Search

The film Her features Samantha, a sophisticated AI audio agent who is capable of understanding both linguistic and paralinguistic information in human speech and delivering real-time responses that are natural, informative and sensitive to emotional subtleties. Moving one step toward more sophisticated audio agent from recent advancement in end-to-end (E2E) speech systems, we propose LUCY, a E2E speech model that (1) senses and responds to user's emotion, (2) deliver responses in a succinct and natural style, and (3) use external tool to answer real-time inquiries. Experiment results show that LUCY is better at emotion control than peer models, generating emotional responses based on linguistic emotional instructions and responding to paralinguistic emotional cues. Lucy is also able to generate responses in a more natural style, as judged by external language models, without sacrificing much performance on general question answering. Finally, LUCY can leverage function calls to answer questions that are out of its knowledge scope., Comment: Demo Link: https://github.com/VITA-MLLM/LUCY

Benchmarks are important tools for tracking the rapid advancements in large language model (LLM) capabilities. However, benchmarks are not keeping pace in difficulty: LLMs now achieve over 90\% accuracy on popular benchmarks like MMLU, limiting informed measurement of state-of-the-art LLM capabilities. In response, we introduce Humanity's Last Exam (HLE), a multi-modal benchmark at the frontier of human knowledge, designed to be the final closed-ended academic benchmark of its kind with broad subject coverage. HLE consists of 3,000 questions across dozens of subjects, including mathematics, humanities, and the natural sciences. HLE is developed globally by subject-matter experts and consists of multiple-choice and short-answer questions suitable for automated grading. Each question has a known solution that is unambiguous and easily verifiable, but cannot be quickly answered via internet retrieval. State-of-the-art LLMs demonstrate low accuracy and calibration on HLE, highlighting a significant gap between current LLM capabilities and the expert human frontier on closed-ended academic questions. To inform research and policymaking upon a clear understanding of model capabilities, we publicly release HLE at https://lastexam.ai., Comment: 26 pages, 6 figures

Current research in audio deepfake detection is gradually transitioning from binary classification to multi-class tasks, referred as audio deepfake source tracing task. However, existing studies on source tracing consider only closed-set scenarios and have not considered the challenges posed by open-set conditions. In this paper, we define the Neural Codec Source Tracing (NCST) task, which is capable of performing open-set neural codec classification and interpretable ALM detection. Specifically, we constructed the ST-Codecfake dataset for the NCST task, which includes bilingual audio samples generated by 11 state-of-the-art neural codec methods and ALM-based out-ofdistribution (OOD) test samples. Furthermore, we establish a comprehensive source tracing benchmark to assess NCST models in open-set conditions. The experimental results reveal that although the NCST models perform well in in-distribution (ID) classification and OOD detection, they lack robustness in classifying unseen real audio. The ST-codecfake dataset and code are available.

Recent Multimodal Large Language Models (MLLMs) have typically focused on integrating visual and textual modalities, with less emphasis placed on the role of speech in enhancing interaction. However, speech plays a crucial role in multimodal dialogue systems, and implementing high-performance in both vision and speech tasks remains a significant challenge due to the fundamental modality differences. In this paper, we propose a carefully designed multi-stage training methodology that progressively trains LLM to understand both visual and speech information, ultimately enabling fluent vision and speech interaction. Our approach not only preserves strong vision-language capacity, but also enables efficient speech-to-speech dialogue capabilities without separate ASR and TTS modules, significantly accelerating multimodal end-to-end response speed. By comparing our method against state-of-the-art counterparts across benchmarks for image, video, and speech tasks, we demonstrate that our model is equipped with both strong visual and speech capabilities, making near real-time vision and speech interaction., Comment: https://github.com/VITA-MLLM/VITA (2K+ Stars by now)

Evaluating the quality of synthesized images remains a significant challenge in the development of text-to-image (T2I) generation. Most existing studies in this area primarily focus on evaluating text-image alignment, image quality, and object composition capabilities, with comparatively fewer studies addressing the evaluation of the factuality of T2I models, particularly when the concepts involved are knowledge-intensive. To mitigate this gap, we present T2I-FactualBench in this work - the largest benchmark to date in terms of the number of concepts and prompts specifically designed to evaluate the factuality of knowledge-intensive concept generation. T2I-FactualBench consists of a three-tiered knowledge-intensive text-to-image generation framework, ranging from the basic memorization of individual knowledge concepts to the more complex composition of multiple knowledge concepts. We further introduce a multi-round visual question answering (VQA) based evaluation framework to assess the factuality of three-tiered knowledge-intensive text-to-image generation tasks. Experiments on T2I-FactualBench indicate that current state-of-the-art (SOTA) T2I models still leave significant room for improvement.

Asian Americans as a group are overrepresented among STEM college graduates and have the highest average college enrollment rate of any racial or ethnic category. Thus, Asian Americans are typically excluded from educational interventions directed at improving STEM education for Students of Color because they are not considered to be underrepresented minorities. However, statistics obscure the individual needs of the more than 20 ethnic subgroups that fall under the umbrella term Asian Americans. Using a participatory action research approach, this paper documents the institutional and sociocultural factors that push out HMoob (or Hmong) American college students from STEM programs at one large, predominantly White university; and the coordinate processes of gatekeeping and transactional advising that either redirect those students toward non-STEM programs or force them out of the university completely.

There has been significant interest in exploring topological disclination states, which effectively probe the band topology of the host material beyond the conventional bulk-edge correspondence. While most studies in this area have primarily focused on Hermitian systems, recent theoretical work predicts that non-Hermiticity can drive topological phase transitions and host topological disclination states associated with fractional charge. However, no experimental observations have been reported to date. Here, we report the first experimental observation of topological disclination states in electric circuits, induced solely by gain and loss. Through admittance matrix measurements and eigenstate analysis, we confirm their emergence and compute the corresponding fractional charge. Moreover, the disclination mode profile and localization effect can be directly visualized via monochromatic field excitation. Additionally, we demonstrate the emergence of degenerate zero-energy topological disclination states, devoid of fractional charge, in distinct non-Hermitian geometries. Our findings open the possibility of non-Hermiticity-induced fractional charges in two-dimensional non-Hermitian lattices, which may pave the way for advancements in active topological photonic devices., Comment: 16 pages, 4 figures

The core-collapse supernova remnant (SNR) Cassiopeia A (Cas A) is one of the brightest galactic radio sources with an angular radius of $\sim$ 2.5 $\arcmin$. Although no extension of this source has been detected in the $\gamma$-ray band, using more than 1000 days of LHAASO data above $\sim 0.8$ TeV, we find that its spectrum is significantly softer than those obtained with Imaging Air Cherenkov Telescopes (IACTs) and its flux near $\sim 1$ TeV is about two times higher. In combination with analyses of more than 16 years of \textit{Fermi}-LAT data covering $0.1 \, \mathrm{GeV} - 1 \, \mathrm{TeV}$, we find that the spectrum above 30 GeV deviates significantly from a single power-law, and is best described by a smoothly broken power-law with a spectral index of $1.90 \pm 0.15_\mathrm{stat}$ ($3.41 \pm 0.19_\mathrm{stat}$) below (above) a break energy of $0.63 \pm 0.21_\mathrm{stat} \, \mathrm{TeV}$. Given differences in the angular resolution of LHAASO-WCDA and IACTs, TeV $\gamma$-ray emission detected with LHAASO may have a significant contribution from regions surrounding the SNR illuminated by particles accelerated earlier, which, however, are treated as background by IACTs. Detailed modelling can be used to constrain acceleration processes of TeV particles in the early stage of SNR evolution.

Due to the effective multi-scale feature fusion capabilities of the Path Aggregation FPN (PAFPN), it has become a widely adopted component in YOLO-based detectors. However, PAFPN struggles to integrate high-level semantic cues with low-level spatial details, limiting its performance in real-world applications, especially with significant scale variations. In this paper, we propose MHAF-YOLO, a novel detection framework featuring a versatile neck design called the Multi-Branch Auxiliary FPN (MAFPN), which consists of two key modules: the Superficial Assisted Fusion (SAF) and Advanced Assisted Fusion (AAF). The SAF bridges the backbone and the neck by fusing shallow features, effectively transferring crucial low-level spatial information with high fidelity. Meanwhile, the AAF integrates multi-scale feature information at deeper neck layers, delivering richer gradient information to the output layer and further enhancing the model learning capacity. To complement MAFPN, we introduce the Global Heterogeneous Flexible Kernel Selection (GHFKS) mechanism and the Reparameterized Heterogeneous Multi-Scale (RepHMS) module to enhance feature fusion. RepHMS is globally integrated into the network, utilizing GHFKS to select larger convolutional kernels for various feature layers, expanding the vertical receptive field and capturing contextual information across spatial hierarchies. Locally, it optimizes convolution by processing both large and small kernels within the same layer, broadening the lateral receptive field and preserving crucial details for detecting smaller targets. The source code of this work is available at: https://github.com/yang0201/MHAF-YOLO., Comment: arXiv admin note: text overlap with arXiv:2407.04381

Spiking Large Language Models have been shown as a good alternative to LLMs in various scenarios. Existing methods for creating Spiking LLMs, i.e., direct training and ANN-SNN conversion, often suffer from performance degradation and relatively high computational costs. To address these issues, we propose a novel Fast ANN-SNN conversion strategy (FAS) that transforms LLMs into spiking LLMs in two stages. The first stage employs a full-parameter fine-tuning of pre-trained models, so it does not need any direct training from scratch. The second stage introduces a coarse-to-fine calibration method to reduce conversion errors and improve accuracy. Our experiments on both language and vision-language tasks across four different scales of LLMs demonstrate that FAS can achieve state-of-the-art performance yet with significantly reduced inference latency and computational costs. For example, FAS only takes 8 timesteps to achieve an accuracy of 3% higher than that of the OPT-7B model, while reducing energy consumption by 96.63%.

This paper presents a method that allows users to design cinematic video shots in the context of image-to-video generation. Shot design, a critical aspect of filmmaking, involves meticulously planning both camera movements and object motions in a scene. However, enabling intuitive shot design in modern image-to-video generation systems presents two main challenges: first, effectively capturing user intentions on the motion design, where both camera movements and scene-space object motions must be specified jointly; and second, representing motion information that can be effectively utilized by a video diffusion model to synthesize the image animations. To address these challenges, we introduce MotionCanvas, a method that integrates user-driven controls into image-to-video (I2V) generation models, allowing users to control both object and camera motions in a scene-aware manner. By connecting insights from classical computer graphics and contemporary video generation techniques, we demonstrate the ability to achieve 3D-aware motion control in I2V synthesis without requiring costly 3D-related training data. MotionCanvas enables users to intuitively depict scene-space motion intentions, and translates them into spatiotemporal motion-conditioning signals for video diffusion models. We demonstrate the effectiveness of our method on a wide range of real-world image content and shot-design scenarios, highlighting its potential to enhance the creative workflows in digital content creation and adapt to various image and video editing applications., Comment: It is best viewed in Acrobat. Project page: https://motion-canvas25.github.io/

The XENONnT experiment has achieved an unprecedented reduction of the $^\text{222}$Rn activity concentration within its liquid xenon dual-phase time projection chamber to a level of (0.90$\,\pm\,$0.01$\,$stat.$\,\pm\,$0.07 sys.)$\,\mu$Bq/kg, equivalent to about 1200 $^\text{222}$Rn atoms per cubic meter of liquid xenon. This represents a 15-fold improvement over the $^\text{222}$Rn levels encountered during XENON1T's main science runs and is a factor five lower compared to other currently operational multi-tonne liquid xenon detectors engaged in dark matter searches. This breakthrough enables the pursuit of various rare event searches that lie beyond the confines of the standard model of particle physics, with world-leading sensitivity. The ultra-low $^\text{222}$Rn levels have diminished the radon-induced background rate in the detector to a point where it is for the first time lower than the solar neutrino-induced background, which is poised to become the primary irreducible background in liquid xenon-based detectors.

We investigate a generalized antiferromagnetic cluster XY model in a transverse magnetic field, where long-range interactions decay algebraically with distance. This model can be exactly solvable within a free fermion framework. By analyzing the gap, we explicitly derive the critical exponents $\nu$ and $z$, finding that the relationship $\nu z = 1$ still holds. However, the values of $\nu$ and $z$ depend on the decaying exponent $\alpha$, in contrast to those for the quantum long-range antiferromagnetic Ising chain. To optimize scaling behavior, we verify these critical exponents using correlation functions and fidelity susceptibility, achieving excellent data collapse across various system sizes by adjusting fitting parameters. Finally, we compute the entanglement entropy at the critical point to determine the central charge $c$, and find it also varies with $\alpha$. This study provides insights into the unique effect of long-range cluster interactions on the critical properties of quantum spin systems., Comment: 8 pages, 7 figures

Current evaluations of synthetic tabular data mainly focus on how well joint distributions are modeled, often overlooking the assessment of their effectiveness in preserving realistic event sequences and coherent entity relationships across columns.This paper proposes three evaluation metrics designed to assess the preservation of logical relationships among columns in synthetic tabular data. We validate these metrics by assessing the performance of both classical and state-of-the-art generation methods on a real-world industrial dataset.Experimental results reveal that existing methods often fail to rigorously maintain logical consistency (e.g., hierarchical relationships in geography or organization) and dependencies (e.g., temporal sequences or mathematical relationships), which are crucial for preserving the fine-grained realism of real-world tabular data. Building on these insights, this study also discusses possible pathways to better capture logical relationships while modeling the distribution of synthetic tabular data.

The first measurement of the $CP$ asymmetry of the decay rate ($A_{CP}$) and the $CP$ average ($\Sigma A_{\text{FB}}$) and $CP$ asymmetry ($\Delta A_{\text{FB}}$) of the forward-backward asymmetry in the muon system of $\mathit{\Lambda}^+_c\to p\mu^+\mu^-$ decays is reported. The measurement is performed using a data sample of proton-proton collisions, recorded by the LHCb experiment from 2016 to 2018 at a center-of-mass energy of 13$\text{ TeV}$, which corresponds to an integrated luminosity of 5.4$\text{ fb}^{-1}$. The asymmetries are measured in two regions of dimuon mass near the $\phi$-meson mass peak. The dimuon-mass integrated results are \begin{align*} A_{CP} &= (-1.1 \pm 4.0 \pm 0.5)\%,\\ \Sigma A_{\text{FB}} &= (\phantom{-}3.9 \pm 4.0 \pm 0.6)\%,\\ \Delta A_{\text{FB}} &= (\phantom{-}3.1 \pm 4.0 \pm 0.4)\%, \end{align*} where the first uncertainty is statistical and the second systematic. The results are consistent with the conservation of $CP$ symmetry and the Standard Model expectations., Comment: All figures and tables, along with machine-readable versions and any supplementary material and additional information, are available at https://lbfence.cern.ch/alcm/public/analysis/full-details/3473/ (LHCb public pages)

Dual-comb spectroscopy has demonstrated remarkable capabilities for rapid and sensitive measurements; however, significant challenges still exist in generating high-power, mutually coherent mid-infrared combs. Here we demonstrate that a pair of near-infrared femtosecond frequency combs can be spectrally translated via a continuous-wave optical parametric oscillator. The pair of spectrally translated combs demonstrated high mutual coherence, power per comb tooth in excess of hundreds of microwatts, and were tunable between 4 um and 5 um. Unlike previous approaches which relied upon synchronous optical parametric oscillation, the present approach avoids challenges associated with comb stabilization, low power per comb tooth, and complex cavity designs. Further it is readily amenable to high repetition rates (gigahertz-level and beyond). The flexible and facile nature of this approach provides a robust path for the spectral translation of mode-locked combs, achieving spectral bandwidths limited only by the phase matching bandwidth of the optical parametric oscillator. This approach holds significant promise for applications in chemical kinetics, remote sensing, combustion science, and precision spectroscopy, where the combination of high powers, broad bandwidths, and high measurement rates are transformative., Comment: 14 pages, 5 figures, 4 extended data figures

Today's open vocabulary scene graph generation (OVSGG) extends traditional SGG by recognizing novel objects and relationships beyond predefined categories, leveraging the knowledge from pre-trained large-scale models. Most existing methods adopt a two-stage pipeline: weakly supervised pre-training with image captions and supervised fine-tuning (SFT) on fully annotated scene graphs. Nonetheless, they omit explicit modeling of interacting objects and treat all objects equally, resulting in mismatched relation pairs. To this end, we propose an interaction-aware OVSGG framework INOVA. During pre-training, INOVA employs an interaction-aware target generation strategy to distinguish interacting objects from non-interacting ones. In SFT, INOVA devises an interaction-guided query selection tactic to prioritize interacting objects during bipartite graph matching. Besides, INOVA is equipped with an interaction-consistent knowledge distillation to enhance the robustness by pushing interacting object pairs away from the background. Extensive experiments on two benchmarks (VG and GQA) show that INOVA achieves state-of-the-art performance, demonstrating the potential of interaction-aware mechanisms for real-world applications.

Autonomous driving technology has advanced significantly, yet detecting driving anomalies remains a major challenge due to the long-tailed distribution of driving events. Existing methods primarily rely on single-modal road condition video data, which limits their ability to capture rare and unpredictable driving incidents. This paper proposes a multimodal driver assistance detection system that integrates road condition video, driver facial video, and audio data to enhance incident recognition accuracy. Our model employs an attention-based intermediate fusion strategy, enabling end-to-end learning without separate feature extraction. To support this approach, we develop a new three-modality dataset using a driving simulator. Experimental results demonstrate that our method effectively captures cross-modal correlations, reducing misjudgments and improving driving safety.

Microcombs, formed in optical microresonators driven by continuous-wave lasers, are miniaturized optical frequency combs with small size, weight and power consumption. Leveraging integrated photonics and laser self-injection locking (SIL), compact and robust microcombs can be constructed via hybrid integration of a semiconductor laser with a chip-based microresonator. While the current linear SIL theory has successfully addressed the linear coupling between the laser cavity and the external microresonator, it fails to manage the complicated nonlinear processes, especially regarding to dark-pulse microcomb formation. Here, we investigate -- theoretically, numerically and experimentally -- the Kerr-thermal dynamics of a semiconductor laser self-injection-locked to an integrated silicon nitride microresonator. We unveil intriguing yet universal dark-pulse formation and switching behaviour with discrete steps, and establish a theoretical model scrutinizing the synergy of laser-microresonator mutual coupling, Kerr nonlinearity, photo-thermal effect. Numerical simulation confirms the experimental result and identifies the origins. Exploiting this unique phenomenon, we showcase an application on low-noise photonic microwave generation with phase noise purified by 23.5 dB. Our study not only add critical insight of pulse formation in laser-microresonator hybrid systems, but also enables all-passive, photonic-chip-based microwave oscillators with high spectral purity., Comment: 8 pages, 3 figures

Universal Machine Learning Interactomic Potentials (MLIPs) enable accelerated simulations for materials discovery. However, current research efforts fail to impactfully utilize MLIPs due to: 1. Overreliance on Density Functional Theory (DFT) for MLIP training data creation; 2. MLIPs' inability to reliably and accurately perform large-scale molecular dynamics (MD) simulations for diverse materials; 3. Limited understanding of MLIPs' underlying capabilities. To address these shortcomings, we aargue that MLIP research efforts should prioritize: 1. Employing more accurate simulation methods for large-scale MLIP training data creation (e.g. Coupled Cluster Theory) that cover a wide range of materials design spaces; 2. Creating MLIP metrology tools that leverage large-scale benchmarking, visualization, and interpretability analyses to provide a deeper understanding of MLIPs' inner workings; 3. Developing computationally efficient MLIPs to execute MD simulations that accurately model a broad set of materials properties. Together, these interdisciplinary research directions can help further the real-world application of MLIPs to accurately model complex materials at device scale.

Generating novel crystalline materials has potential to lead to advancements in fields such as electronics, energy storage, and catalysis. The defining characteristic of crystals is their symmetry, which plays a central role in determining their physical properties. However, existing crystal generation methods either fail to generate materials that display the symmetries of real-world crystals, or simply replicate the symmetry information from examples in a database. To address this limitation, we propose SymmCD, a novel diffusion-based generative model that explicitly incorporates crystallographic symmetry into the generative process. We decompose crystals into two components and learn their joint distribution through diffusion: 1) the asymmetric unit, the smallest subset of the crystal which can generate the whole crystal through symmetry transformations, and; 2) the symmetry transformations needed to be applied to each atom in the asymmetric unit. We also use a novel and interpretable representation for these transformations, enabling generalization across different crystallographic symmetry groups. We showcase the competitive performance of SymmCD on a subset of the Materials Project, obtaining diverse and valid crystals with realistic symmetries and predicted properties.

Large Vision-Language Models (LVLMs) can reason effectively over both textual and visual inputs, but they tend to hallucinate syntactically coherent yet visually ungrounded contents. In this paper, we investigate the internal dynamics of hallucination by examining the tokens logits rankings throughout the generation process, revealing three key patterns in how LVLMs process information: (1) gradual visual information loss -- visually grounded tokens gradually become less favored throughout generation, and (2) early excitation -- semantically meaningful tokens achieve peak activation in the layers earlier than the final layer. (3) hidden genuine information -- visually grounded tokens though not being eventually decided still retain relatively high rankings at inference. Based on these insights, we propose VISTA (Visual Information Steering with Token-logit Augmentation), a training-free inference-time intervention framework that reduces hallucination while promoting genuine information. VISTA works by combining two complementary approaches: reinforcing visual information in activation space and leveraging early layer activations to promote semantically meaningful decoding. Compared to existing methods, VISTA requires no external supervision and is applicable to various decoding strategies. Extensive experiments show that VISTA on average reduces hallucination by abount 40% on evaluated open-ended generation task, and it consistently outperforms existing methods on four benchmarks across four architectures under three decoding strategies.

The application of deep learning for partial differential equation (PDE)-constrained control is gaining increasing attention. However, existing methods rarely consider safety requirements crucial in real-world applications. To address this limitation, we propose Safe Diffusion Models for PDE Control (SafeDiffCon), which introduce the uncertainty quantile as model uncertainty quantification to achieve optimal control under safety constraints through both post-training and inference phases. Firstly, our approach post-trains a pre-trained diffusion model to generate control sequences that better satisfy safety constraints while achieving improved control objectives via a reweighted diffusion loss, which incorporates the uncertainty quantile estimated using conformal prediction. Secondly, during inference, the diffusion model dynamically adjusts both its generation process and parameters through iterative guidance and fine-tuning, conditioned on control targets while simultaneously integrating the estimated uncertainty quantile. We evaluate SafeDiffCon on three control tasks: 1D Burgers' equation, 2D incompressible fluid, and controlled nuclear fusion problem. Results demonstrate that SafeDiffCon is the only method that satisfies all safety constraints, whereas other classical and deep learning baselines fail. Furthermore, while adhering to safety constraints, SafeDiffCon achieves the best control performance.

Recent advances in machine learning have opened new avenues for optimizing detector designs in high-energy physics, where the complex interplay of geometry, materials, and physics processes has traditionally posed a significant challenge. In this work, we introduce the $\textit{end-to-end}$ optimization framework AIDO that leverages a diffusion model as a surrogate for the full simulation and reconstruction chain, enabling gradient-based design exploration in both continuous and discrete parameter spaces. Although this framework is applicable to a broad range of detectors, we illustrate its power using the specific example of a sampling calorimeter, focusing on charged pions and photons as representative incident particles. Our results demonstrate that the diffusion model effectively captures critical performance metrics for calorimeter design, guiding the automatic search for layer arrangement and material composition that aligns with known calorimeter principles. The success of this proof-of-concept study provides a foundation for future applications of end-to-end optimization to more complex detector systems, offering a promising path toward systematically exploring the vast design space in next-generation experiments., Comment: 14 pages, 9 figures, submitted to MDPI particles

In this paper, several Bohr-type inequalities are generalized to the form with two parameters for the bounded analytic function. Most of the results are sharp.

In this article, some Bohr inequalities for analytical functions on the unit disk are generalized to the forms with two parameters. One of our results is sharp.

In this article, Bohr type inequalities for some complex valued harmonic functions defined on the unit disk are given. All the results are sharp.

Head computed tomography (CT) imaging is a widely-used imaging modality with multitudes of medical indications, particularly in assessing pathology of the brain, skull, and cerebrovascular system. It is commonly the first-line imaging in neurologic emergencies given its rapidity of image acquisition, safety, cost, and ubiquity. Deep learning models may facilitate detection of a wide range of diseases. However, the scarcity of high-quality labels and annotations, particularly among less common conditions, significantly hinders the development of powerful models. To address this challenge, we introduce FM-CT: a Foundation Model for Head CT for generalizable disease detection, trained using self-supervised learning. Our approach pre-trains a deep learning model on a large, diverse dataset of 361,663 non-contrast 3D head CT scans without the need for manual annotations, enabling the model to learn robust, generalizable features. To investigate the potential of self-supervised learning in head CT, we employed both discrimination with self-distillation and masked image modeling, and we construct our model in 3D rather than at the slice level (2D) to exploit the structure of head CT scans more comprehensively and efficiently. The model's downstream classification performance is evaluated using internal and three external datasets, encompassing both in-distribution (ID) and out-of-distribution (OOD) data. Our results demonstrate that the self-supervised foundation model significantly improves performance on downstream diagnostic tasks compared to models trained from scratch and previous 3D CT foundation models on scarce annotated datasets. This work highlights the effectiveness of self-supervised learning in medical imaging and sets a new benchmark for head CT image analysis in 3D, enabling broader use of artificial intelligence for head CT-based diagnosis., Comment: Under Review Preprint

Neural operators have shown great potential in surrogate modeling. However, training a well-performing neural operator typically requires a substantial amount of data, which can pose a major challenge in complex applications. In such scenarios, detailed physical knowledge can be unavailable or difficult to obtain, and collecting extensive data is often prohibitively expensive. To mitigate this challenge, we propose the Pseudo Physics-Informed Neural Operator (PPI-NO) framework. PPI-NO constructs a surrogate physics system for the target system using partial differential equations (PDEs) derived from simple, rudimentary physics principles, such as basic differential operators. This surrogate system is coupled with a neural operator model, using an alternating update and learning process to iteratively enhance the model's predictive power. While the physics derived via PPI-NO may not mirror the ground-truth underlying physical laws -- hence the term ``pseudo physics'' -- this approach significantly improves the accuracy of standard operator learning models in data-scarce scenarios, which is evidenced by extensive evaluations across five benchmark tasks and a fatigue modeling application.

The advent of 6G communication demands seamlessly integrated terminals operating above 100 GHz with low power consumption for human-centric applications. In this work, we report high-performance, flexible radio-frequency (RF) transistors based on aligned carbon nanotube (CNT) arrays, achieving, for the first time, as-measured current gain cutoff frequency ($f_{\mathrm{T}}$) and power gain cutoff frequency ($f_{\mathrm{max}}$) both exceeding 100 GHz. Electro-thermal co-design improves both heat dissipation and RF performance, despite the low thermal conductivity of the flexible substrate. The transistors deliver 0.947 mA/ $\mathrm{\mu}$m on-state current and 0.728 mS/ $\mathrm{\mu}$m transconductance. Peak extrinsic $f_{\mathrm{T}}$ and $f_{\mathrm{max}}$ reach 152 GHz and 102 GHz, with low power consumption of 199 mW/mm and 147 mW/mm, respectively, setting new performance records for flexible CNT-based RF transistors by nearly $100\times$, outperforming all other flexible RF devices. Additionally, flexible RF amplifiers achieve output power of 64 mW/mm and power gain of 11 dB in the K-band (18 GHz), marking a significant milestone in the development of flexible RF technologies for next-generation wireless communication systems., Comment: 34 pages, 15 figures

Although deep models have been widely explored in solving partial differential equations (PDEs), previous works are primarily limited to data only with up to tens of thousands of mesh points, far from the million-point scale required by industrial simulations that involve complex geometries. In the spirit of advancing neural PDE solvers to real industrial applications, we present Transolver++, a highly parallel and efficient neural solver that can accurately solve PDEs on million-scale geometries. Building upon previous advancements in solving PDEs by learning physical states via Transolver, Transolver++ is further equipped with an extremely optimized parallelism framework and a local adaptive mechanism to efficiently capture eidetic physical states from massive mesh points, successfully tackling the thorny challenges in computation and physics learning when scaling up input mesh size. Transolver++ increases the single-GPU input capacity to million-scale points for the first time and is capable of continuously scaling input size in linear complexity by increasing GPUs. Experimentally, Transolver++ yields 13% relative promotion across six standard PDE benchmarks and achieves over 20% performance gain in million-scale high-fidelity industrial simulations, whose sizes are 100$\times$ larger than previous benchmarks, covering car and 3D aircraft designs.

Heterogeneity in sensors and actuators across environments poses a significant challenge to building large-scale pre-trained world models on top of this low-dimensional sensor information. In this work, we explore pre-training world models for heterogeneous environments by addressing key transfer barriers in both data diversity and model flexibility. We introduce UniTraj, a unified dataset comprising over one million trajectories from 80 environments, designed to scale data while preserving critical diversity. Additionally, we propose TrajWorld, a novel architecture capable of flexibly handling varying sensor and actuator information and capturing environment dynamics in-context. Pre-training TrajWorld on UniTraj demonstrates significant improvements in transition prediction and achieves a new state-of-the-art for off-policy evaluation. To the best of our knowledge, this work, for the first time, demonstrates the transfer benefits of world models across heterogeneous and complex control environments.

Retrieval-Augmented Generation (RAG) has demonstrated remarkable success in enhancing Large Language Models (LLMs) through external knowledge integration, yet its application has primarily focused on textual content, leaving the rich domain of multi-modal video knowledge predominantly unexplored. This paper introduces VideoRAG, the first retrieval-augmented generation framework specifically designed for processing and understanding extremely long-context videos. Our core innovation lies in its dual-channel architecture that seamlessly integrates (i) graph-based textual knowledge grounding for capturing cross-video semantic relationships, and (ii) multi-modal context encoding for efficiently preserving visual features. This novel design empowers VideoRAG to process unlimited-length videos by constructing precise knowledge graphs that span multiple videos while maintaining semantic dependencies through specialized multi-modal retrieval paradigms. Through comprehensive empirical evaluation on our proposed LongerVideos benchmark-comprising over 160 videos totaling 134+ hours across lecture, documentary, and entertainment categories-VideoRAG demonstrates substantial performance compared to existing RAG alternatives and long video understanding methods. The source code of VideoRAG implementation and the benchmark dataset are openly available at: https://github.com/HKUDS/VideoRAG.

While Transformers have become the dominant architecture for visual generation, linear attention models, such as the state-space models (SSM), are increasingly recognized for their efficiency in processing long visual sequences. However, the essential efficiency of these models comes from formulating a limited recurrent state, enforcing causality among tokens that are prone to inconsistent modeling of N-dimensional visual data, leaving questions on their capacity to generate long non-causal sequences. In this paper, we explore the boundary of SSM on image and video generation by building the largest-scale diffusion SSM-Transformer hybrid model to date (5B parameters) based on the sub-quadratic bi-directional Hydra and self-attention, and generate up to 2K images and 360p 8 seconds (16 FPS) videos. Our results demonstrate that the model can produce faithful results aligned with complex text prompts and temporal consistent videos with high dynamics, suggesting the great potential of using SSMs for visual generation tasks., Comment: 21 pages, paper under review

Large language models (LLMs) for formal theorem proving have become a prominent research focus. At present, the proving ability of these LLMs is mainly evaluated through proof pass rates on datasets such as miniF2F. However, this evaluation method overlooks the varying importance of theorems. As a result, it fails to highlight the real performance disparities between LLMs and leads to high evaluation costs. This study proposes a psychometric-based evaluation method for theorem proving with LLMs, comprising two main components: Dataset Annotation and Adaptive Evaluation. First, we propose a metric calculation method to annotate the dataset with difficulty and discrimination metrics. Specifically, we annotate each theorem in the miniF2F dataset and grade them into varying difficulty levels according to the performance of LLMs, resulting in an enhanced dataset: miniF2F-Graded. Experimental results show that the difficulty grading in miniF2F-Graded better reflects the theorem difficulty perceived by LLMs. Secondly, we design an adaptive evaluation method to dynamically select the most suitable theorems for testing based on the annotated metrics and the real-time performance of LLMs. We apply this method to evaluate 10 LLMs. The results show that our method finely highlights the performance disparities between LLMs. It also reduces evaluation costs by using only 23% of the theorems in the dataset.

The advancement of large language models (LLMs) has opened new frontiers in natural language processing, particularly in specialized domains like healthcare. In this paper, we propose the Incremental Curriculum-Based Fine-Tuning (ICFT) framework to enhance the generative capabilities of medical large language models (MLLMs). ICFT combines curriculum-based learning, dual-stage memory coordination, and parameter-efficient fine-tuning to enable a progressive transition from general linguistic knowledge to strong domain-specific expertise. Experimental results across diverse medical NLP tasks, including question answering, preference classification, and response generation, demonstrate that ICFT consistently outperforms state-of-the-art baselines, achieving improvements in both accuracy and efficiency. Further analysis reveals the framework's ability to generalize to unseen data, reduce errors, and deliver diverse, contextually relevant medical responses. These findings establish ICFT as a robust and scalable solution for adapting LLMs to the medical domain, offering practical benefits for real-world healthcare applications.

We introduce Sundial, a family of native, flexible, and scalable time series foundation models. To predict the next-patch's distribution, we propose a TimeFlow Loss based on flow-matching, which facilitates native pre-training of Transformers on time series without discrete tokenization. Conditioned on arbitrary-length time series, our model is pre-trained without specifying any prior distribution and can generate multiple probable predictions, achieving flexibility in representation learning beyond using parametric densities. Towards time series foundation models, we leverage minimal but crucial adaptations of Transformers and curate TimeBench with 1 trillion time points, comprising mostly real-world datasets and synthetic data. By mitigating mode collapse through TimeFlow Loss, we pre-train a family of Sundial models on TimeBench, which exhibit unprecedented model capacity and generalization performance on zero-shot forecasting. In addition to presenting good scaling behavior, Sundial achieves new state-of-the-art on both point forecasting and probabilistic forecasting benchmarks. We believe that Sundial's pioneering generative paradigm will facilitate a wide variety of forecasting scenarios.

Physics-informed neural networks (PINNs) have earned high expectations in solving partial differential equations (PDEs), but their optimization usually faces thorny challenges due to the unique derivative-dependent loss function. By analyzing the loss distribution, previous research observed the propagation failure phenomenon of PINNs, intuitively described as the correct supervision for model outputs cannot ``propagate'' from initial states or boundaries to the interior domain. Going beyond intuitive understanding, this paper provides the first formal and in-depth study of propagation failure and its root cause. Based on a detailed comparison with classical finite element methods, we ascribe the failure to the conventional single-point-processing architecture of PINNs and further prove that propagation failure is essentially caused by the lower gradient correlation of PINN models on nearby collocation points. Compared to superficial loss maps, this new perspective provides a more precise quantitative criterion to identify where and why PINN fails. The theoretical finding also inspires us to present a new PINN architecture, named ProPINN, which can effectively unite the gradient of region points for better propagation. ProPINN can reliably resolve PINN failure modes and significantly surpass advanced Transformer-based models with 46% relative promotion.

Recent research on tool integration for math Large Language Models (LLMs) aims to combine complementary strengths of chain-of-thought (CoT) reasoning and code execution. However, we discover a critical limitation: current tool-integrated math LLMs rely on externally dictated instructions to decide whether to use CoT or code, lacking the autonomy to choose the most appropriate method independently. This prompts us to study \emph{Autonomous Code integration} for math LLMs, which enables models to \emph{independently} develop their own methodology-selection strategy in the absence of reliable supervision. To address this challenge, we propose an innovative Expectation-Maximization (EM) formulation that refines the model's decision-making through the exploration of its capabilities. This framework alternates between (a) computing a reference strategy that improves the model's belief over its capabilities through self-exploration, and (b) updating the model based on the refined belief. We further enhance this framework with an efficient implementation, incorporating a novel data synthesis strategy and off-policy reinforcement learning. Extensive experiments demonstrate that our approach, using only a public query set, significantly boosts the performance of existing math LLMs, raising accuracy by nearly 20\% to 65.28\% on the challenging MATH benchmark, while reducing code executions by up to 65\% .

Listing k-cliques plays a fundamental role in various data mining tasks, such as community detection and mining of cohesive substructures. Existing algorithms for the k-clique listing problem are built upon a general framework, which finds k-cliques by recursively finding (k-1)-cliques within subgraphs induced by the out-neighbors of each vertex. However, this framework has inherent inefficiency of finding smaller cliques within certain subgraphs repeatedly. In this paper, we propose an algorithm DIST for the k-clique listing problem. In contrast to existing works, the main idea in our approach is to compute each clique in the given graph only once and store it into a data structure called Induced Subgraph Trie, which allows us to retrieve the cliques efficiently. Furthermore, we propose a method to prune search space based on a novel concept called soft embedding of an l-tree, which further improves the running time. We show the superiority of our approach in terms of time and space usage through comprehensive experiments conducted on real-world networks; DIST outperforms the state-of-the-art algorithm by up to two orders of magnitude in both single-threaded and parallel experiments.

Image denoising of low-dose computed tomography (LDCT) is an important problem for clinical diagnosis with reduced radiation exposure. Previous methods are mostly trained with pairs of synthetic or misaligned LDCT and normal-dose CT (NDCT) images. However, trained with synthetic noise or misaligned LDCT/NDCT image pairs, the denoising networks would suffer from blurry structure or motion artifacts. Since non-contrast CT (NCCT) images share the content characteristics to the corresponding NDCT images in a three-phase scan, they can potentially provide useful information for real-world LDCT image denoising. To exploit this aspect, in this paper, we propose to incorporate clean NCCT images as useful guidance for the learning of real-world LDCT image denoising networks. To alleviate the issue of spatial misalignment in training data, we design a new Patch Triplet Similarity Purification (PTSP) strategy to select highly similar patch (instead of image) triplets of LDCT, NDCT, and NCCT images for network training. Furthermore, we modify two image denoising transformers of SwinIR and HAT to accommodate the NCCT image guidance, by replacing vanilla self-attention with cross-attention. On our collected clinical dataset, the modified transformers trained with the data selected by our PTSP strategy show better performance than 15 comparison methods on real-world LDCT image denoising. Ablation studies validate the effectiveness of our NCCT image guidance and PTSP strategy. We will publicly release our data and code.

Simultaneous Localization and Mapping (SLAM) is essential for precise surgical interventions and robotic tasks in minimally invasive procedures. While recent advancements in 3D Gaussian Splatting (3DGS) have improved SLAM with high-quality novel view synthesis and fast rendering, these systems struggle with accurate depth and surface reconstruction due to multi-view inconsistencies. Simply incorporating SLAM and 3DGS leads to mismatches between the reconstructed frames. In this work, we present Endo-2DTAM, a real-time endoscopic SLAM system with 2D Gaussian Splatting (2DGS) to address these challenges. Endo-2DTAM incorporates a surface normal-aware pipeline, which consists of tracking, mapping, and bundle adjustment modules for geometrically accurate reconstruction. Our robust tracking module combines point-to-point and point-to-plane distance metrics, while the mapping module utilizes normal consistency and depth distortion to enhance surface reconstruction quality. We also introduce a pose-consistent strategy for efficient and geometrically coherent keyframe sampling. Extensive experiments on public endoscopic datasets demonstrate that Endo-2DTAM achieves an RMSE of $1.87\pm 0.63$ mm for depth reconstruction of surgical scenes while maintaining computationally efficient tracking, high-quality visual appearance, and real-time rendering. Our code will be released at github.com/lastbasket/Endo-2DTAM., Comment: Accepted by ICRA 2025

The Great Outdoors (GO) dataset is a multi-modal annotated data resource aimed at advancing ground robotics research in unstructured environments. This dataset provides the most comprehensive set of data modalities and annotations compared to existing off-road datasets. In total, the GO dataset includes six unique sensor types with high-quality semantic annotations and GPS traces to support tasks such as semantic segmentation, object detection, and SLAM. The diverse environmental conditions represented in the dataset present significant real-world challenges that provide opportunities to develop more robust solutions to support the continued advancement of field robotics, autonomous exploration, and perception systems in natural environments. The dataset can be downloaded at: https://www.unmannedlab.org/the-great-outdoors-dataset/, Comment: 5 pages, 5 figures

In this work, we propose a method that leverages CLIP feature distillation, achieving efficient 3D segmentation through language guidance. Unlike previous methods that rely on multi-scale CLIP features and are limited by processing speed and storage requirements, our approach aims to streamline the workflow by directly and effectively distilling dense CLIP features, thereby achieving precise segmentation of 3D scenes using text. To achieve this, we introduce an adapter module and mitigate the noise issue in the dense CLIP feature distillation process through a self-cross-training strategy. Moreover, to enhance the accuracy of segmentation edges, this work presents a low-rank transient query attention mechanism. To ensure the consistency of segmentation for similar colors under different viewpoints, we convert the segmentation task into a classification task through label volume, which significantly improves the consistency of segmentation in color-similar areas. We also propose a simplified text augmentation strategy to alleviate the issue of ambiguity in the correspondence between CLIP features and text. Extensive experimental results show that our method surpasses current state-of-the-art technologies in both training speed and performance. Our code is available on: https://github.com/xingy038/Laser.git., Comment: Accepted by IEEE Transactions on Pattern Analysis and Machine Intelligence

This paper addresses the challenges in learning-based monocular positioning by proposing VKFPos, a novel approach that integrates Absolute Pose Regression (APR) and Relative Pose Regression (RPR) via an Extended Kalman Filter (EKF) within a variational Bayesian inference framework. Our method shows that the essential posterior probability of the monocular positioning problem can be decomposed into APR and RPR components. This decomposition is embedded in the deep learning model by predicting covariances in both APR and RPR branches, allowing them to account for associated uncertainties. These covariances enhance the loss functions and facilitate EKF integration. Experimental evaluations on both indoor and outdoor datasets show that the single-shot APR branch achieves accuracy on par with state-of-the-art methods. Furthermore, for temporal positioning, where consecutive images allow for RPR and EKF integration, VKFPos outperforms temporal APR and model-based integration methods, achieving superior accuracy.

Concept erasure has emerged as a promising technique for mitigating the risk of harmful content generation in diffusion models by selectively unlearning undesirable concepts. The common principle of previous works to remove a specific concept is to map it to a fixed generic concept, such as a neutral concept or just an empty text prompt. In this paper, we demonstrate that this fixed-target strategy is suboptimal, as it fails to account for the impact of erasing one concept on the others. To address this limitation, we model the concept space as a graph and empirically analyze the effects of erasing one concept on the remaining concepts. Our analysis uncovers intriguing geometric properties of the concept space, where the influence of erasing a concept is confined to a local region. Building on this insight, we propose the Adaptive Guided Erasure (AGE) method, which \emph{dynamically} selects optimal target concepts tailored to each undesirable concept, minimizing unintended side effects. Experimental results show that AGE significantly outperforms state-of-the-art erasure methods on preserving unrelated concepts while maintaining effective erasure performance. Our code is published at {https://github.com/tuananhbui89/Adaptive-Guided-Erasure}.

Test-time adaptation of pre-trained vision-language models has emerged as a technique for tackling distribution shifts during the test time. Although existing methods, especially those based on Test-time Prompt Tuning (TPT), have shown promising results, their high computational cost associated with parameter optimization presents challenges for scalability and practical application. This paper unveils the unnecessary nature of backpropagation in existing methods from a loss landscape perspective. Building on this insight, this paper proposes a simple yet effective framework called Test-time Loss Landscape Adaptation (TLLA). TLLA leverages the relative position between the training minimum and test loss landscapes to guide the adaptation process, avoiding the update of model parameters at test time. Specifically, it mainly consists of two main stages: In the prompt tuning stage, a Sharpness-Aware Prompt Tuning (SAPT) method is introduced to identify the training flat minimum, setting the foundation for the subsequent test-time adaptation; In the test stage, a Sharpness-based Test Sample Selection (STSS) approach is utilized to ensure the alignment of flat minima within the training loss landscape and each augmented test sample's loss landscape. Extensive experiments on both domain generalization and cross-dataset benchmarks demonstrate that TLLA achieves state-of-the-art performances while significantly reducing computational overhead. Notably, TLLA surpasses TPT by an average of 5.32\% and 6.98\% on four ImageNet variant datasets when employing ResNet50 and ViT-B/16 image encoders, respectively. The code will be available soon.

Low-rank Adaptation (LoRA) has demonstrated remarkable capabilities for task specific fine-tuning. However, in scenarios that involve multiple tasks, training a separate LoRA model for each one results in considerable inefficiency in terms of storage and inference. Moreover, existing parameter generation methods fail to capture the correlations among these tasks, making multi-task LoRA parameter generation challenging. To address these limitations, we propose In-Context Meta LoRA (ICM-LoRA), a novel approach that efficiently achieves task-specific customization of large language models (LLMs). Specifically, we use training data from all tasks to train a tailored generator, Conditional Variational Autoencoder (CVAE). CVAE takes task descriptions as inputs and produces task-aware LoRA weights as outputs. These LoRA weights are then merged with LLMs to create task-specialized models without the need for additional fine-tuning. Furthermore, we utilize in-context meta-learning for knowledge enhancement and task mapping, to capture the relationship between tasks and parameter distributions. As a result, our method achieves more accurate LoRA parameter generation for diverse tasks using CVAE. ICM-LoRA enables more accurate LoRA parameter reconstruction than current parameter reconstruction methods and is useful for implementing task-specific enhancements of LoRA parameters. At the same time, our method occupies 283MB, only 1\% storage compared with the original LoRA.

We present ultraviolet (UV) spectroscopic observations covering three distinct accretion states of the low-mass X-ray binary (LMXB) MAXI J1820+070: the luminous hard state, a hard-intermediate state and the soft state. Our observations were obtained during the 2018 eruption of MAXI J1820+070 with the Hubble Space Telescope (HST) and AstroSat observatory. The extinction towards the source turns out to be low - $\rm E_{B-V} = 0.2 \pm 0.05$ - making it one of the best UV accretion laboratories among LMXBs. Remarkably, we observe only moderate differences between all three states, with all spectra displaying similar continuum shapes and emission lines. Moreover, the continua are not well-described by physically plausible irradiated disc models. All of this challenges the standard reprocessing picture for UV emission from erupting LMXBs. The UV emission lines are double-peaked, with high-ionization lines displaying higher peak-to-peak velocities. None of the lines display obvious outflow signatures, even though blue-shifted absorption features have been seen in optical and near-infrared lines during the hard state. The emission line ratios are consistent with normal abundances, suggesting that the donor mass at birth was low enough to avoid CNO processing ($\rm M_{2,i} \lesssim 1.0 - 1.5 {\mathrm M_{\odot}}$). Finally, we study the evolution of UV variability in our time-resolved HST observations (hard and hard-intermediate states). All UV power spectra can be modelled with a broken power-law, superposed on which we tentatively detect the $\simeq 18$s quasi-periodic oscillation (QPO) that has been seen in other spectral bands., Comment: 19 pages, 14 figures, submitted to MNRAS