Search

Iodinated contrast agents are widely utilized in numerous interventional procedures, yet posing substantial health risks to patients. This paper presents CAS-GAN, a novel GAN framework that serves as a ``virtual contrast agent" to synthesize X-ray angiographies via disentanglement representation learning and vessel semantic guidance, thereby reducing the reliance on iodinated agents during interventional procedures. Specifically, our approach disentangles X-ray angiographies into background and vessel components, leveraging medical prior knowledge. A specialized predictor then learns to map the interrelationships between these components. Additionally, a vessel semantic-guided generator and a corresponding loss function are introduced to enhance the visual fidelity of generated images. Experimental results on the XCAD dataset demonstrate the state-of-the-art performance of our CAS-GAN, achieving a FID of 5.94 and a MMD of 0.017. These promising results highlight CAS-GAN's potential for clinical applications., Comment: 8 pages, 4 figures

System level synthesis (SLS) is a controller parameterization technique that facilitates distributed structured control via convex techniques. Results on SLS are primarily in the discrete-time setting; this paper extends SLS to the continuous-time setting. We translate the parametrization and associated constraints to continuous time, and propose a controller design procedure consisting of two steps: (1) pole selection and (2) optimization over closed-loops. We provide SLS reformulations of H2 and Hinf control, and show that the proposed procedure allows for convex design of structured H2 and Hinf controllers. We verify our methods in simulation on a grid of linearized swing equations. The resulting structured (i.e. sparse) controllers perform similarly (in some cases within 1\% cost) as the centralized (i.e. dense) controllers. The proposed procedure preserves the scalability and disturbance-rejection features of the original discrete-time SLS framework., Comment: 8 pages, 6 figures, conference

There is a growing interest in expanding the input capacity of language models (LMs) across various domains. However, simply increasing the context window does not guarantee robust performance across diverse long-input processing tasks, such as understanding extensive documents and extracting detailed information from lengthy and noisy data. In response, we introduce SEGMENT+, a general framework that enables LMs to handle extended inputs within limited context windows efficiently. SEGMENT+ utilizes structured notes and a filtering module to manage information flow, resulting in a system that is both controllable and interpretable. Our extensive experiments across various model sizes, focusing on long-document question-answering and Needle-in-a-Haystack tasks, demonstrate the effectiveness of SEGMENT+ in improving performance., Comment: EMNLP 2024

Embodied AI represents a paradigm in AI research where artificial agents are situated within and interact with physical or virtual environments. Despite the recent progress in Embodied AI, it is still very challenging to learn the generalizable manipulation skills that can handle large deformation and topological changes on soft-body objects, such as clay, water, and soil. In this work, we proposed an effective policy, namely GP2E behavior cloning policy, which can guide the agent to learn the generalizable manipulation skills from soft-body tasks, including pouring, filling, hanging, excavating, pinching, and writing. Concretely, we build our policy from three insights:(1) Extracting intricate semantic features from point cloud data and seamlessly integrating them into the robot's end-effector frame; (2) Capturing long-distance interactions in long-horizon tasks through the incorporation of our guided self-attention module; (3) Mitigating overfitting concerns and facilitating model convergence to higher accuracy levels via the introduction of our two-stage fine-tuning strategy. Through extensive experiments, we demonstrate the effectiveness of our approach by achieving the 1st prize in the soft-body track of the ManiSkill2 Challenge at the CVPR 2023 4th Embodied AI workshop. Our findings highlight the potential of our method to improve the generalization abilities of Embodied AI models and pave the way for their practical applications in real-world scenarios.

The non-Abelian braiding describes the exchange behavior of anyons, which can be leveraged to encode qubits for quantum computing. Recently, this concept has been realized in classical photonic and acoustic systems. However, these implementations are constrained by adiabatic conditions, necessitating long operation distances and impeding practical applications. Here, we conceive and demonstrate a shortcut to adiabatic (STA) braiding of telecommunication light in three-dimensional silicon photonic chips. Our device comprises tri-layer silicon waveguides stacked and embedded in the SU-8 polymer, employing an STA strategy to expedite the braiding operations and give rise to compact devices that function as photonic quantum X, Y, and Z gates. We further experimentally observed non-Abelian braiding behaviors based on this STA-braiding scheme. Remarkably, this achievement represents the most compact braiding apparatus ever reported, with a size reduction of nearly three orders of magnitude compared to previous works. This work presents a feasible approach to accelerating adiabatic braiding evolutions, paving the way for compact, CMOS-compatible non-Abelian photonic devices.

We examine the implicit bias of mirror flow in univariate least squares error regression with wide and shallow neural networks. For a broad class of potential functions, we show that mirror flow exhibits lazy training and has the same implicit bias as ordinary gradient flow when the network width tends to infinity. For ReLU networks, we characterize this bias through a variational problem in function space. Our analysis includes prior results for ordinary gradient flow as a special case and lifts limitations which required either an intractable adjustment of the training data or networks with skip connections. We further introduce scaled potentials and show that for these, mirror flow still exhibits lazy training but is not in the kernel regime. For networks with absolute value activations, we show that mirror flow with scaled potentials induces a rich class of biases, which generally cannot be captured by an RKHS norm. A takeaway is that whereas the parameter initialization determines how strongly the curvature of the learned function is penalized at different locations of the input space, the scaled potential determines how the different magnitudes of the curvature are penalized.

Polymorphism, commonly denoting the variety of molecular or crystal structures, is a vital element in many natural science disciplines. In van der Waals layered antiferromagnets, a new type of magnetic polymorphism is allowed by having multiple layer-selective magnetic structures with the same total magnetization. However, resolving and manipulating such magnetic polymorphs remain a great challenge. Here we use the phase-resolved magnetic second-harmonic generation microscopy to elucidate such magnetic polymorphism in the 2D semiconducting layered antiferromagnet chromium sulfur bromide (CrSBr), and demonstrate how the magnetic polymorphs can be deterministically switched in an unprecedented layer-selective manner. With the nonlinear magneto-optical technique unveiling the magnetic symmetry information through the amplitude and phase of light, we could unambiguously resolve the polymorphic spin-flip transitions in CrSBr bilayers and tetralayers. Remarkably, the deterministic routing of polymorphic transitions originates from the breaking of energy degeneracy via a magnetic layer-sharing effect: the spin-flip transitions in a tetralayer are governed by the laterally extended bilayer, which acts as a control bit. We envision such controllable magnetic polymorphism to be ubiquitous for van der Waals layered antiferromagnets, and could lead to conceptually new design and construction of spintronic and opto-spintronic devices for probabilistic computation and neuromorphic engineering., Comment: 22 pages, 6 figures

The cause of the speed-accuracy tradeoff (typically quantified via Fitts' Law) is a debated topic of interest in motor neuroscience, and is commonly studied using tools from control theory. Two prominent theories involve the presence of signal dependent motor noise and planning variability -- these factors are generally incorporated separately. In this work, we study how well the simultaneous presence of both factors explains the speed-accuracy tradeoff. A human arm reaching model is developed with bio-realistic signal dependent motor noise, and a Gaussian noise model is used to deterministically approximate the motor noise. Both offline trajectory optimization and online model predictive control are used to simulate the planning and execution of several different reaching tasks with varying target sizes and movement durations. These reaching trajectories are then compared to experimental human reaching data, revealing that both models produce behavior consistent with humans, and the speed-accuracy tradeoff is present in both online and offline control. These results suggest the speed-accuracy tradeoff is likely caused by a combination of these two factors, and also that it plays a role in both offline and online computation., Comment: Submitted to IEEE American Control Conference

We study the task of balancing a human on a log that is fixed in place. Balancing on a log is substantially more challenging than balancing on a flat surface -- to achieve stability, we use a switched multi-layer controller. The controller consists of an upper-layer LQR planner (akin to the central nervous system) that coordinates ankle and hip torques, and lower-layer PID trackers (akin to local motor units) that follow this plan subject to nonlinear dynamics. Additionally, the controller switches between three operational modes depending on the current state of the human. The efficacy of the controller is verified in simulation, where our controller is able to stabilize the human for a variety of initial conditions. We also show that this controller is compatible with muscle-based actuation and imperfect sensing, making it a promising candidate for modeling motor control under challenging conditions in a more bio-realistic way., Comment: Submitted to IEEE American Control Conference

Animal sensorimotor behavior is frequently modeled using optimal controllers. However, it is unclear how the neuronal circuits within the animal's nervous system implement optimal controller-like behavior. In this work, we study the question of implementing a delayed linear quadratic regulator with linear dynamical "neurons" on a muscle model. We show that for any second-order controller, there are three minimal neural circuit configurations that implement the same controller. Furthermore, the firing rate characteristics of each circuit can vary drastically, even as the overall controller behavior is preserved. Along the way, we introduce concepts that bridge controller realizations to neural implementations that are compatible with known neuronal delay structures., Comment: Submitted to IEEE American Control Conference

In the framework of general relativity, the dynamics of a general barotropic fluid are coupled to the Einstein equations, which govern the structure of the underlying spacetime. We establish a priori estimates and well-posedness in Sobolev spaces for this model with a free boundary. Within the frame parallel-transported by the fluid velocity, we decompose the fluid and geometric quantities. The fluid components are estimated via a coupled interior-boundary wave equation, while the geometric quantities are analyzed through the Bianchi equations. Compared to a previous work, the results in present paper allow general equations of state and non-zero vorticities.

We investigate LoRA in federated learning through the lens of the asymmetry analysis of the learned $A$ and $B$ matrices. In doing so, we uncover that $A$ matrices are responsible for learning general knowledge, while $B$ matrices focus on capturing client-specific knowledge. Based on this finding, we introduce Federated Share-A Low-Rank Adaptation (FedSA-LoRA), which employs two low-rank trainable matrices $A$ and $B$ to model the weight update, but only $A$ matrices are shared with the server for aggregation. Moreover, we delve into the relationship between the learned $A$ and $B$ matrices in other LoRA variants, such as rsLoRA and VeRA, revealing a consistent pattern. Consequently, we extend our FedSA-LoRA method to these LoRA variants, resulting in FedSA-rsLoRA and FedSA-VeRA. In this way, we establish a general paradigm for integrating LoRA with FL, offering guidance for future work on subsequent LoRA variants combined with FL. Extensive experimental results on natural language understanding and generation tasks demonstrate the effectiveness of the proposed method. Our code is available at https://github.com/Pengxin-Guo/FedSA-LoRA.

In the domain of battery research, the processing of high-resolution microscopy images is a challenging task, as it involves dealing with complex images and requires a prior understanding of the components involved. The utilization of deep learning methodologies for image analysis has attracted considerable interest in recent years, with multiple investigations employing such techniques for image segmentation and analysis within the realm of battery research. However, the automated analysis of high-resolution microscopy images for detecting phases and components in composite materials is still an underexplored area. This work proposes a novel workflow for detecting components and phase segmentation from raw high resolution transmission electron microscopy (TEM) images using a trained U-Net segmentation model. The developed model can expedite the detection of components and phase segmentation, diminishing the temporal and cognitive demands associated with scrutinizing an extensive array of TEM images, thereby mitigating the potential for human errors. This approach presents a novel and efficient image analysis approach with broad applicability beyond the battery field and holds potential for application in other related domains characterized by phase and composition distribution, such as alloy production.

Augmenting the multi-step reasoning abilities of Large Language Models (LLMs) has been a persistent challenge. Recently, verification has shown promise in improving solution consistency by evaluating generated outputs. However, current verification approaches suffer from sampling inefficiencies, requiring a large number of samples to achieve satisfactory performance. Additionally, training an effective verifier often depends on extensive process supervision, which is costly to acquire. In this paper, we address these limitations by introducing a novel verification method based on Twisted Sequential Monte Carlo (TSMC). TSMC sequentially refines its sampling effort to focus exploration on promising candidates, resulting in more efficient generation of high-quality solutions. We apply TSMC to LLMs by estimating the expected future rewards at partial solutions. This approach results in a more straightforward training target that eliminates the need for step-wise human annotations. We empirically demonstrate the advantages of our method across multiple math benchmarks, and also validate our theoretical analysis of both our approach and existing verification methods.

I present the Lower Biased Teacher model, an enhancement of the Unbiased Teacher model, specifically tailored for semi-supervised object detection tasks. The primary innovation of this model is the integration of a localization loss into the teacher model, which significantly improves the accuracy of pseudo-label generation. By addressing key issues such as class imbalance and the precision of bounding boxes, the Lower Biased Teacher model demonstrates superior performance in object detection tasks. Extensive experiments on multiple semi-supervised object detection datasets show that the Lower Biased Teacher model not only reduces the pseudo-labeling bias caused by class imbalances but also mitigates errors arising from incorrect bounding boxes. As a result, the model achieves higher mAP scores and more reliable detection outcomes compared to existing methods. This research underscores the importance of accurate pseudo-label generation and provides a robust framework for future advancements in semi-supervised learning for object detection., Comment: 12pages,2 figures,2 tables, several fomulas. arXiv admin note: text overlap with arXiv:2102.09480 by other authors

The coalescence of binary neutron stars (BNS) is a prolific source of gravitational waves (GWs) and electromagnetic (EM) radiation, offering a dual observational window into the Universe. Lowering the signal-to-noise ratio (S/N) threshold is a simple and cost-effective way to enhance the detection probability of GWs from BNS mergers. In this study, we introduce a metric of the purity of joint GW and EM detections $P_{\rm joint}$, which is in analogue to $P_{\rm astro}$ in GW only observations. By simulating BNS merger GWs jointly detected by the HLV network and EM counterparts (kilonovae and short Gamma-ray bursts, sGRBs) with an assumed merger rate density of BNS, we generate catalogs of GW events and EM counterparts. Through this simulation, we analyze joint detection pairs, both correct and misidentified. We find the following: 1. For kilonovae, requiring $P_{\rm joint}>$ 95\% instead of $P_{\rm astro}>95\%$ reduces the S/N from 9.2 to 8.5-8.8, allowing 5-13 additional joint detections per year and increasing the GW detection volume by 9-17\%; 2. For sGRBs, requiring $P_{\rm joint}>$ 95\% instead of $P_{\rm astro}$ reduces the S/N from 9.2 to 8.1-8.5; 3. Increasing kilonova or sGRB detection capability does not improve $P_{\rm joint}$ due to a higher rate of misidentifications. We also show that sub-threshold GW and kilonova detections can reduce the uncertainty in measuring the Hubble constant to 89-92\% of its original value, and sub-threshold GW and sGRB observations can enhance the precision of constraining the speed of GWs to 88\% of previously established values., Comment: Accepted for publication in MNRAS, 8 pages, 7 figures

Quantum key distribution (QKD) enables the generation of secure keys between two distant users. Security proof of QKD against general coherent attacks is challenging, while the one against collective attacks is much easier. As an effective and general solution, the postselection method tries to extend security analyses of collective attacks to be against coherent attacks. However, it gives a bad performance. To overcome this drawback, instead of directly calculating key rate by postselection method, we propose a method correlating the failure probabilities of phase error estimation against collective and coherent attacks, enabling the use of the independent and identically distributed assumption in parameter estimation against coherent attacks. Then the key rate can be obtained by uncertainty relation of entropy. Our method can be applied to various QKD protocols, providing better performance compared with the traditional postselection method. For instance, we give the finite-key analyses of the side-channel-secure (SCS) QKD and the no-phase-postselection (NPP) twin-field (TF) QKD to show their performance improvements with the proposed method.

As a promising candidate for altermagnet, CrSb possesses a distinctive compensated spin split band structure that could bring groundbreaking concepts to the field of spintronics. In this work, we have grown high-quality CrSb single crystals and comprehensively investigated their electronic and magneto-transport properties. We have observed large, positive, and non-saturated magnetoresistance (MR) in CrSb, which well obeys Kohler's rule, indicating its classic Lorentz scattering origins. Remarkably, a nonlinear magnetic field dependence of Hall effect resembling the spontaneous anomalous Hall is identified over a wide temperature range. After careful analysis of the transport data, we conclude the non-linearity mainly stems from the incorporation of different carriers in the magnetoconductivity. According to the Fermi surface analyses of CrSb, we applied the three-carrier model to fit the conductivity data, yielding good agreement. The extracted carrier concentration and mobility indicates that CrSb behaves more like a semimetal, with the highest mobility reaching 3*103 cm2V-1s-1. Furthermore, calculations using the semiclassical Boltzmann transport theory have successfully reproduced the main features of the experimental MR and Hall effect in CrSb. These exceptional transport properties make CrSb unique for applications in spintronics as an altermagnet.

Ghost imaging (GI) is a high-resolution imaging technology that has been a subject of interest to many fields in the past 20 years. Most GI researchers focus on the reconstruction of signal under-sampling, nevertheless, how to use information redundancy to improve the result's belief in a complex environment has hardly been studied. Motivated by this, we propose a computational GI system based on the low-density parity-check (LDPC) coded radiation field by exploiting the signal redundancy. The non-ideal factors generated within the imaging process can be eliminated by setting up the matching fading channel model. We have derived the analytical lower bound on the bit error rate for the proposed LDPC-coded GI system. The effectiveness and performance of the LDPC-coded GI system are further validated through numerical and experiment results.

During the fourth observing run, the LIGO-Virgo-KAGRA Collaboration reported the detection of a coalescing compact binary (GW230529$_{-}$181500) with component masses estimated at $2.5-4.5\, M_\odot$ and $1.2-2.0\, M_\odot$ with 90\% credibility. Given the current constraints on the maximum neutron star (NS) mass, this event is most likely a lower-mass-gap (LMG) black hole-neutron star (BHNS) binary. The spin magnitude of the BH, especially when aligned with the orbital angular momentum, is critical in determining whether the NS is tidally disrupted. An LMG BHNS merger with a rapidly spinning BH is an ideal candidate for producing electromagnetic counterparts. However, no such signals have been detected. In this study, we employ a detailed binary evolution model, incorporating new dynamical tide implementations, to explore the origin of BH spin in an LMG BHNS binary. If the NS forms first, the BH progenitor (He-rich star) must begin in orbit shorter than 0.35 days to spin up efficiently, potentially achieving a spin magnitude of $\chi_{\rm BH} > 0.3$. Alternatively, if a non-spinning BH (e.g., $M_{\rm BH} = 3.6\, M_\odot$) forms first, it can accrete up to $\sim 0.2\, M_\odot$ via Case BA mass transfer (MT), reaching a spin magnitude of $\chi_{\rm BH} \sim 0.18$ under Eddington-limited accretion. With a higher Eddington accretion limit (i.e., 10.0 $\Dot{M}_{\rm Edd}$), the BH can attain a significantly higher spin magnitude of $\chi_{\rm BH} \sim\,0.65$ by accreting approximately $1.0\, M_\odot$ during Case BA MT phase., Comment: 9 pages, 8 figures; submitted to AA; comments welcome!

The purpose of face super-resolution (FSR) is to reconstruct high-resolution (HR) face images from low-resolution (LR) inputs. With the continuous advancement of deep learning technologies, contemporary prior-guided FSR methods initially estimate facial priors and then use this information to assist in the super-resolution reconstruction process. However, ensuring the accuracy of prior estimation remains challenging, and straightforward cascading and convolutional operations often fail to fully leverage prior knowledge. Inaccurate or insufficiently utilized prior information inevitably degrades FSR performance. To address this issue, we propose a prior knowledge distillation network (PKDN) for FSR, which involves transferring prior information from the teacher network to the student network. This approach enables the network to learn priors during the training stage while relying solely on low-resolution facial images during the testing stage, thus mitigating the adverse effects of prior estimation inaccuracies. Additionally, we incorporate robust attention mechanisms to design a parsing map fusion block that effectively utilizes prior information. To prevent feature loss, we retain multi-scale features during the feature extraction stage and employ them in the subsequent super-resolution reconstruction process. Experimental results on benchmark datasets demonstrate that our PKDN approach surpasses existing FSR methods in generating high-quality face images.

Understanding atomic hydrogen (H) diffusion in multi-principal element alloys (MPEAs) is essential for advancing clean energy technologies such as H transport, storage, and nuclear fusion applications. However, the vast compositional space and the intricate chemical environments inherent in MPEAs pose significant obstacles. In this work, we address this challenge by developing a multifaceted machine learning framework that integrates machine-learning force field, neural network-driven kinetic Monte Carlo, and machine-learning symbolic regression. This framework allows for accurate investigation of H diffusion across the entire compositional space of body-centered cubic (BCC) refractory MoNbTaW alloys, achieving density functional theory accuracy. For the first time, we discover that H diffusion in MPEAs exhibits super-Arrhenius behavior, described by the Vogel-Fulcher-Tammann model, where the Vogel temperature correlates with the 5th percentile of H solution energy spectrum. We also derive robust analytical expressions that can be used to predict H diffusivity in general BCC MPEAs. Our findings further elucidate that chemical short-range order (SRO) generally does not impact H diffusion, except it enhances diffusion when "H-favoring" elements (notably Nb and Ta) are present in low concentrations. These findings not only enhance our understanding of H diffusion dynamics in general MPEAs but also guide the development of advanced MPEAs in H-related applications by manipulating element type, composition and SRO.

With the development of 3D and 2D data acquisition techniques, it has become easy to obtain point clouds and images of scenes simultaneously, which further facilitates dual-modal semantic segmentation. Most existing methods for simultaneously segmenting point clouds and images rely heavily on the quantity and quality of the labeled training data. However, massive point-wise and pixel-wise labeling procedures are time-consuming and labor-intensive. To address this issue, we propose a parallel dual-stream network to handle the semi-supervised dual-modal semantic segmentation task, called PD-Net, by jointly utilizing a small number of labeled point clouds, a large number of unlabeled point clouds, and unlabeled images. The proposed PD-Net consists of two parallel streams (called original stream and pseudo-label prediction stream). The pseudo-label prediction stream predicts the pseudo labels of unlabeled point clouds and their corresponding images. Then, the unlabeled data is sent to the original stream for self-training. Each stream contains two encoder-decoder branches for 3D and 2D data respectively. In each stream, multiple dual-modal fusion modules are explored for fusing the dual-modal features. In addition, a pseudo-label optimization module is explored to optimize the pseudo labels output by the pseudo-label prediction stream. Experimental results on two public datasets demonstrate that the proposed PD-Net not only outperforms the comparative semi-supervised methods but also achieves competitive performances with some fully-supervised methods in most cases.

Hyperspectral image (HSI) plays a vital role in various fields such as agriculture and environmental monitoring. However, due to the expensive acquisition cost, the number of hyperspectral images is limited, degenerating the performance of downstream tasks. Although some recent studies have attempted to employ diffusion models to synthesize HSIs, they still struggle with the scarcity of HSIs, affecting the reliability and diversity of the generated images. Some studies propose to incorporate multi-modal data to enhance spatial diversity, but the spectral fidelity cannot be ensured. In addition, existing HSI synthesis models are typically uncontrollable or only support single-condition control, limiting their ability to generate accurate and reliable HSIs. To alleviate these issues, we propose HSIGene, a novel HSI generation foundation model which is based on latent diffusion and supports multi-condition control, allowing for more precise and reliable HSI generation. To enhance the spatial diversity of the training data while preserving spectral fidelity, we propose a new data augmentation method based on spatial super-resolution, in which HSIs are upscaled first, and thus abundant training patches could be obtained by cropping the high-resolution HSIs. In addition, to improve the perceptual quality of the augmented data, we introduce a novel two-stage HSI super-resolution framework, which first applies RGB bands super-resolution and then utilizes our proposed Rectangular Guided Attention Network (RGAN) for guided HSI super-resolution. Experiments demonstrate that the proposed model is capable of generating a vast quantity of realistic HSIs for downstream tasks such as denoising and super-resolution. The code and models are available at https://github.com/LiPang/HSIGene.

On April 25th, 2019, the LIGO-Virgo Collaboration discovered a Gravitational-wave (GW) signal from a binary neutron star (BNS) merger, i.e., GW190425. Due to the inferred large total mass, the origin of GW190425 remains unclear. We perform detailed stellar structure and binary evolution calculations that take into account mass-loss, internal differential rotation, and tidal interactions between a He-rich star and a NS companion. We explore the parameter space of the initial binary properties, including initial NS and He-rich masses and initial orbital period. We find that the immediate post-common-envelope progenitor system, consisting of a primary $\sim2.0\,M_\odot$ ($\sim1.7\,M_\odot$) NS and a secondary He-rich star with an initial mass of $\sim3.0-5.5\,M_\odot$ ($\sim5.5-6.0\,M_\odot$) in a close binary with an initial period of $\sim0.08-0.5\,{\rm{days}}$ ($\sim 0.08-0.4\,{\rm{days}}$), that experiences stable Case BB/BC mass transfer (MT) during binary evolution, can reproduce the formation of GW190425-like BNS events. Our studies reveal that the secondary He-rich star of the GW190425's progenitor before its core collapse can be efficiently spun up through tidal interaction, finally remaining as a NS with rotational energy even reaching $\sim10^{52}\,{\rm{erg}}$, which is always much higher than the neutrino-driven energy of the supernova (SN) explosion. If the newborn secondary NS is a magnetar, we expect that GW190425 can be the remnant of a magnetar-driven SN, e.g., a magnetar-driven ultra-stripped SN, a superluminous SN, or a broad-line Type Ic SN. Our results show that GW190425 could be formed through the isolated binary evolution, which involves a stable Case BB/BC MT just after the common envelope phase. On top of that, we show the He-rich star can be tidally spun up, potentially forming a spinning magnetized NS (magnetar) during the second SN explosion., Comment: 7 pages, 8 figures, accepted for publication in A&A

The brightest-of-all-time gamma-ray burst (GRB), GRB 221009A, is the first GRB observed to have emission line (up to 37 MeV) in its prompt emission spectra. It is naturally explained as \pair annihilation line that was Doppler boosted in the relativistic jet of the GRB. In this work, we repeatedly apply the simple optical depth argument to different physical processes necessary to produce an observable \pair annihilation line. This approach results in robust constraints on the physics of the line: We conclude that in GRB 221009A, the \pair pairs were produced at a radius greater than $4.3\times 10^{15}$\,cm from the central engine, and annihilated in a region between $1.4\times 10^{16}$\,cm and $4.3\times 10^{16}$\,cm. From these constraints, we established a self-consistent picture of \pair production, cooling, and annihilation. We also derived a criterion for pair production in the GRB prompt emission: $E_{\rm{iso}} \gtrsim3.3\times 10^{53} E_{\rm{peak},100} (1+z) R^2_{\rm{prod},16}~\text{erg}$. Using this criterion, we find tens of candidate GRBs that could have produced \pair in prompt emissions to annihilate. GRB 221009A is with the highest likelihood according to this criterion. We also predict the presence of a thermal radiation, with a time-evolving black body temperature, sweeping through soft X-ray during the prompt emission phase., Comment: 10 pages, 4 figures, submitted to MNRAS

We tackle the problem of mosaicing bundle adjustment (i.e., simultaneous refinement of camera orientations and scene map) for a purely rotating event camera. We formulate the problem as a regularized non-linear least squares optimization. The objective function is defined using the linearized event generation model in the camera orientations and the panoramic gradient map of the scene. We show that this BA optimization has an exploitable block-diagonal sparsity structure, so that the problem can be solved efficiently. To the best of our knowledge, this is the first work to leverage such sparsity to speed up the optimization in the context of event-based cameras, without the need to convert events into image-like representations. We evaluate our method, called EMBA, on both synthetic and real-world datasets to show its effectiveness (50% photometric error decrease), yielding results of unprecedented quality. In addition, we demonstrate EMBA using high spatial resolution event cameras, yielding delicate panoramas in the wild, even without an initial map. Project page: https://github.com/tub-rip/emba, Comment: 14+11 pages, 11 figures, 10 tables, https://github.com/tub-rip/emba

Time-varying effects have unveiled new possibilities for manipulating electromagnetic waves through the temporal dimension. In this study, we experimentally explore these effects in the nonlinear optical process of terahertz (THz) generation using optically pumped indium tin oxide (ITO) films. The ultrafast carrier dynamics in the ITO film endow the second-order nonlinear susceptibility (\c{hi}(2)) with sub-picosecond temporal evolution, establishing a temporal boundary for the generated THz waves. We observe significant amplitude and frequency modulations in the THz generation at various transients, attributed to the time-varying complex amplitude of the \c{hi}(2). Moreover, we also observed polarization modulations when further exploiting the tensor properties of \c{hi}(2). This work advances the exploration of time-varying effects into the nonlinear regime through frequency down-conversion, effectively transferring the strong time-varying material response from the near-infrared (NIR) band to the THz band. These findings open up new opportunities for realizing time-varying phenomena that specifically require single-cycle modulation.

The Agatston score, which is the sum of the calcification in the four main coronary arteries, has been widely used in the diagnosis of coronary artery disease (CAD). However, many studies have emphasized the importance of the vessel-specific Agatston score, as calcification in a specific vessel is significantly correlated with the occurrence of coronary heart disease (CHD). In this paper, we propose the Residual-block Inspired Coordinate Attention U-Net (RICAU-Net), which incorporates coordinate attention in two distinct manners and a customized combo loss function for lesion-specific coronary artery calcium (CAC) segmentation. This approach aims to tackle the high class-imbalance issue associated with small and sparse CAC lesions. Experimental results and the ablation study demonstrate that the proposed method outperforms the four other U-Net based methods used in medical applications, by achieving the highest per-lesion Dice scores across all four lesions., Comment: This work has been submitted to the IEEE for possible publication. Copyright may be transferred without notice, after which this version may no longer be accessible

Diffusion models have achieved remarkable success in sequential decision-making by leveraging the highly expressive model capabilities in policy learning. A central problem for learning diffusion policies is to align the policy output with human intents in various tasks. To achieve this, previous methods conduct return-conditioned policy generation or Reinforcement Learning (RL)-based policy optimization, while they both rely on pre-defined reward functions. In this work, we propose a novel framework, Forward KL regularized Preference optimization for aligning Diffusion policies, to align the diffusion policy with preferences directly. We first train a diffusion policy from the offline dataset without considering the preference, and then align the policy to the preference data via direct preference optimization. During the alignment phase, we formulate direct preference learning in a diffusion policy, where the forward KL regularization is employed in preference optimization to avoid generating out-of-distribution actions. We conduct extensive experiments for MetaWorld manipulation and D4RL tasks. The results show our method exhibits superior alignment with preferences and outperforms previous state-of-the-art algorithms.

For audio-driven visual dubbing, it remains a considerable challenge to uphold and highlight speaker's persona while synthesizing accurate lip synchronization. Existing methods fall short of capturing speaker's unique speaking style or preserving facial details. In this paper, we present PersonaTalk, an attention-based two-stage framework, including geometry construction and face rendering, for high-fidelity and personalized visual dubbing. In the first stage, we propose a style-aware audio encoding module that injects speaking style into audio features through a cross-attention layer. The stylized audio features are then used to drive speaker's template geometry to obtain lip-synced geometries. In the second stage, a dual-attention face renderer is introduced to render textures for the target geometries. It consists of two parallel cross-attention layers, namely Lip-Attention and Face-Attention, which respectively sample textures from different reference frames to render the entire face. With our innovative design, intricate facial details can be well preserved. Comprehensive experiments and user studies demonstrate our advantages over other state-of-the-art methods in terms of visual quality, lip-sync accuracy and persona preservation. Furthermore, as a person-generic framework, PersonaTalk can achieve competitive performance as state-of-the-art person-specific methods. Project Page: https://grisoon.github.io/PersonaTalk/., Comment: Accepted at SIGGRAPH Asia 2024 (Conference Track)

High-Definition Maps (HD maps) are essential for the precise navigation and decision-making of autonomous vehicles, yet their creation and upkeep present significant cost and timeliness challenges. The online construction of HD maps using on-board sensors has emerged as a promising solution; however, these methods can be impeded by incomplete data due to occlusions and inclement weather. This paper proposes the PriorDrive framework to addresses these limitations by harnessing the power of prior maps, significantly enhancing the robustness and accuracy of online HD map construction. Our approach integrates a variety of prior maps, such as OpenStreetMap's Standard Definition Maps (SD maps), outdated HD maps from vendors, and locally constructed maps from historical vehicle data. To effectively encode this prior information into online mapping models, we introduce a Hybrid Prior Representation (HPQuery) that standardizes the representation of diverse map elements. At the core of PriorDrive is the Unified Vector Encoder (UVE), which employs a dual encoding mechanism to process vector data. The intra-vector encoder captures fine-grained local features, while the inter-vector encoder integrates global context. Furthermore, we propose a segment-level and point-level pre-training strategy that enables the UVE to learn the prior distribution of vector data, thereby improving the encoder's generalizability and performance. Through extensive testing on the nuScenes dataset, we demonstrate that PriorDrive is highly compatible with various online mapping models and substantially improves map prediction capabilities. The integration of prior maps through the PriorDrive framework offers a robust solution to the challenges of single-perception data, paving the way for more reliable autonomous vehicle navigation.

It remains a long-standing problem, unsettled for almost two decades in General Relativity (GR) community, ever since Griffiths and Podolsky demonstrated in Ref. [J.B. Griffiths and J. Podolsky, Class. Quant. Grav. 22, 3467 (2005)] that the type-D NUT C-metric seems to be absent from the most general family of the type-D Pleba\'nski-Demia\'nski (P-D) solution. However, Astorino [Phys. Rev. D 109, 084038 (2024)] presented a different form of rotating and accelerating black holes and showed that all known four-dimensional type-D accelerating black holes (without the NUT charge) can be recovered via various different limits in a definitive fashion. In particular, he provided, for the first time, the correct expressions for the type-D static accelerating black holes with a nonzero NUT charge, which was previously impossible using the traditional parametrization of the familiar P-D solution. Nevertheless, it still remains elusive that how these two different forms of the four-dimensional rotating and accelerating solutions are related. In this paper, we aim to fill this gap by finding the obvious coordinate transformations and parameter identifications between the vacuum metrics after two different parameterizations of the generated solution via the inverse scattering method from the seed metric -- the Rindler vacuum background. We then resolve this ``missing" puzzle by providing another M\"{o}bius transformation and linear combinations of the Killing coordinates, which clearly cast the type-D NUT C-metric into the familiar form of the P-D solution. Additionally, we propose an alternative new routine for the normalization of the obtained metric derived via the inverse scattering method from the vacuum seed solution, which could be potentially useful for the construction of higher-dimensional solutions using the trivial vacuum background as the seed metric., Comment: 16 pages, No figure, revtex4-1. During the arXiv's onhold of our submission (submit/5815878) on August 27, 2409.02308 appears with a partial overlap with the subject of our paper

Recently, prompt learning has garnered considerable attention for its success in various Vision-Language (VL) tasks. However, existing prompt-based models are primarily focused on studying prompt generation and prompt strategies with complete modality settings, which does not accurately reflect real-world scenarios where partial modality information may be missing. In this paper, we present the first comprehensive investigation into prompt learning behavior when modalities are incomplete, revealing the high sensitivity of prompt-based models to missing modalities. To this end, we propose a novel Multi-step Adaptive Prompt Learning (MuAP) framework, aiming to generate multimodal prompts and perform multi-step prompt tuning, which adaptively learns knowledge by iteratively aligning modalities. Specifically, we generate multimodal prompts for each modality and devise prompt strategies to integrate them into the Transformer model. Subsequently, we sequentially perform prompt tuning from single-stage and alignment-stage, allowing each modality-prompt to be autonomously and adaptively learned, thereby mitigating the imbalance issue caused by only textual prompts that are learnable in previous works. Extensive experiments demonstrate the effectiveness of our MuAP and this model achieves significant improvements compared to the state-of-the-art on all benchmark datasets

In addressing the persistent challenges of data-sparsity and cold-start issues in domain-expert recommender systems, Cross-Domain Recommendation (CDR) emerges as a promising methodology. CDR aims at enhancing prediction performance in the target domain by leveraging interaction knowledge from related source domains, particularly through users or items that span across multiple domains (e.g., Short-Video and Living-Room). For academic research purposes, there are a number of distinct aspects to guide CDR method designing, including the auxiliary domain number, domain-overlapped element, user-item interaction types, and downstream tasks. With so many different CDR combination scenario settings, the proposed scenario-expert approaches are tailored to address a specific vertical CDR scenario, and often lack the capacity to adapt to multiple horizontal scenarios. In an effect to coherently adapt to various scenarios, and drawing inspiration from the concept of domain-invariant transfer learning, we extend the former SOTA model UniCDR in five different aspects, named as UniCDR+. Our work was successfully deployed on the Kuaishou Living-Room RecSys., Comment: Work in progress

Modeling outdoor scenes for the synthetic 3D environment requires the recovery of reflectance/albedo information from raw images, which is an ill-posed problem due to the complicated unmodeled physics in this process (e.g., indirect lighting, volume scattering, specular reflection). The problem remains unsolved in a practical context. The recovered albedo can facilitate model relighting and shading, which can further enhance the realism of rendered models and the applications of digital twins. Typically, photogrammetric 3D models simply take the source images as texture materials, which inherently embed unwanted lighting artifacts (at the time of capture) into the texture. Therefore, these polluted textures are suboptimal for a synthetic environment to enable realistic rendering. In addition, these embedded environmental lightings further bring challenges to photo-consistencies across different images that cause image-matching uncertainties. This paper presents a general image formation model for albedo recovery from typical aerial photogrammetric images under natural illuminations and derives the inverse model to resolve the albedo information through inverse rendering intrinsic image decomposition. Our approach builds on the fact that both the sun illumination and scene geometry are estimable in aerial photogrammetry, thus they can provide direct inputs for this ill-posed problem. This physics-based approach does not require additional input other than data acquired through the typical drone-based photogrammetric collection and was shown to favorably outperform existing approaches. We also demonstrate that the recovered albedo image can in turn improve typical image processing tasks in photogrammetry such as feature and dense matching, edge, and line extraction., Comment: ISPRS Journal of Photogrammetry and Remote Sensing

A critical step in the digital surface models(DSM) generation is feature matching. Off-track (or multi-date) satellite stereo images, in particular, can challenge the performance of feature matching due to spectral distortions between images, long baseline, and wide intersection angles. Feature matching methods have evolved over the years from handcrafted methods (e.g., SIFT) to learning-based methods (e.g., SuperPoint and SuperGlue). In this paper, we compare the performance of different features, also known as feature extraction and matching methods, applied to satellite imagery. A wide range of stereo pairs(~500) covering two separate study sites are used. SIFT, as a widely used classic feature extraction and matching algorithm, is compared with seven deep-learning matching methods: SuperGlue, LightGlue, LoFTR, ASpanFormer, DKM, GIM-LightGlue, and GIM-DKM. Results demonstrate that traditional matching methods are still competitive in this age of deep learning, although for particular scenarios learning-based methods are very promising., Comment: ECCV2024 Workshop - TradiCV

To improve storage and transmission, images are generally compressed. Vector quantization (VQ) is a popular compression method as it has a high compression ratio that suppresses other compression techniques. Despite this, existing adversarial attack methods on image classification are mostly performed in the pixel domain with few exceptions in the compressed domain, making them less applicable in real-world scenarios. In this paper, we propose a novel one-index attack method in the VQ domain to generate adversarial images by a differential evolution algorithm, successfully resulting in image misclassification in victim models. The one-index attack method modifies a single index in the compressed data stream so that the decompressed image is misclassified. It only needs to modify a single VQ index to realize an attack, which limits the number of perturbed indexes. The proposed method belongs to a semi-black-box attack, which is more in line with the actual attack scenario. We apply our method to attack three popular image classification models, i.e., Resnet, NIN, and VGG16. On average, 55.9% and 77.4% of the images in CIFAR-10 and Fashion MNIST, respectively, are successfully attacked, with a high level of misclassification confidence and a low level of image perturbation.

Recent advancements in large language models (LLMs) have enabled understanding webpage contexts, product details, and human instructions. Utilizing LLMs as the foundational architecture for either reward models or policies in reinforcement learning has gained popularity -- a notable achievement is the success of InstructGPT. RL algorithms have been instrumental in maximizing long-term customer satisfaction and avoiding short-term, myopic goals in industrial recommender systems, which often rely on deep learning models to predict immediate clicks or purchases. In this project, several RL methods are implemented and evaluated using the WebShop benchmark environment, data, simulator, and pre-trained model checkpoints. The goal is to train an RL agent to maximize the purchase reward given a detailed human instruction describing a desired product. The RL agents are developed by fine-tuning a pre-trained BERT model with various objectives, learning from preferences without a reward model, and employing contemporary training techniques such as Proximal Policy Optimization (PPO) as used in InstructGPT, and Direct Preference Optimization (DPO). This report also evaluates the RL agents trained using generative trajectories. Evaluations were conducted using Thompson sampling in the WebShop simulator environment. The simulated online experiments demonstrate that agents trained on generated trajectories exhibited comparable task performance to those trained using human trajectories. This has demonstrated an example of an extremely low-cost data-efficient way of training reinforcement learning agents. Also, with limited training time (<2hours), without utilizing any images, a DPO agent achieved a 19% success rate after approximately 3000 steps or 30 minutes of training on T4 GPUs, compared to a PPO agent, which reached a 15% success rate.

Automatic disease diagnosis has become increasingly valuable in clinical practice. The advent of large language models (LLMs) has catalyzed a paradigm shift in artificial intelligence, with growing evidence supporting the efficacy of LLMs in diagnostic tasks. Despite the increasing attention in this field, a holistic view is still lacking. Many critical aspects remain unclear, such as the diseases and clinical data to which LLMs have been applied, the LLM techniques employed, and the evaluation methods used. In this article, we perform a comprehensive review of LLM-based methods for disease diagnosis. Our review examines the existing literature across various dimensions, including disease types and associated clinical specialties, clinical data, LLM techniques, and evaluation methods. Additionally, we offer recommendations for applying and evaluating LLMs for diagnostic tasks. Furthermore, we assess the limitations of current research and discuss future directions. To our knowledge, this is the first comprehensive review for LLM-based disease diagnosis., Comment: 69 pages

We explore the post-Newtonian dynamics of spinning black hole (BH) binaries in Einstein-ScalarGauss-Bonnet (ESGB) gravity, a theory that modifies general relativity by introducing a massless scalar field coupled nonminimally to gravity via the Gauss-Bonnet term. By employing an effective field theory (EFT) approach, we extend the Routhian formalism to incorporate spin effects in scalar interactions. In this formalism, we derive for the first time the effective two-body potential for spinning BHs up to the third post-Newtonian (3PN) order in generic ESGB gravity theories. This potential is expressed in terms of the sensitivities of the BHs, which are then obtained through a matching procedure using analytic BH solutions derived here within a slow-rotation approximation, accurate to seventh order in the Gauss-Bonnet coupling and fifth order in the BH spin. We also examine the thermodynamic properties of these rotating solutions, which, as shown in previous work, yield important insights into the inspiral phase evolution in ESGB gravity., Comment: 25 pages, 8 figures. Contains ancillary file

Federated Learning (FL) is a rising approach towards collaborative and privacy-preserving machine learning where large-scale medical datasets remain localized to each client. However, the issue of data heterogeneity among clients often compels local models to diverge, leading to suboptimal global models. To mitigate the impact of data heterogeneity on FL performance, we start with analyzing how FL training influence FL performance by decomposing the global loss into three terms: local loss, distribution shift loss and aggregation loss. Remarkably, our loss decomposition reveals that existing local training-based FL methods attempt to reduce the distribution shift loss, while the global aggregation-based FL methods propose better aggregation strategies to reduce the aggregation loss. Nevertheless, a comprehensive joint effort to minimize all three terms is currently limited in the literature, leading to subpar performance when dealing with data heterogeneity challenges. To fill this gap, we propose a novel FL method based on global loss decomposition, called FedLD, to jointly reduce these three loss terms. Our FedLD involves a margin control regularization in local training to reduce the distribution shift loss, and a principal gradient-based server aggregation strategy to reduce the aggregation loss. Notably, under different levels of data heterogeneity, our strategies achieve better and more robust performance on retinal and chest X-ray classification compared to other FL algorithms. Our code is available at https://github.com/Zeng-Shuang/FedLD., Comment: Modify some typos (MICCAI 2024)

Legal Judgment Prediction (LJP) aims to form legal judgments based on the criminal fact description. However, researchers struggle to classify confusing criminal cases, such as robbery and theft, which requires LJP models to distinguish the nuances between similar crimes. Existing methods usually design handcrafted features to pick up necessary semantic legal clues to make more accurate legal judgment predictions. In this paper, we propose a Semantic-Aware Dual Encoder Model (SEMDR), which designs a novel legal clue tracing mechanism to conduct fine-grained semantic reasoning between criminal facts and instruments. Our legal clue tracing mechanism is built from three reasoning levels: 1) Lexicon-Tracing, which aims to extract criminal facts from criminal descriptions; 2) Sentence Representation Learning, which contrastively trains language models to better represent confusing criminal facts; 3) Multi-Fact Reasoning, which builds a reasons graph to propagate semantic clues among fact nodes to capture the subtle difference among criminal facts. Our legal clue tracing mechanism helps SEMDR achieve state-of-the-art on the CAIL2018 dataset and shows its advance in few-shot scenarios. Our experiments show that SEMDR has a strong ability to learn more uniform and distinguished representations for criminal facts, which helps to make more accurate predictions on confusing criminal cases and reduces the model uncertainty during making judgments. All codes will be released via GitHub.

We present a particle tracking method for unstructured meshes in two and three dimensions. Our tracking algorithm is based on a patch searching and includes two steps. We first locate the given point to a patch near a vertex, and then the host element is determined within the patch domain. We prove that the patch in the first step can be rapidly identified by constructing an auxiliary Cartesian grid with a prescribed resolution. Then, the second step can be converted into a searching problem, which can be easily solved by searching algorithms. Only coordinates to particles are required in our method. We conduct a series of numerical tests in two and three dimensions to illustrate the robustness and efficiency in our method.

The interplay of topology, magnetism, and correlations gives rise to intriguing phases of matter. In this study, through state-of-the-art angle-resolved photoemission spectroscopy, density functional theory and dynamical mean-field theory calculations, we visualize a fourfold degenerate Dirac nodal line at the boundary of the bulk Brillouin zone in the antiferromagnet YMn2Ge2. We further demonstrate that this gapless, antiferromagnetic Dirac nodal line is enforced by the combination of magnetism, space-time inversion symmetry and nonsymmorphic lattice symmetry. The corresponding drumhead surface states traverse the whole surface Brillouin zone. YMn2Ge2 thus serves as a platform to exhibit the interplay of multiple degenerate nodal physics and antiferromagnetism. Interestingly, the magnetic nodal line displays a d-orbital dependent renormalization along its trajectory in momentum space, thereby manifesting Hund coupling. Our findings offer insights into the effect of electronic correlations on magnetic Dirac nodal lines, leading to an antiferromagnetic Hund nodal line.

$\pi$ modes are unique topological edge states appearing in Floquet systems with periodic modulations of the underlying lattice structure in evolution variable, such as dynamically modulated Su-Schrieffer-Heeger (SSH) lattices. These edge states are anomalous states usually appearing between Floquet replicas of the same band, even if standard topological index remains zero for this band. While linear and nonlinear $\pi$ modes were observed in conservative systems, they have never been studied in nonlinear regime in the non-Hermitian systems with structured gain and losses. Here we show that SSH waveguide array with periodically oscillating waveguide positions in propagation direction and with parity-time symmetric refractive index landscape, can support $\pi$ modes that are damped or amplified at different ends of the array. By including nonlinearity and nonlinear absorption into our continuous system, we achieve stable lasing in $\pi$ mode at one end of the array. The representative feature of this system is that lasing in it is thresholdless and it occurs even at low gain-loss amplitudes. The degree of localization of lasing $\pi$ modes can be flexibly controlled by the amplitude of transverse waveguide oscillations. This work therefore introduces a new type of topological Floquet laser and a route to manipulation of $\pi$ modes by structured gain and losses., Comment: 9 pages, 3 figures, to appear in APL Photonics

Generative Flow Networks (GFlowNets) aim to generate diverse trajectories from a distribution in which the final states of the trajectories are proportional to the reward, serving as a powerful alternative to reinforcement learning for exploratory control tasks. However, the individual-flow matching constraint in GFlowNets limits their applications for multi-agent systems, especially continuous joint-control problems. In this paper, we propose a novel Multi-Agent generative Continuous Flow Networks (MACFN) method to enable multiple agents to perform cooperative exploration for various compositional continuous objects. Technically, MACFN trains decentralized individual-flow-based policies in a centralized global-flow-based matching fashion. During centralized training, MACFN introduces a continuous flow decomposition network to deduce the flow contributions of each agent in the presence of only global rewards. Then agents can deliver actions solely based on their assigned local flow in a decentralized way, forming a joint policy distribution proportional to the rewards. To guarantee the expressiveness of continuous flow decomposition, we theoretically derive a consistency condition on the decomposition network. Experimental results demonstrate that the proposed method yields results superior to the state-of-the-art counterparts and better exploration capability. Our code is available at https://github.com/isluoshuang/MACFN.

Glossy objects present a significant challenge for 3D reconstruction from multi-view input images under natural lighting. In this paper, we introduce PBIR-NIE, an inverse rendering framework designed to holistically capture the geometry, material attributes, and surrounding illumination of such objects. We propose a novel parallax-aware non-distant environment map as a lightweight and efficient lighting representation, accurately modeling the near-field background of the scene, which is commonly encountered in real-world capture setups. This feature allows our framework to accommodate complex parallax effects beyond the capabilities of standard infinite-distance environment maps. Our method optimizes an underlying signed distance field (SDF) through physics-based differentiable rendering, seamlessly connecting surface gradients between a triangle mesh and the SDF via neural implicit evolution (NIE). To address the intricacies of highly glossy BRDFs in differentiable rendering, we integrate the antithetic sampling algorithm to mitigate variance in the Monte Carlo gradient estimator. Consequently, our framework exhibits robust capabilities in handling glossy object reconstruction, showcasing superior quality in geometry, relighting, and material estimation.

The scaling of large language models (LLMs) has revolutionized their capabilities in various tasks, yet this growth must be matched with efficient computational strategies. The Mixture-of-Experts (MoE) architecture stands out for its ability to scale model size without significantly increasing training costs. Despite their advantages, current MoE models often display parameter inefficiency. For instance, a pre-trained MoE-based LLM with 52 billion parameters might perform comparably to a standard model with 6.7 billion parameters. Being a crucial part of MoE, current routers in different layers independently assign tokens without leveraging historical routing information, potentially leading to suboptimal token-expert combinations and the parameter inefficiency problem. To alleviate this issue, we introduce the Layerwise Recurrent Router for Mixture-of-Experts (RMoE). RMoE leverages a Gated Recurrent Unit (GRU) to establish dependencies between routing decisions across consecutive layers. Such layerwise recurrence can be efficiently parallelly computed for input tokens and introduces negotiable costs. Our extensive empirical evaluations demonstrate that RMoE-based language models consistently outperform a spectrum of baseline models. Furthermore, RMoE integrates a novel computation stage orthogonal to existing methods, allowing seamless compatibility with other MoE architectures. Our analyses attribute RMoE's gains to its effective cross-layer information sharing, which also improves expert selection and diversity. Our code is at https://github.com/qiuzh20/RMoE

Kuaishou, is one of the largest short-video and live-streaming platform, compared with short-video recommendations, live-streaming recommendation is more complex because of: (1) temporarily-alive to distribution, (2) user may watch for a long time with feedback delay, (3) content is unpredictable and changes over time. Actually, even if a user is interested in the live-streaming author, it still may be an negative watching (e.g., short-view < 3s) since the real-time content is not attractive enough. Therefore, for live-streaming recommendation, there exists a challenging task: how do we recommend the live-streaming at right moment for users? Additionally, our platform's major exposure content is short short-video, and the amount of exposed short-video is 9x more than exposed live-streaming. Thus users will leave more behaviors on short-videos, which leads to a serious data imbalance problem making the live-streaming data could not fully reflect user interests. In such case, there raises another challenging task: how do we utilize users' short-video behaviors to make live-streaming recommendation better?, Comment: Work in progress