arXiv subjects

Liang Li

Publications and source records attributed to Liang Li.

At least 91 records · Page 5Linked to original sources

arXiv2026

STaR: Sensitive Trajectory Regulation for Unlearning in Large Reasoning Models

Large Reasoning Models (LRMs) have advanced automated multi-step reasoning, but their ability to generate complex Chain-of-Thought (CoT) trajectories introduces severe privacy risks, as sensitive information may be deeply embedded throughout the reasoning process. Existing Large Language Models (LLMs) unlearning approaches that typically focus on modifying only final answers are insufficient for LRMs, as they fail to remove sensitive content from intermediate steps, leading to persistent privacy leakage and degraded security. To address these challenges, we propose Sensitive Trajectory Regulation (STaR), a parameter-free, inference-time unlearning framework that achieves robust privacy protection throughout the reasoning process. Specifically, we first identify sensitive content via semantic-aware detection. Then, we inject global safety constraints through secure prompt prefix. Next, we perform trajectory-aware suppression to dynamically block sensitive content across the entire reasoning chain. Finally, we apply token-level adaptive filtering to prevent both exact and paraphrased sensitive tokens during generation. Furthermore, to overcome the inadequacies of existing evaluation protocols, we introduce two metrics: Multi-Decoding Consistency Assessment (MCS), which measures the consistency of unlearning across diverse decoding strategies, and Multi-Granularity Membership Inference Attack (MIA) Evaluation, which quantifies privacy protection at both answer and reasoning-chain levels. Experiments on the R-TOFU benchmark demonstrate that STaR achieves comprehensive and stable unlearning with minimal utility loss, setting a new standard for privacy-preserving reasoning in LRMs.

cs.AI↗

arXiv2026

Large Language Models for Limited Noisy Data: A Gravitational Wave Identification Study

This work investigates whether large language models (LLMs) offer advantages over traditional neural networks for astronomical data processing, in regimes with non-Gaussian, non-stationary noise and limited labeled samples. Gravitational wave observations provide an suitable test case, using only 90 LIGO events, finetuned LLMs achieve 97.4\% accuracy for identifying signals. Further experiments show that, in contrast to traditional networks that rely on large simulated datasets, additional simulated samples do not improve LLM performance, while scaling studies reveal predictable gains with increasing model size and dataset size. These results indicate that LLMs can extract discriminative structure directly from observational data and provide an efficient assessment for gravitational wave identification. The same strategy may extend to other astronomical domains with similar noise properties, such as radio or pulsar observations.

astro-ph.IM↗

arXiv2025

Classification and Characteristics of Double-trigger Gamma-ray Bursts

Over the past two decades, the \textit{Swift} and \textit{Fermi} missions have identified a rare class of ``double-trigger'' gamma-ray bursts (GRBs) that produce two independent trigger events. These events are characterized by a sufficiently long quiescent period during which the on-board trigger system can reset, resulting in the subsequent emission being recorded as a second independent event. Consistent sky localization confirms that both trigger events originated from the same astrophysical source. Here, we present a systematic classification and characteristics study of three such cases: GRB 091024A, GRB 110709B, and GRB 220627A. We investigate each trigger episode emission independently using standard classification diagnostics, including duration ($T_{90}$), hardness ratio, minimum variability timescale (MVT), spectral lag ($τ_{\rm lag}$), peak energy ($E_{\rm p}$), and energetics. We compare these properties with those of typical long GRBs (LGRBs) and with single-trigger LGRBs that exhibit extended quiescent periods. Our analysis reveals that all sub-bursts from the three double-trigger events consistently lie within the LGRB classification region. These results indicate that double-trigger GRBs are not a physically distinct subclass, but rather products of LGRB central engines that undergo extended dormancy and subsequent reactivation.

astro-ph.HE↗

arXiv2025

Seedance 1.5 pro: A Native Audio-Visual Joint Generation Foundation Model

Recent strides in video generation have paved the way for unified audio-visual generation. In this work, we present Seedance 1.5 pro, a foundational model engineered specifically for native, joint audio-video generation. Leveraging a dual-branch Diffusion Transformer architecture, the model integrates a cross-modal joint module with a specialized multi-stage data pipeline, achieving exceptional audio-visual synchronization and superior generation quality. To ensure practical utility, we implement meticulous post-training optimizations, including Supervised Fine-Tuning (SFT) on high-quality datasets and Reinforcement Learning from Human Feedback (RLHF) with multi-dimensional reward models. Furthermore, we introduce an acceleration framework that boosts inference speed by over 10X. Seedance 1.5 pro distinguishes itself through precise multilingual and dialect lip-syncing, dynamic cinematic camera control, and enhanced narrative coherence, positioning it as a robust engine for professional-grade content creation. Seedance 1.5 pro is now accessible on Volcano Engine at https://console.volcengine.com/ark/region:ark+cn-beijing/experience/vision?type=GenVideo.

cs.CV↗

arXiv2025

Adaptive Frequency Domain Alignment Network for Medical image segmentation

High-quality annotated data plays a crucial role in achieving accurate segmentation. However, such data for medical image segmentation are often scarce due to the time-consuming and labor-intensive nature of manual annotation. To address this challenge, we propose the Adaptive Frequency Domain Alignment Network (AFDAN)--a novel domain adaptation framework designed to align features in the frequency domain and alleviate data scarcity. AFDAN integrates three core components to enable robust cross-domain knowledge transfer: an Adversarial Domain Learning Module that transfers features from the source to the target domain; a Source-Target Frequency Fusion Module that blends frequency representations across domains; and a Spatial-Frequency Integration Module that combines both frequency and spatial features to further enhance segmentation accuracy across domains. Extensive experiments demonstrate the effectiveness of AFDAN: it achieves an Intersection over Union (IoU) of 90.9% for vitiligo segmentation in the newly constructed VITILIGO2025 dataset and a competitive IoU of 82.6% on the retinal vessel segmentation benchmark DRIVE, surpassing existing state-of-the-art approaches.

cs.CV↗

arXiv2025

InstructDubber: Instruction-based Alignment for Zero-shot Movie Dubbing

Movie dubbing seeks to synthesize speech from a given script using a specific voice, while ensuring accurate lip synchronization and emotion-prosody alignment with the character's visual performance. However, existing alignment approaches based on visual features face two key limitations: (1)they rely on complex, handcrafted visual preprocessing pipelines, including facial landmark detection and feature extraction; and (2) they generalize poorly to unseen visual domains, often resulting in degraded alignment and dubbing quality. To address these issues, we propose InstructDubber, a novel instruction-based alignment dubbing method for both robust in-domain and zero-shot movie dubbing. Specifically, we first feed the video, script, and corresponding prompts into a multimodal large language model to generate natural language dubbing instructions regarding the speaking rate and emotion state depicted in the video, which is robust to visual domain variations. Second, we design an instructed duration distilling module to mine discriminative duration cues from speaking rate instructions to predict lip-aligned phoneme-level pronunciation duration. Third, for emotion-prosody alignment, we devise an instructed emotion calibrating module, which finetunes an LLM-based instruction analyzer using ground truth dubbing emotion as supervision and predicts prosody based on the calibrated emotion analysis. Finally, the predicted duration and prosody, together with the script, are fed into the audio decoder to generate video-aligned dubbing. Extensive experiments on three major benchmarks demonstrate that InstructDubber outperforms state-of-the-art approaches across both in-domain and zero-shot scenarios.

cs.SD↗

arXiv2025

Seedream 4.0: Toward Next-generation Multimodal Image Generation

We introduce Seedream 4.0, an efficient and high-performance multimodal image generation system that unifies text-to-image (T2I) synthesis, image editing, and multi-image composition within a single framework. We develop a highly efficient diffusion transformer with a powerful VAE which also can reduce the number of image tokens considerably. This allows for efficient training of our model, and enables it to fast generate native high-resolution images (e.g., 1K-4K). Seedream 4.0 is pretrained on billions of text-image pairs spanning diverse taxonomies and knowledge-centric concepts. Comprehensive data collection across hundreds of vertical scenarios, coupled with optimized strategies, ensures stable and large-scale training, with strong generalization. By incorporating a carefully fine-tuned VLM model, we perform multi-modal post-training for training both T2I and image editing tasks jointly. For inference acceleration, we integrate adversarial distillation, distribution matching, and quantization, as well as speculative decoding. It achieves an inference time of up to 1.8 seconds for generating a 2K image (without a LLM/VLM as PE model). Comprehensive evaluations reveal that Seedream 4.0 can achieve state-of-the-art results on both T2I and multimodal image editing. In particular, it demonstrates exceptional multimodal capabilities in complex tasks, including precise image editing and in-context reasoning, and also allows for multi-image reference, and can generate multiple output images. This extends traditional T2I systems into an more interactive and multidimensional creative tool, pushing the boundary of generative AI for both creativity and professional applications. We further scale our model and data as Seedream 4.5. Seedream 4.0 and Seedream 4.5 are accessible on Volcano Engine https://www.volcengine.com/experience/ark?launch=seedream.

cs.CV↗

arXiv2025

NeuMatC: A General Neural Framework for Fast Parametric Matrix Operation

Matrix operations (e.g., inversion and singular value decomposition (SVD)) are fundamental in science and engineering. In many emerging real-world applications (such as wireless communication and signal processing), these operations must be performed repeatedly over matrices with parameters varying continuously. However, conventional methods tackle each matrix operation independently, underexploring the inherent low-rankness and continuity along the parameter dimension, resulting in significantly redundant computation. To address this challenge, we propose \textbf{\textit{Neural Matrix Computation Framework} (NeuMatC)}, which elegantly tackles general parametric matrix operation tasks by leveraging the underlying low-rankness and continuity along the parameter dimension. Specifically, NeuMatC unsupervisedly learns a low-rank and continuous mapping from parameters to their corresponding matrix operation results. Once trained, NeuMatC enables efficient computations at arbitrary parameters using only a few basic operations (e.g., matrix multiplications and nonlinear activations), significantly reducing redundant computations. Experimental results on both synthetic and real-world datasets demonstrate the promising performance of NeuMatC, exemplified by over $3\times$ speedup in parametric inversion and $10\times$ speedup in parametric SVD compared to the widely used NumPy baseline in wireless communication, while maintaining acceptable accuracy.

cs.CV↗

arXiv2025

A fifth-order absolutely convergent fixed-point fast sweeping hybrid alternative WENO scheme for steady state of hyperbolic conservation laws

In this paper, we extend the previous work on absolutely convergent fixed-point fast sweeping WENO methods by Li et al. (J. Comput. Phys. 443: 110516, 2021) and design a fifth-order hybrid fast sweeping scheme for solving steady state problems of hyperbolic conservation laws. Unlike many other fast sweeping methods, the explicit property of fixed-point fast sweeping methods provides flexibility to apply the alternative weighted essentially non-oscillatory (AWENO) scheme with unequal-sized substencils as the local solver, which facilitates the usage of arbitrary monotone numerical fluxes. Furthermore, a novel hybrid technique is designed in the local solver to combine the nonlinear AWENO interpolation with the linear scheme for an additional improvement in efficiency of the high-order fast sweeping iterations. Numerical examples show that the developed fixed-point fast sweeping hybrid AWENO method with unequal-sized substencils can achieve absolute convergence (i.e., the residue of the fast sweeping iterations converges to machine zero / round off errors) more easily than the original AWENO method with equal-sized substencils, and is more efficient than the popular third-order total variation diminishing (TVD) Runge-Kutta time-marching approach to converge to steady state solutions.

math.NA↗

arXiv2025

Kondo-coupled van der Waals antiferromagnet with high-mobility quasiparticles

Two-dimensional van der Waals (vdW) materials exhibit high carrier mobility and tunability, making them suitable for low-power, high-performance electronic and spintronic applications. Incorporating narrow-band electronic correlation effects could further promote tunability, though mass renormalization may impact carrier mobility. It is therefore challenging to identify a vdW material with both high mobility and strong correlation. Herein, by a combination of optical spectroscopy and high-field quantum-oscillation measurements, we observe significant effective-mass enhancement in CeTe$_3$ at low temperature, arising from not only the band-structure modulation by antiferromagnetic ordering but also the narrow-band correlation effect. Despite the mass enhancement, the quantum mobility surprisingly \textit{increases} and reaches $\sim$2403 cm$^2$/Vs, likely benefiting from topological protection. Remarkably, these unique properties are maintained in atomically thin nanoflakes with quantum mobility enhanced to $\sim$3158 cm$^2$/Vs. Thus, CeTe$_3$ emerges as a promising Kondo-coupled vdW antiferromagnetic metal with high-mobility quasiparticles, potentially unlocking new device concepts.

cond-mat.str-el↗

arXiv2025

Search for Long-lived Particles at Future Lepton Colliders Using Deep Learning Techniques

Long-lived particles (LLPs) provide an unambiguous signal for physics beyond the Standard Model (BSM). They have a distinct detector signature, with decay lengths corresponding to lifetimes of around nanoseconds or longer. Lepton colliders allow LLP searches to be conducted in a clean environment, and such searches can reach their full physics potential when combined with machine learning (ML) techniques.This experimental study, utilizing comprehensive full simulation data samples, focuses on LLP searches resulting from Higgs decay in $e^+e^-\to ZH$. We demonstrate that, by employing deep neural network approaches the LLP signal efficiency can be improved up to 95\% for an LLP mass around 50 GeV and a lifetime of approximately 1 nanosecond, while rejecting all SM backgrounds. Furthermore, the signal sensitivity for the branching ratio of Higgs decaying into LLPs reaches a state-of-art limit of $1.0 \times 10^{-6}$ with a statistics of $4 \times 10^{6}$ Higgs.

hep-ex↗

arXiv2025

Development of the CEPC analog hadron calorimeter prototype

The Circular Electron Positron Collider (CEPC) is a next-generation electron$-$positron collider proposed for the precise measurement of the properties of the Higgs boson. To emphasize boson separation and jet reconstruction, the baseline design of the CEPC detector was guided by the particle flow algorithm (PFA) concept. As one of the calorimeter options, the analogue hadron calorimeter (AHCAL) was proposed. The CEPC AHCAL comprises a 40-layer sandwich structure using steel plates as absorbers and scintillator tiles coupled with silicon photomultipliers (SiPM) as sensitive units. To validate the feasibility of the AHCAL option, a series of studies were conducted to develop a prototype. This AHCAL prototype underwent an electronic test and a cosmic ray test to assess its performance and ensure it was ready for three beam tests performed in 2022 and 2023. The test beam data is currently under analysis, and the results are expected to deepen our understanding of hadron showers, validate the concept of Particle Flow Algorithm (PFA), and ultimately refine the design of the CEPC detector.

physics.ins-det↗

arXiv2025

Hybrid-graph neural network method for muon fast reconstruction in neutrino telescopes

Fast and accurate muon reconstruction is crucial for neutrino telescopes to improve experimental sensitivity and enable online triggering. This paper introduces a hybrid-graph neural network (GNN) method tailored for efficient muon track reconstruction, leveraging the robustness of GNNs, alongside traditional physics-based approaches. The "light GNN model" achieves a run-time of 0.19-0.29 ms per event on GPUs, offering a 3 orders of magnitude speedup compared to traditional likelihood-based methods, while maintaining a high reconstruction accuracy. For high-energy muons (10-100 TeV), the median angular error is approximately 0.1°, with errors in reconstructed Cherenkov photon emission positions being below 3-5 m, depending on the GNN model used. Furthermore, the semi-GNN method offers a mechanism to assess the quality of event reconstruction, enabling the identification and exclusion of poorly reconstructed events. These results establish the GNN-based approach as a promising solution for next-generation neutrino telescope data reconstruction.

hep-ex↗

arXiv2025

Adaptive Budget Allocation for Orthogonal-Subspace Adapter Tuning in LLMs Continual Learning

Large language models (LLMs) often suffer from catastrophic forgetting in continual learning (CL) scenarios, where performance on previously learned tasks degrades severely while training on sequentially arriving tasks. Although pioneering CL approaches using orthogonal subspaces can mitigate task interference, they typically employ fixed budget allocation, neglecting the varying complexity across tasks and layers. Besides, recent budget-adaptive tuning methods for LLMs often adopt multi-stage paradigms that decouple optimization and budget allocation. Such decoupling results in potential misalignment, which hinders those approaches' practical application in CL scenarios. To address these limitations, we propose OA-Adapter, a novel parameter-efficient approach for continual learning in LLMs that unifies dynamic budget adaptation with orthogonal subspace learning in an end-to-end training stage. Specifically, OA-Adapter introduces a dynamic bottleneck dimension adaptation mechanism that simultaneously allocates an efficient parameter budget and optimizes task objectives without misalignment.To effectively preserve previously acquired knowledge while coordinating with the dynamic budget allocation, orthogonal constraints are applied specifically between the parameter subspace of the current task and the dynamically allocated parameter subspaces of historical tasks. Experimental results on continual learning benchmarks demonstrate that OA-Adapter outperforms state-of-the-art methods in both accuracy and parameter efficiency. OA-Adapter achieves higher average accuracy while using 58.5% fewer parameters on the standard CL benchmark, and maintains its advantages on two larger benchmarks comprising 15 tasks.

cs.LG↗

arXiv2025

Fast Authenticated and Interoperable Multimedia Healthcare Data over Hybrid-Storage Blockchains

The integration of blockchain technology into healthcare presents a paradigm shift for secure data management, enabling decentralized and tamper-proof storage and sharing of sensitive Electronic Health Records (EHRs). However, existing blockchain-based healthcare systems, while providing robust access control, commonly overlook the high latency in user-side re-computation of hashes for integrity verification of large multimedia data, impairing their practicality, especially in time-sensitive clinical scenarios. In this paper, we propose FAITH, an innovative scheme for \underline{F}ast \underline{A}uthenticated and \underline{I}nteroperable mul\underline{T}imedia \underline{H}ealthcare data storage and sharing over hybrid-storage blockchains. Rather than user-side hash re-computations, FAITH lets an off-chain storage provider generate verifiable proofs using recursive Zero-Knowledge Proofs (ZKPs), while the user only needs to perform lightweight verification. For flexible access authorization, we leverage Proxy Re-Encryption (PRE) and enable the provider to conduct ciphertext re-encryption, in which the re-encryption correctness can be verified via ZKPs against the malicious provider. All metadata and proofs are recorded on-chain for public verification. We provide a comprehensive analysis of FAITH's security regarding data privacy and integrity. We implemented a prototype of FAITH, and extensive experiments demonstrated its practicality for time-critical healthcare applications, dramatically reducing user-side verification latency by up to $98\%$, bringing it from $4$ s down to around $70$ ms for a $5$ GB encrypted file.

cs.CR↗

arXiv2025

Multi-Granularity Sequence Denoising with Weakly Supervised Signal for Sequential Recommendation

Sequential recommendation aims to predict the next item based on user interests in historical interaction sequences. Historical interaction sequences often contain irrelevant noisy items, which significantly hinders the performance of recommendation systems. Existing research employs unsupervised methods that indirectly identify item-granularity irrelevant noise by predicting the ground truth item. Since these methods lack explicit noise labels, they are prone to misidentify users' interested items as noise. Additionally, while these methods focus on removing item-granularity noise driven by the ground truth item, they overlook interest-granularity noise, limiting their ability to perform broader denoising based on user interests. To address these issues, we propose Multi-Granularity Sequence Denoising with Weakly Supervised Signal for Sequential Recommendation(MGSD-WSS). MGSD-WSS first introduces the Multiple Gaussian Kernel Perceptron module to map the original and enhance sequence into a common representation space and utilizes weakly supervised signals to accurately identify noisy items in the historical interaction sequence. Subsequently, it employs the item-granularity denoising module with noise-weighted contrastive learning to obtain denoised item representations. Then, it extracts target interest representations from the ground truth item and applies noise-weighted contrastive learning to obtain denoised interest representations. Finally, based on the denoised item and interest representations, MGSD-WSS predicts the next item. Extensive experiments on five datasets demonstrate that the proposed method significantly outperforms state-of-the-art sequence recommendation and denoising models. Our code is available at https://github.com/lalunex/MGSD-WSS.

cs.IR↗

arXiv2025

New Physics Search at the CEPC: a General Perspective

The Circular Electron-Positron Collider (CEPC), a proposed next-generation Higgs factory, provides new opportunities to explore physics beyond the Standard Model (SM). With its clean electron-positron collision environment and the ability to collect large samples of Higgs, W, and Z bosons, the CEPC enables precision measurements and searches for new physics. This white paper outlines the CEPC's discovery potential, including studies of exotic decays of the Higgs, Z, and top quarks, dark matter and dark sector phenomena, long-lived particles, supersymmetry, and neutrino-related signatures. Advanced detector technologies and reconstruction techniques, such as one-to-one correspondence reconstruction and jet origin identification, significantly improve sensitivity to rare and weakly interacting processes. The CEPC is particularly well suited to probe the electroweak phase transition and test models of electroweak baryogenesis and dark sector interactions. In addition, global fit analyses highlight the CEPC's complementary role in constraining a wide range of new physics scenarios. These features position the CEPC as a powerful tool for exploring the next frontier in fundamental particle physics in the post-Higgs discovery era.

hep-ex↗

arXiv2025

VAMamba: An Efficient Visual Adaptive Mamba for Image Restoration

Recent Mamba-based image restoration methods have achieved promising results but remain limited by fixed scanning patterns and inefficient feature utilization. Conventional Mamba architectures rely on predetermined paths that cannot adapt to diverse degradations, constraining both restoration performance and computational efficiency. To overcome these limitations, we propose VAMamba, a Visual Adaptive Mamba framework with two key innovations. First, QCLAM(Queue-basedCacheLow-rankAdaptiveMemory)enhancesfeaturelearningthrougha FIFO cache that stores historical representations. Similarity between current LoRA-adapted and cached features guides intelligent fusion, enabling dynamic reuse while effectively controlling memorygrowth.Second, GPS-SS2D(GreedyPathScanSS2D)introducesadaptive scanning. A Vision Transformer generates score maps to estimate pixel importance, and a greedy strategy de termines optimal forward and backward scanning paths. These learned trajectories replace rigid patterns, enabling SS2D to perform targeted feature extraction. The integration of QCLAM and GPS-SS2D allows VAMamba to adaptively focus on degraded regions while maintaining high computational efficiency. Extensive experiments across diverse restoration tasks demonstrate that VAMamba consistently outperforms existing approaches in both restoration quality and efficiency, establishing new benchmarks for adaptive image restoration. Our code is available at https://github.com/WaterHQH/VAMamba.

cs.CV↗