arXiv subjects

Tao Xu

Publications and source records attributed to Tao Xu.

At least 55 records · Page 3Linked to original sources

arXiv2025

Quantum-Boosted High-Fidelity Deep Learning

A fundamental limitation of probabilistic deep learning is its predominant reliance on Gaussian priors. This simplistic assumption prevents models from accurately capturing the complex, non-Gaussian landscapes of natural data, particularly in demanding domains like complex biological data, severely hindering the fidelity of the model for scientific discovery. The physically-grounded Boltzmann distribution offers a more expressive alternative, but it is computationally intractable on classical computers. To date, quantum approaches have been hampered by the insufficient qubit scale and operational stability required for the iterative demands of deep learning. Here, we bridge this gap by introducing the Quantum Boltzmann Machine-Variational Autoencoder (QBM-VAE), a large-scale and long-time stable hybrid quantum-classical architecture. Our framework leverages a quantum processor for efficient sampling from the Boltzmann distribution, enabling its use as a powerful prior within a deep generative model. Applied to million-scale single-cell datasets from multiple sources, the QBM-VAE generates a latent space that better preserves complex biological structures, consistently outperforming conventional Gaussian-based deep learning models like VAE and SCVI in essential tasks such as omics data integration, cell-type classification, and trajectory inference. It also provides a typical example of introducing a physics priori into deep learning to drive the model to acquire scientific discovery capabilities that breaks through data limitations. This work provides the demonstration of a practical quantum advantage in deep learning on a large-scale scientific problem and offers a transferable blueprint for developing hybrid quantum AI models.

cs.LG↗

arXiv2025

Detecting Axion-Like Particles with Primordial Black Holes

Future gamma-ray experiments, such as the e-ASTROGAM and AMEGO telescopes, can detect the Hawking radiation of photons from primordial black holes (PBHs) if they make up a fraction or all of dark matter. PBHs can analogously also Hawking radiate new particles, which is especially interesting if these particles are mostly secluded from the Standard Model (SM) sector, since they might therefore be less accessible otherwise. A well-motivated example of this type is axion-like particles (ALPs) with a tiny coupling to photons. We assume that the ALPs produced by PBHs decay into photons well before reaching the earth, so these will augment the photons directly radiated by the PBHs. Remarkably, we find that the peaks in the energy distributions of ALPs produced from PBHs are different than the corresponding ones for Hawking radiated photons due to the spin-dependent greybody factor. Therefore, we demonstrate that this process will in fact distinctively modify the PBHs' gamma-ray spectrum relative to the SM prediction. We use monochromatic asteroid-mass PBHs as an example to show that e-ASTROGAM can observe the PBH-produced ALP gamma-ray signal (for masses up to ~60 MeV) and further distinguish it from Hawking radiation without ALPs. By measuring the gamma-ray signals, e-ASTROGAM can thereby probe yet unexplored parameters in the ALP mass and photon coupling.

hep-ph↗

arXiv2025

Reinforcement Learning in hyperbolic space for multi-step reasoning

Multi-step reasoning is a fundamental challenge in artificial intelligence, with applications ranging from mathematical problem-solving to decision-making in dynamic environments. Reinforcement Learning (RL) has shown promise in enabling agents to perform multi-step reasoning by optimizing long-term rewards. However, conventional RL methods struggle with complex reasoning tasks due to issues such as credit assignment, high-dimensional state representations, and stability concerns. Recent advancements in Transformer architectures and hyperbolic geometry have provided novel solutions to these challenges. This paper introduces a new framework that integrates hyperbolic Transformers into RL for multi-step reasoning. The proposed approach leverages hyperbolic embeddings to model hierarchical structures effectively. We present theoretical insights, algorithmic details, and experimental results that include Frontier Math and nonlinear optimal control problems. Compared to RL with vanilla transformer, the hyperbolic RL largely improves accuracy by (32%~44%) on FrontierMath benchmark, (43%~45%) on nonlinear optimal control benchmark, while achieving impressive reduction in computational time by (16%~32%) on FrontierMath benchmark, (16%~17%) on nonlinear optimal control benchmark. Our work demonstrates the potential of hyperbolic Transformers in reinforcement learning, particularly for multi-step reasoning tasks that involve hierarchical structures.

cs.LG↗

arXiv2025

Inverse scattering transform for the defocusing nonlinear Schrödinger equation with local and nonlocal nonlinearities under non-zero boundary conditions

Within the framework of the Riemann-Hilbert problem, the theory of inverse scattering transform is established for the defocusing nonlinear Schrödinger equation with local and nonlocal nonlinearities (which originates from the parity-symmetric reduction of the Manakov system) under non-zero boundary conditions. First, the adjoint Lax pair and auxiliary eigenfunctions are introduced for the direct scattering, and the analyticity, symmetries of eigenfunctions and scattering matrix are studied in detail. Then, the distribution of discrete eigenvalues is examined, and the asymptotic behaviors of the eigenfunctions and scattering coefficients are analyzed rigorously. Compared with the Manakov system, the reverse-space nonlocality introduces an additional symmetry, leading to stricter constraints on eigenfunctions, scattering coefficients and norming constants. Further, the Riemann-Hilbert problem is formulated for the the inverse problem with the scattering coefficients admitting an arbitrary number of simple zeros. For the reflectionless case, the N-soliton solutions are presented in the determinant form. With N=1, the dark and beating one-soliton solutions are obtained, which are respectively associated with a pair of discrete eigenvalues lying on and off the circle on the spectrum plane. Via the asymptotic analysis, the two-soliton solutions are found to admit the interactions between two dark solitons or two beating solitons, as well as the superpositions of two beating solitons or one beating soliton and one dark soliton.

nlin.SI↗

arXiv2025

Astra: Toward General-Purpose Mobile Robots via Hierarchical Multimodal Learning

Modern robot navigation systems encounter difficulties in diverse and complex indoor environments. Traditional approaches rely on multiple modules with small models or rule-based systems and thus lack adaptability to new environments. To address this, we developed Astra, a comprehensive dual-model architecture, Astra-Global and Astra-Local, for mobile robot navigation. Astra-Global, a multimodal LLM, processes vision and language inputs to perform self and goal localization using a hybrid topological-semantic graph as the global map, and outperforms traditional visual place recognition methods. Astra-Local, a multitask network, handles local path planning and odometry estimation. Its 4D spatial-temporal encoder, trained through self-supervised learning, generates robust 4D features for downstream tasks. The planning head utilizes flow matching and a novel masked ESDF loss to minimize collision risks for generating local trajectories, and the odometry head integrates multi-sensor inputs via a transformer encoder to predict the relative pose of the robot. Deployed on real in-house mobile robots, Astra achieves high end-to-end mission success rate across diverse indoor environments.

cs.RO↗

arXiv2025

Gravitational wave signatures of primordial black hole accretion during early matter domination

We present a scenario in which primordial black holes (PBHs) form in a post-inflationary radiation-dominated (RD) phase and then experience significant accretion during a phase of early matter dominated (EMD). We show that PBH masses could grow by up to two orders of magnitude. Restricting to the linear perturbation regime, we compute the gravitational wave (GW) spectrum that features two peaks. The high-frequency peak is associated with the PBH formation in the RD phase, while the low-frequency peak is due to the sudden transition from EMD to the later, standard RD phase. We identify a PBH mass range where one or both peaks can be observed by a combination of different GW detectors. Finally, we show the signal-to-noise ratio of the total GW spectrum for PBHs in the asteroid mass window, where they could comprise the totality of dark matter.

hep-ph↗

arXiv2025

LinGen: Towards High-Resolution Minute-Length Text-to-Video Generation with Linear Computational Complexity

Text-to-video generation enhances content creation but is highly computationally intensive: The computational cost of Diffusion Transformers (DiTs) scales quadratically in the number of pixels. This makes minute-length video generation extremely expensive, limiting most existing models to generating videos of only 10-20 seconds length. We propose a Linear-complexity text-to-video Generation (LinGen) framework whose cost scales linearly in the number of pixels. For the first time, LinGen enables high-resolution minute-length video generation on a single GPU without compromising quality. It replaces the computationally-dominant and quadratic-complexity block, self-attention, with a linear-complexity block called MATE, which consists of an MA-branch and a TE-branch. The MA-branch targets short-to-long-range correlations, combining a bidirectional Mamba2 block with our token rearrangement method, Rotary Major Scan, and our review tokens developed for long video generation. The TE-branch is a novel TEmporal Swin Attention block that focuses on temporal correlations between adjacent tokens and medium-range tokens. The MATE block addresses the adjacency preservation issue of Mamba and improves the consistency of generated videos significantly. Experimental results show that LinGen outperforms DiT (with a 75.6% win rate) in video quality with up to 15$\times$ (11.5$\times$) FLOPs (latency) reduction. Furthermore, both automatic metrics and human evaluation demonstrate our LinGen-4B yields comparable video quality to state-of-the-art models (with a 50.5%, 52.1%, 49.1% win rate with respect to Gen-3, LumaLabs, and Kling, respectively). This paves the way to hour-length movie generation and real-time interactive video generation. We provide 68s video generation results and more examples in our project website: https://lineargen.github.io/.

cs.CV↗

arXiv2025

Token-Shuffle: Towards High-Resolution Image Generation with Autoregressive Models

Autoregressive (AR) models, long dominant in language generation, are increasingly applied to image synthesis but are often considered less competitive than Diffusion-based models. A primary limitation is the substantial number of image tokens required for AR models, which constrains both training and inference efficiency, as well as image resolution. To address this, we present Token-Shuffle, a novel yet simple method that reduces the number of image tokens in Transformer. Our key insight is the dimensional redundancy of visual vocabularies in Multimodal Large Language Models (MLLMs), where low-dimensional visual codes from visual encoder are directly mapped to high-dimensional language vocabularies. Leveraging this, we consider two key operations: token-shuffle, which merges spatially local tokens along channel dimension to decrease the input token number, and token-unshuffle, which untangles the inferred tokens after Transformer blocks to restore the spatial arrangement for output. Jointly training with textual prompts, our strategy requires no additional pretrained text-encoder and enables MLLMs to support extremely high-resolution image synthesis in a unified next-token prediction way while maintaining efficient training and inference. For the first time, we push the boundary of AR text-to-image generation to a resolution of 2048x2048 with gratifying generation performance. In GenAI-benchmark, our 2.7B model achieves 0.77 overall score on hard prompts, outperforming AR models LlamaGen by 0.18 and diffusion models LDM by 0.15. Exhaustive large-scale human evaluations also demonstrate our prominent image generation ability in terms of text-alignment, visual flaw, and visual appearance. We hope that Token-Shuffle can serve as a foundational design for efficient high-resolution image generation within MLLMs.

cs.CV↗

arXiv2025

Blue Loops, Cepheids, and Forays into Axions

The blue loop stage of intermediate mass stars has been called a "magnifying glass", where even seemingly small effects in prior stages of evolution, as well as assumptions about stellar composition, rotation, and convection, produce discernible changes. As such, blue loops, and especially the existence and properties of Cepheids, can serve as a laboratory where feebly connected Beyond Standard Model particles such as axions can be gainfully studied. We undertake a careful study of the effects of these putative particles on the blue loop, paying close attention to the evolution of the core potential and the hydrogen profile. Our simulations, performed with MESA, place bounds on the axion-photon coupling using the galactic Cepheid S Mus, with dynamically-determined mass of $6 M_\odot$, as a benchmark. The effects of varying convective overshoot on the core potential and hydrogen profile, and the ensuing changes in the axion constraints, are carefully studied. Along the way, we explore the "mirror principle" induced by the hydrogen burning shell and contrast our results with those existing in the literature. Less conservative (but more stringent) bounds on the axion-photon coupling are given for a $9 M_\odot$ model, which is the heaviest that can be simulated if overshoot is incorporated, and tentative projections are given for a $12 M_\odot$ model, which is approximately the heaviest tail of the mass distribution of galactic Cepheids determined by pulsation models using Gaia DR2. Our main message is that the reliable simulation and observation (ideally, through dynamical mass determination) of massive Cepheids constitutes an important frontier in axion searches, challenges in modeling uncertainties in the microphysics of the blue loop stage notwithstanding.

hep-ph↗

arXiv2025

DRExplainer: Quantifiable Interpretability in Drug Response Prediction with Directed Graph Convolutional Network

Predicting the response of a cancer cell line to a therapeutic drug is pivotal for personalized medicine. Despite numerous deep learning methods that have been developed for drug response prediction, integrating diverse information about biological entities and predicting the directional response remain major challenges. Here, we propose a novel interpretable predictive model, DRExplainer, which leverages a directed graph convolutional network to enhance the prediction in a directed bipartite network framework. DRExplainer constructs a directed bipartite network integrating multi-omics profiles of cell lines, the chemical structure of drugs and known drug response to achieve directed prediction. Then, DRExplainer identifies the most relevant subgraph to each prediction in this directed bipartite network by learning a mask, facilitating critical medical decision-making. Additionally, we introduce a quantifiable method for model interpretability that leverages a ground truth benchmark dataset curated from biological features. In computational experiments, DRExplainer outperforms state-of-the-art predictive methods and another graph-based explanation method under the same experimental setting. Finally, the case studies further validate the interpretability and the effectiveness of DRExplainer in predictive novel drug response. Our code is available at: https://github.com/vshy-dream/DRExplainer.

cs.LG↗

arXiv2025

Adaptive Wavelet Filters as Practical Texture Feature Amplifiers for Parkinson's Disease Screening in OCT

Parkinson's disease (PD) is a prevalent neurodegenerative disorder globally. The eye's retina is an extension of the brain and has great potential in PD screening. Recent studies have suggested that texture features extracted from retinal layers can be adopted as biomarkers for PD diagnosis under optical coherence tomography (OCT) images. Frequency domain learning techniques can enhance the feature representations of deep neural networks (DNNs) by decomposing frequency components involving rich texture features. Additionally, previous works have not exploited texture features for automated PD screening in OCT. Motivated by the above analysis, we propose a novel Adaptive Wavelet Filter (AWF) that serves as the Practical Texture Feature Amplifier to fully leverage the merits of texture features to boost the PD screening performance of DNNs with the aid of frequency domain learning. Specifically, AWF first enhances texture feature representation diversities via channel mixer, then emphasizes informative texture feature representations with the well-designed adaptive wavelet filtering token mixer. By combining the AWFs with the DNN stem, AWFNet is constructed for automated PD screening. Additionally, we introduce a novel Balanced Confidence (BC) Loss by mining the potential of sample-wise predicted probabilities of all classes and class frequency prior, to further boost the PD screening performance and trustworthiness of AWFNet. The extensive experiments manifest the superiority of our AWFNet and BC over state-of-the-art methods in terms of PD screening performance and trustworthiness.

eess.IV↗

arXiv2025

Movie Gen: A Cast of Media Foundation Models

We present Movie Gen, a cast of foundation models that generates high-quality, 1080p HD videos with different aspect ratios and synchronized audio. We also show additional capabilities such as precise instruction-based video editing and generation of personalized videos based on a user's image. Our models set a new state-of-the-art on multiple tasks: text-to-video synthesis, video personalization, video editing, video-to-audio generation, and text-to-audio generation. Our largest video generation model is a 30B parameter transformer trained with a maximum context length of 73K video tokens, corresponding to a generated video of 16 seconds at 16 frames-per-second. We show multiple technical innovations and simplifications on the architecture, latent spaces, training objectives and recipes, data curation, evaluation protocols, parallelization techniques, and inference optimizations that allow us to reap the benefits of scaling pre-training data, model size, and training compute for training large scale media generation models. We hope this paper helps the research community to accelerate progress and innovation in media generation models. All videos from this paper are available at https://go.fb.me/MovieGenResearchVideos.

cs.CV↗

arXiv2025

Pseudo-Measurement Enhancement in Power Distribution Systems

With the rapid development of smart distribution networks (DNs), the integrity and accuracy of grid measurement data are crucial to the safety and stability of the entire system. However, the quality of the user power consumption data cannot be guaranteed during the collection and transmission process. To this end, this paper proposes a low-rank tensor completion model based on CANDECOMP/PARAFAC decomposition (CPD-LRTC) to enhance the quality of the measurement data of the DNs. Firstly, the causes and the associated characteristics of the missing data are analyzed, and a third-order standard tensor is constructed as a mathematical model of the measurement data of the DN. Then, a completion model is established based on the characteristics of measurement data and the low rank of the completion tensor, and the alternating direction method of multipliers (ADMM) is used to solve it iteratively. Finally, the proposed model is verified through two case studies, the completion accuracy, the computational efficiency, and the memory usage are compared to traditional methods.

eess.SY↗

arXiv2025

Dancing with invisible partners: Three-body exchanges with primordial black holes

The abundance of massive primordial black holes has historically been constrained by dynamical probes. Since these objects can participate in hard few-body scattering processes, they can readily transfer energy to stellar systems, and, in particular, can disrupt wide binaries. However, disruption is not the only possible outcome of such few-body processes. Primordial black holes could also participate in exchange processes, in which one component of a binary system is ejected and replaced by the black hole itself. In this case, the remaining object in the binary would dynamically appear to have an invisible companion. We study the rate of exchange processes for primordial black holes as a component of dark matter and evaluate possible mechanisms for detecting such binaries. We find that many such binaries plausibly exist in the Solar neighborhood, and show that this process can account for observed binary systems whose properties run counter to the predictions of isolated binary evolution.

hep-ph↗

arXiv2025

Multi-messenger Probes of Asteroid Mass Primordial Black Holes: Superradiance Spectroscopy, Hawking Radiation, and Microlensing

Superradiance provides a unique opportunity for investigating dark sectors as well as primordial black holes, which themselves are candidates for dark matter over a wide mass range. Using axion-like particles as an example, we show that line signals emerging from a superradiated axion cloud combined with black hole Hawking radiation in extragalactic and galactic halos, along with microlensing observations lead to complementary constraints on parameter space combinations including the axion-photon coupling, axion mass, black hole mass, and its dark matter fraction, $f_{\rm PBH}$. For the asteroid mass range $\sim10^{16}-10^{22}~{\rm g}$, where primordial black holes can provide the totality of dark matter, we demonstrate that ongoing and upcoming observations such as SXI, JWST, and AMEGO-X will be sensitive to possible line and continuum signals, respectively, providing probes of previously inaccessible regions of $f_{\rm PBH}$ parameter space. Further complementarity from a stochastic gravitational-wave background emerging from the black hole formation mechanism is also considered.

hep-ph↗

arXiv2025

Learnings from Scaling Visual Tokenizers for Reconstruction and Generation

Visual tokenization via auto-encoding empowers state-of-the-art image and video generative models by compressing pixels into a latent space. Although scaling Transformer-based generators has been central to recent advances, the tokenizer component itself is rarely scaled, leaving open questions about how auto-encoder design choices influence both its objective of reconstruction and downstream generative performance. Our work aims to conduct an exploration of scaling in auto-encoders to fill in this blank. To facilitate this exploration, we replace the typical convolutional backbone with an enhanced Vision Transformer architecture for Tokenization (ViTok). We train ViTok on large-scale image and video datasets far exceeding ImageNet-1K, removing data constraints on tokenizer scaling. We first study how scaling the auto-encoder bottleneck affects both reconstruction and generation -- and find that while it is highly correlated with reconstruction, its relationship with generation is more complex. We next explored the effect of separately scaling the auto-encoders' encoder and decoder on reconstruction and generation performance. Crucially, we find that scaling the encoder yields minimal gains for either reconstruction or generation, while scaling the decoder boosts reconstruction but the benefits for generation are mixed. Building on our exploration, we design ViTok as a lightweight auto-encoder that achieves competitive performance with state-of-the-art auto-encoders on ImageNet-1K and COCO reconstruction tasks (256p and 512p) while outperforming existing auto-encoders on 16-frame 128p video reconstruction for UCF-101, all with 2-5x fewer FLOPs. When integrated with Diffusion Transformers, ViTok demonstrates competitive performance on image generation for ImageNet-1K and sets new state-of-the-art benchmarks for class-conditional video generation on UCF-101.

cs.CV↗

arXiv2024

Reject Threshold Adaptation for Open-Set Model Attribution of Deepfake Audio

Open environment oriented open set model attribution of deepfake audio is an emerging research topic, aiming to identify the generation models of deepfake audio. Most previous work requires manually setting a rejection threshold for unknown classes to compare with predicted probabilities. However, models often overfit training instances and generate overly confident predictions. Moreover, thresholds that effectively distinguish unknown categories in the current dataset may not be suitable for identifying known and unknown categories in another data distribution. To address the issues, we propose a novel framework for open set model attribution of deepfake audio with rejection threshold adaptation (ReTA). Specifically, the reconstruction error learning module trains by combining the representation of system fingerprints with labels corresponding to either the target class or a randomly chosen other class label. This process generates matching and non-matching reconstructed samples, establishing the reconstruction error distributions for each class and laying the foundation for the reject threshold calculation module. The reject threshold calculation module utilizes gaussian probability estimation to fit the distributions of matching and non-matching reconstruction errors. It then computes adaptive reject thresholds for all classes through probability minimization criteria. The experimental results demonstrate the effectiveness of ReTA in improving the open set model attributes of deepfake audio.

cs.SD↗

arXiv2024

Smart Predict-then-Optimize Method with Dependent Data: Risk Bounds and Calibration of Autoregression

The predict-then-optimize (PTO) framework is indispensable for addressing practical stochastic decision-making tasks. It consists of two crucial steps: initially predicting unknown parameters of an optimization model and subsequently solving the problem based on these predictions. Elmachtoub and Grigas [1] introduced the Smart Predict-then-Optimize (SPO) loss for the framework, which gauges the decision error arising from predicted parameters, and a convex surrogate, the SPO+ loss, which incorporates the underlying structure of the optimization model. The consistency of these different loss functions is guaranteed under the assumption of i.i.d. training data. Nevertheless, various types of data are often dependent, such as power load fluctuations over time. This dependent nature can lead to diminished model performance in testing or real-world applications. Motivated to make intelligent predictions for time series data, we present an autoregressive SPO method directly targeting the optimization problem at the decision stage in this paper, where the conditions of consistency are no longer met. Therefore, we first analyze the generalization bounds of the SPO loss within our autoregressive model. Subsequently, the uniform calibration results in Liu and Grigas [2] are extended in the proposed model. Finally, we conduct experiments to empirically demonstrate the effectiveness of the SPO+ surrogate compared to the absolute loss and the least squares loss, especially when the cost vectors are determined by stationary dynamical systems and demonstrate the relationship between normalized regret and mixing coefficients.

eess.SY↗