the effect of thermal noise is minimal. Our ZNE results agree with exact calculations of the coherent evolution over a range of annealing times that exceeds the coherent annealing range by almost an order of magnitude., Comment: 10 pages, 5 figures

Quantum computers are nearing the thousand qubit mark, with the current focus on scaling to improve computational performance. As quantum processors grow in complexity, new challenges arise such as the management of device variability and the interface with supporting electronics. Spin qubits in silicon quantum dots are poised to address these challenges with their proven control fidelities and potential for compatibility with large-scale integration. Here, we demonstrate the integration of 1024 silicon quantum dots with on-chip digital and analogue electronics, all operating below 1 K. A high-frequency analogue multiplexer provides fast access to all devices with minimal electrical connections, enabling characteristic data across the quantum dot array to be acquired in just 5 minutes. We achieve this by leveraging radio-frequency reflectometry with state-of-the-art signal integrity, reaching a minimum integration time of 160 ps. Key quantum dot parameters are extracted by fast automated machine learning routines to assess quantum dot yield and understand the impact of device design. We find correlations between quantum dot parameters and room temperature transistor behaviour that may be used as a proxy for in-line process monitoring. Our results show how rapid large-scale studies of silicon quantum devices can be performed at lower temperatures and measurement rates orders of magnitude faster than current probing techniques, and form a platform for the future on-chip addressing of large scale qubit arrays., Comment: Main text: 14 pages, 8 figures, 1 table Supplementary: 8 pages, 6 figures

Training unsupervised speech recognition systems presents challenges due to GAN-associated instability, misalignment between speech and text, and significant memory demands. To tackle these challenges, we introduce a novel ASR system, ESPUM. This system harnesses the power of lower-order N-skipgrams (up to N=3) combined with positional unigram statistics gathered from a small batch of samples. Evaluated on the TIMIT benchmark, our model showcases competitive performance in ASR and phoneme segmentation tasks. Access our publicly available code at https://github.com/lwang114/GraphUnsupASR.

Scalable quantum processors require high-fidelity universal quantum logic operations in a manufacturable physical platform. Donors in silicon provide atomic size, excellent quantum coherence and compatibility with standard semiconductor processing, but no entanglement between donor-bound electron spins has been demonstrated to date. Here we present the experimental demonstration and tomography of universal 1- and 2-qubit gates in a system of two weakly exchange-coupled electrons, bound to single phosphorus donors introduced in silicon by ion implantation. We surprisingly observe that the exchange interaction has no effect on the qubit coherence. We quantify the fidelity of the quantum operations using gate set tomography (GST), and we use the universal gate set to create entangled Bell states of the electrons spins, with fidelity ~ 93%, and concurrence 0.91 +/- 0.08. These results form the necessary basis for scaling up donor-based quantum computers.

The assessment of children at risk of autism typically involves a clinician observing, taking notes, and rating children's behaviors. A machine learning model that can label adult and child audio may largely save labor in coding children's behaviors, helping clinicians capture critical events and better communicate with parents. In this study, we leverage Wav2Vec 2.0 (W2V2), pre-trained on 4300-hour of home audio of children under 5 years old, to build a unified system for tasks of clinician-child speaker diarization and vocalization classification (VC). To enhance children's VC, we build a W2V2 phoneme recognition system for children under 4 years old, and we incorporate its phonetically-tuned embeddings as auxiliary features or recognize pseudo phonetic transcripts as an auxiliary task. We test our method on two corpora (Rapid-ABC and BabbleCor) and obtain consistent improvements. Additionally, we outperform the state-of-the-art performance on the reproducible subset of BabbleCor. Code available at https://huggingface.co/lijialudew, Comment: Accepted to Interspeech 2024

Text-to-Text Transfer Transformer (T5) has recently been considered for the Grapheme-to-Phoneme (G2P) transduction. As a follow-up, a tokenizer-free byte-level model based on T5 referred to as ByT5, recently gave promising results on word-level G2P conversion by representing each input character with its corresponding UTF-8 encoding. Although it is generally understood that sentence-level or paragraph-level G2P can improve usability in real-world applications as it is better suited to perform on heteronyms and linking sounds between words, we find that using ByT5 for these scenarios is nontrivial. Since ByT5 operates on the character level, it requires longer decoding steps, which deteriorates the performance due to the exposure bias commonly observed in auto-regressive generation models. This paper shows that the performance of sentence-level and paragraph-level G2P can be improved by mitigating such exposure bias using our proposed loss-based sampling method., Comment: INTERSPEECH 2023

Quantum nondemolition (QND) measurements are a precious resource for quantum information processing. Repetitive QND measurements can boost the fidelity of qubit preparation and measurement, even when the underlying single-shot measurements are of low fidelity. However, this fidelity boost is limited by the degree in which the physical system allows for a truly QND process -- slight deviations from ideal QND measurement result in bit flip errors (`quantum jumps') if the measurement is repeated too often. Here, we develop a theoretical framework to understand and quantify the resulting error arising from deviation from perfect QND measurement in model spin qubit systems. We first develop our model on the ubiquitous example of exchange-coupled electron spins qubits tunnel-coupled to a charge reservoir. We then extend it to electron-nuclear spin systems, to illustrate the crucial similarities and differences between the two limits. Applied to the well-understood platform of a donor nuclear spin in silicon, the model shows excellent agreement with experiments. For added generality, we conclude the work by considering the effect of anisotropic spin couplings.

Infant sleep is critical to brain and behavioral development. Prior studies on infant sleep/wake classification have been largely limited to reliance on expensive and burdensome polysomnography (PSG) tests in the laboratory or wearable devices that collect single-modality data. To facilitate data collection and accuracy of detection, we aimed to advance this field of study by using a multi-modal wearable device, LittleBeats (LB), to collect audio, electrocardiogram (ECG), and inertial measurement unit (IMU) data among a cohort of 28 infants. We employed a 3-branch (audio/ECG/IMU) large scale transformer-based neural network (NN) to demonstrate the potential of such multi-modal data. We pretrained each branch independently with its respective modality, then finetuned the model by fusing the pretrained transformer layers with cross-attention. We show that multi-modal data significantly improves sleep/wake classification (accuracy = 0.880), compared with use of a single modality (accuracy = 0.732). Our approach to multi-modal mid-level fusion may be adaptable to a diverse range of architectures and tasks, expanding future directions of infant behavioral research., Comment: Preprint for APSIPA2023

Efficient scaling and flexible control are key aspects of useful quantum computing hardware. Spins in semiconductors combine quantum information processing with electrons, holes or nuclei, control with electric or magnetic fields, and scalable coupling via exchange or dipole interaction. However, accessing large Hilbert space dimensions has remained challenging, due to the short-distance nature of the interactions. Here, we present an atom-based semiconductor platform where a 16-dimensional Hilbert space is built by the combined electron-nuclear states of a single antimony donor in silicon. We demonstrate the ability to navigate this large Hilbert space using both electric and magnetic fields, with gate fidelity exceeding 99.8% on the nuclear spin, and unveil fine details of the system Hamiltonian and its susceptibility to control and noise fields. These results establish high-spin donors as a rich platform for practical quantum information and to explore quantum foundations., Comment: 31 pages and 19 figures including Supplementary Materials

Unsupervised speech recognition (ASR-U) is the problem of learning automatic speech recognition (ASR) systems from unpaired speech-only and text-only corpora. While various algorithms exist to solve this problem, a theoretical framework is missing from studying their properties and addressing such issues as sensitivity to hyperparameters and training instability. In this paper, we proposed a general theoretical framework to study the properties of ASR-U systems based on random matrix theory and the theory of neural tangent kernels. Such a framework allows us to prove various learnability conditions and sample complexity bounds of ASR-U. Extensive ASR-U experiments on synthetic languages with three classes of transition graphs provide strong empirical evidence for our theory (code available at cactuswiththoughts/UnsupASRTheory.git).

Automatic Speech Recognition (ASR) systems have attained unprecedented performance with large speech models pre-trained based on self-supervised speech representation learning. However, these pre-trained speech models suffer from representational bias as they tend to better represent those prominent accents (i.e., native (L1) English accent) in the pre-training speech corpus than less represented accents, resulting in a deteriorated performance for non-native (L2) English accents. Although there have been some approaches to mitigate this issue, all of these methods require updating the pre-trained model weights. In this paper, we propose Information Theoretic Adversarial Prompt Tuning (INTapt), which introduces prompts concatenated to the original input that can re-modulate the attention of the pre-trained model such that the corresponding input resembles a native (L1) English speech without updating the backbone weights. INTapt is trained simultaneously in the following two manners: (1) adversarial training to reduce accent feature dependence between the original input and the prompt-concatenated input and (2) training to minimize CTC loss for improving ASR performance to a prompt-concatenated input. Experimental results show that INTapt improves the performance of L2 English and increases feature similarity between L2 and L1 accents., Comment: ACL2023

Large Language Models (LLMs) are claimed to be capable of Natural Language Inference (NLI), necessary for applied tasks like question answering and summarization. We present a series of behavioral studies on several LLM families (LLaMA, GPT-3.5, and PaLM) which probe their behavior using controlled experiments. We establish two biases originating from pretraining which predict much of their behavior, and show that these are major sources of hallucination in generative LLMs. First, memorization at the level of sentences: we show that, regardless of the premise, models falsely label NLI test samples as entailing when the hypothesis is attested in training data, and that entities are used as ``indices'' to access the memorized data. Second, statistical patterns of usage learned at the level of corpora: we further show a similar effect when the premise predicate is less frequent than that of the hypothesis in the training data, a bias following from previous studies. We demonstrate that LLMs perform significantly worse on NLI test samples which do not conform to these biases than those which do, and we offer these as valuable controls for future LLM evaluation., Comment: Findings of EMNLP 2023

To perform automatic family audio analysis, past studies have collected recordings using phone, video, or audio-only recording devices like LENA, investigated supervised learning methods, and used or fine-tuned general-purpose embeddings learned from large pretrained models. In this study, we advance the audio component of a new infant wearable multi-modal device called LittleBeats (LB) by learning family audio representation via wav2vec 2.0 (W2V2) pertaining. We show given a limited number of labeled LB home recordings, W2V2 pretrained using 1k-hour of unlabeled home recordings outperforms oracle W2V2 pretrained on 960-hour unlabeled LibriSpeech in terms of parent/infant speaker diarization (SD) and vocalization classifications (VC) at home. Extra relevant external unlabeled and labeled data further benefit W2V2 pretraining and fine-tuning. With SpecAug and environmental speech corruptions, we obtain 12% relative gain on SD and moderate boost on VC. Code and model weights are available., Comment: Proceedings of Interspeech 2023; v4 version updates: correction of W2V2-base pretrained on 960-hour of LibriSpeech and number of families participated for LENA home recordings

Word Sense Disambiguation (WSD) is an NLP task aimed at determining the correct sense of a word in a sentence from discrete sense choices. Although current systems have attained unprecedented performances for such tasks, the nonuniform distribution of word senses during training generally results in systems performing poorly on rare senses. To this end, we consider data augmentation to increase the frequency of these least frequent senses (LFS) to reduce the distributional bias of senses during training. We propose Sense-Maintained Sentence Mixup (SMSMix), a novel word-level mixup method that maintains the sense of a target word. SMSMix smoothly blends two sentences using mask prediction while preserving the relevant span determined by saliency scores to maintain a specific word's sense. To the best of our knowledge, this is the first attempt to apply mixup in NLP while preserving the meaning of a specific word. With extensive experiments, we validate that our augmentation method can effectively give more information about rare senses during training with maintained target sense label., Comment: EMNLP2022

Variability in sea ice conditions, combined with strong couplings to the atmosphere and the ocean, lead to a broad range of complex sea ice dynamics. More in-situ measurements are needed to better identify the phenomena and mechanisms that govern sea ice growth, drift, and breakup. To this end, we have gathered a dataset of in-situ observations of sea ice drift and waves in ice. A total of 15 deployments were performed over a period of 5 years in both the Arctic and Antarctic, involving 72 instruments. These provide both GPS drift tracks, and measurements of waves in ice. The data can, in turn, be used for tuning sea ice drift models, investigating waves damping by sea ice, and helping calibrate other sea ice measurement techniques, such as satellite based observations.