arXiv subjects

Wei Wu

Publications and source records attributed to Wei Wu.

At least 127 records · Page 7Linked to original sources

arXiv2025

Towards Comprehensive Sampling of SMT Solutions

This work focuses on effectively generating diverse solutions for satisfiability modulo theories (SMT) formulas, targeting the theories of bit-vectors, arrays, and uninterpreted functions, which is a critical task in software and hardware testing. Generating diverse SMT solutions helps uncover faults and detect safety violations during the verification and testing process, resulting in the SMT sampling problem, i.e., constructing a small number of solutions while achieving comprehensive coverage of the constraint space. While high coverage is crucial for exploring system behaviors, reducing the number of solutions is of great importance, as excessive solutions increase testing time and resource usage, undermining efficiency. In this work, we introduce PanSampler, a novel SMT sampler that achieves high coverage with a small number of solutions. It incorporates three novel techniques, i.e., diversity-aware SMT algorithm, abstract syntax tree (AST)-guided scoring function and post-sampling optimization technology, enhancing its practical performance. It iteratively samples solutions, evaluates candidates, and employs local search to refine solutions, ensuring high coverage with a small number of samples. Extensive experiments on practical benchmarks demonstrate that PanSampler exhibits a significantly stronger capability to reach high target coverage, while requiring fewer solutions than current samplers to achieve the same coverage level. Furthermore, our empirical evaluation on practical subjects, which are collected from real-world software systems, shows that PanSampler achieves higher fault detection capability and reduces the number of required test cases from 32.6\% to 76.4\% to reach the same fault detection effectiveness, leading to a substantial improvement in testing efficiency. PanSampler advances SMT sampling, reducing the cost of software testing and hardware verification.

cs.SE↗

arXiv2025

DynaAct: Large Language Model Reasoning with Dynamic Action Spaces

In modern sequential decision-making systems, the construction of an optimal candidate action space is critical to efficient inference. However, existing approaches either rely on manually defined action spaces that lack scalability or utilize unstructured spaces that render exhaustive search computationally prohibitive. In this paper, we propose a novel framework named \textsc{DynaAct} for automatically constructing a compact action space to enhance sequential reasoning in complex problem-solving scenarios. Our method first estimates a proxy for the complete action space by extracting general sketches observed in a corpus covering diverse complex reasoning problems using large language models. We then formulate a submodular function that jointly evaluates candidate actions based on their utility to the current state and their diversity, and employ a greedy algorithm to select an optimal candidate set. Extensive experiments on six diverse standard benchmarks demonstrate that our approach significantly improves overall performance, while maintaining efficient inference without introducing substantial latency. The implementation is available at https://github.com/zhaoxlpku/DynaAct.

cs.LG↗

arXiv2025

AgentPRM: Process Reward Models for LLM Agents via Step-Wise Promise and Progress

Despite rapid development, large language models (LLMs) still encounter challenges in multi-turn decision-making tasks (i.e., agent tasks) like web shopping and browser navigation, which require making a sequence of intelligent decisions based on environmental feedback. Previous work for LLM agents typically relies on elaborate prompt engineering or fine-tuning with expert trajectories to improve performance. In this work, we take a different perspective: we explore constructing process reward models (PRMs) to evaluate each decision and guide the agent's decision-making process. Unlike LLM reasoning, where each step is scored based on correctness, actions in agent tasks do not have a clear-cut correctness. Instead, they should be evaluated based on their proximity to the goal and the progress they have made. Building on this insight, we propose a re-defined PRM for agent tasks, named AgentPRM, to capture both the interdependence between sequential decisions and their contribution to the final goal. This enables better progress tracking and exploration-exploitation balance. To scalably obtain labeled data for training AgentPRM, we employ a Temporal Difference-based (TD-based) estimation method combined with Generalized Advantage Estimation (GAE), which proves more sample-efficient than prior methods. Extensive experiments across different agentic tasks show that AgentPRM is over $8\times$ more compute-efficient than baselines, and it demonstrates robust improvement when scaling up test-time compute. Moreover, we perform detailed analyses to show how our method works and offer more insights, e.g., applying AgentPRM to the reinforcement learning of LLM agents.

cs.CL↗

arXiv2025

Introducing LongCat-Flash-Thinking: A Technical Report

We present LongCat-Flash-Thinking, an efficient 560-billion-parameter open-source Mixture-of-Experts (MoE) reasoning model. Its advanced capabilities are cultivated through a meticulously crafted training process, beginning with long Chain-of-Thought (CoT) data cold-start and culminating in large-scale Reinforcement Learning (RL). We first employ a well-designed cold-start training strategy, which significantly enhances the reasoning potential and equips the model with specialized skills in both formal and agentic reasoning. Then, a core innovation is our domain-parallel training scheme, which decouples optimization across distinct domains (e.g., STEM, Code, Agentic) and subsequently fuses the resulting expert models into a single, nearly Pareto-optimal model. This entire process is powered by our Dynamic ORchestration for Asynchronous rollout (DORA) system, a large-scale RL framework that delivers a greater than threefold training speedup over synchronous methods on tens of thousands of accelerators. As a result, LongCat-Flash-Thinking achieves state-of-the-art performance among open-source models on a suite of complex reasoning tasks. The model exhibits exceptional efficiency in agentic reasoning, reducing average token consumption by 64.5% (from 19, 653 to 6, 965) on AIME-25, without degrading task accuracy. We release LongCat-Flash-Thinking to promote further advances in reasoning systems and agentic AI research.

cs.AI↗

arXiv2025

Uniqueness for phaseless inverse elastic scattering problem for periodic structures

This paper establishes uniqueness results of inverse elastic scattering problem with phaseless near-field data in periodic structures in $\mathbb{R}^2$ and periodic/biperiodic structures in $\mathbb{R}^3$. We use a superposition of two point sources in each periodic unit with different positions as the incident field, and measures the phaseless near-field data on a line parallel to $x_1$-axis in 2D, or on a plane parallel to $(x_1,x_2)$-plane in 3D. We first calculate the explicit formula of quasi-periodic/biperiodic Green's functions of Lamé system in $\mathbb{R}^3$. Then, to establish the uniqueness results, the reciprocity relations for point sources, scattered fields, and total fields are derived. Finally, with the help of Rayleigh's expansion, the uniqueness results are proved. The quasi-periodic/biperiodic Green's functions of Lamé system in $\mathbb{R}^3$, the reciprocity relations, and Rayleigh's expansion in $\mathbb{R}^3$ are novel results as important by-products in the proof process.

math.AP↗

arXiv2025

Lithium Niobate Vertical Cavity Electro-Optic Modulator

Electro-optic modulators (EOMs) are vital for optical imaging and information processing, with free-space devices enabling LiDAR and beam control. Lithium niobate (LN), powered by the strong Pockels effect and scalable LN-on-insulator (LNOI) platform, has become a leading material for high-performance EOMs. Here we realize a vertical-cavity EOM in which an LN membrane is sandwiched between two photonic crystal (PhC) mirrors with integrated electrodes. The cavity supports sharp defect-mode resonances that shift efficiently under the Pockels effect, enabling strong modulation of transmission. Experiments show a depth of 43 % at 50 V and a bandwidth of 5 MHz. This architecture combines free-space compatibility with fabrication simplicity, opening new routes to compact electro-optic platforms for ranging, holography, and beam steering.

physics.optics↗

arXiv2025

Mermin-Wagner theorem for dimers, monomer double-dimers, and spatial random permutations

We study a generalisation of the double-dimer model that encompasses several models of interest, including the monomer double-dimer model, spatial random permutations, the dimer model, and the spin $O(N)$ model, and which is also related to the loop $O(N)$ model. We show that on two-dimensional-like graphs (such as slabs), both the correlation function and the probability that a loop visits two vertices decay to zero as the distance between the vertices diverges. Our approach is based on the introduction of a new complex spin representation for all models in this class, together with a new proof of the Mermin-Wagner theorem that does not require positivity of the Gibbs measure. Even for the well-studied dimer and double-dimer models our results are new: since they do not rely on exact solvability or Kasteleyn's theorem, they apply beyond the planar-graph setting.

math.PR↗

arXiv2025

Hardware-aligned Hierarchical Sparse Attention for Efficient Long-term Memory Access

A key advantage of Recurrent Neural Networks (RNNs) over Transformers is their linear computational and space complexity enables faster training and inference for long sequences. However, RNNs are fundamentally unable to randomly access historical context, and simply integrating attention mechanisms may undermine their efficiency advantages. To overcome this limitation, we propose Hierarchical Sparse Attention (HSA), a novel attention mechanism that enhances RNNs with long-range random access flexibility while preserving their merits in efficiency and length generalization. HSA divides inputs into chunks, selects the top-$k$ chunks and hierarchically aggregates information. The core innovation lies in learning token-to-chunk relevance based on fine-grained token-level information inside each chunk. This approach enhances the precision of chunk selection across both in-domain and out-of-domain context lengths. To make HSA efficient, we further introduce a hardware-aligned kernel design. By combining HSA with Mamba, we introduce RAMba, which achieves perfect accuracy in passkey retrieval across 64 million contexts despite pre-training on only 4K-length contexts, and significant improvements on various downstream tasks, with nearly constant memory footprint. These results show RAMba's huge potential in long-context modeling.

cs.CL↗

arXiv2025

Absence of magnetic order and magnetic fluctuations in RuO$_{2}$

A novel magnetic class blending ferromagnetism and antiferromagnetism, termed altermagnetism, has gained significant attention for its staggered order in coordinate and momentum spaces, time-reversal symmetry-breaking phenomena, and promising applications in spintronics. Ruthenium dioxide (RuO$_{2}$) has been considered a candidate material for altermagnetism, yet the presence of magnetic moments on Ru atoms remains a subject of debate. In this study, we systematically investigated the magnetic properties of RuO$_{2}$ powder using nuclear quadrupole resonance (NQR) measurements. The NQR spectra show that there is no internal magnetic field. Furthermore, the temperature independence of spin-lattice relaxation rate, $1/T_1T$, proves that there are no magnetic fluctuations. Our results unambiguously demonstrate that Ru atoms in RuO$_{2}$ possess neither static magnetic moments nor fluctuating magnetic moments, and thus RuO$_{2}$ does not possess the magnetic characteristics essential for altermagnetism.

cond-mat.str-el↗

arXiv2025

Experimental Proposal on Scalable Radio-Frequency Magnetometer with Trapped Ions

Quantum magnetometry represents a fundamental component of quantum metrology, where trapped-ion systems have achieved $\rm{pT}/\sqrt{\rm{Hz}}$ sensitivity in single-ion radio-frequency magnetic field measurements via dressed states based dynamical decoupling. Here we propose a scalable trapped-ion magnetometer utilizing the mixed dynamical decoupling method, combining dressed states with periodic sequences to suppress decoherence and spatial magnetic field inhomogeneity. With numerical simulations for a $10^4$ ion system with realistic experimental parameters, we demonstrate that a sensitivity of 13 $\rm{fT}/\sqrt{\rm{Hz}}$ for the radio-frequency field could be reached. Such a sensitivity could be obtained via robust resilience to magnetic field drift noise and inhomogeneity, where coherence time could be extended to the order of several minutes on average. This method enables scalable trapped-ion magnetometry, demonstrating its potential as a robust and practical solution for advancing quantum sensing applications.

quant-ph↗

arXiv2025

Restoring Quantum Superiority of Noisy Quantum Illumination

Quantum illumination uses quantum entanglement as a resource to enable higher-resolution detection of low-reflectivity targets than is possible with classical techniques. This revolutionary technology could transform modern radar. However, it is widely believed that the decoherence induced by the ubiquitous quantum noise destroys the superiority of quantum illumination, severely constraining its performance and application in our present noisy intermediate-scale quantum era. Here, we propose a method to restore the quantum superiority of the quantum illumination in the presence of quantum noises. Going beyond the widely used Born-Markov approximation, we discover that the resolution of noisy quantum illumination is highly sensitive to the energy spectrum of the composite system formed by each of the two light modes and its local quantum noise. When a bound state is present in the energy spectrum, the resolution asymptotically approaches its ideal form. Our result establishes a physical principle to preserve the quantum superiority and paves the way for the realization of high-resolution quantum illumination in noisy situations.

quant-ph↗

arXiv2025

Primal-dual splitting methods for phase-field surfactant model with moving contact lines

Surfactants have important effects on the dynamics of droplets on solid surfaces, which has inspired many industrial applications. Phase-field surfactant model with moving contact lines (PFS-MCL) has been employed to investigate the complex droplet dynamics with surfactants, while its numerical simulation remains challenging due to the coupling of gradient flows with respect to transport distances involving nonlinear and degenerate mobilities. We propose a novel structure-preserving variational scheme for PFS-MCL model with the dynamic boundary condition based on the minimizing movement scheme and optimal transport theory for Wasserstein gradient flows. The proposed scheme consists of a series of convex minimization problems and can be efficiently solved by our proposed primal-dual splitting method and its accelerated versions. By respecting the underlying PDE's variational structure with respect to the transport distance, the proposed scheme is proved to inherits the desirable properties including original energy dissipation, bound-preserving, and mass conservation. Through a suite of numerical simulations, we validate the performance of the proposed scheme and investigate the effects of surfactants on the droplet dynamics.

math.NA↗

arXiv2025

UniMamba: Unified Spatial-Channel Representation Learning with Group-Efficient Mamba for LiDAR-based 3D Object Detection

Recent advances in LiDAR 3D detection have demonstrated the effectiveness of Transformer-based frameworks in capturing the global dependencies from point cloud spaces, which serialize the 3D voxels into the flattened 1D sequence for iterative self-attention. However, the spatial structure of 3D voxels will be inevitably destroyed during the serialization process. Besides, due to the considerable number of 3D voxels and quadratic complexity of Transformers, multiple sequences are grouped before feeding to Transformers, leading to a limited receptive field. Inspired by the impressive performance of State Space Models (SSM) achieved in the field of 2D vision tasks, in this paper, we propose a novel Unified Mamba (UniMamba), which seamlessly integrates the merits of 3D convolution and SSM in a concise multi-head manner, aiming to perform "local and global" spatial context aggregation efficiently and simultaneously. Specifically, a UniMamba block is designed which mainly consists of spatial locality modeling, complementary Z-order serialization and local-global sequential aggregator. The spatial locality modeling module integrates 3D submanifold convolution to capture the dynamic spatial position embedding before serialization. Then the efficient Z-order curve is adopted for serialization both horizontally and vertically. Furthermore, the local-global sequential aggregator adopts the channel grouping strategy to efficiently encode both "local and global" spatial inter-dependencies using multi-head SSM. Additionally, an encoder-decoder architecture with stacked UniMamba blocks is formed to facilitate multi-scale spatial learning hierarchically. Extensive experiments are conducted on three popular datasets: nuScenes, Waymo and Argoverse 2. Particularly, our UniMamba achieves 70.2 mAP on the nuScenes dataset.

cs.CV↗

arXiv2025

FreqPDE: Rethinking Positional Depth Embedding for Multi-View 3D Object Detection Transformers

Detecting 3D objects accurately from multi-view 2D images is a challenging yet essential task in the field of autonomous driving. Current methods resort to integrating depth prediction to recover the spatial information for object query decoding, which necessitates explicit supervision from LiDAR points during the training phase. However, the predicted depth quality is still unsatisfactory such as depth discontinuity of object boundaries and indistinction of small objects, which are mainly caused by the sparse supervision of projected points and the use of high-level image features for depth prediction. Besides, cross-view consistency and scale invariance are also overlooked in previous methods. In this paper, we introduce Frequency-aware Positional Depth Embedding (FreqPDE) to equip 2D image features with spatial information for 3D detection transformer decoder, which can be obtained through three main modules. Specifically, the Frequency-aware Spatial Pyramid Encoder (FSPE) constructs a feature pyramid by combining high-frequency edge clues and low-frequency semantics from different levels respectively. Then the Cross-view Scale-invariant Depth Predictor (CSDP) estimates the pixel-level depth distribution with cross-view and efficient channel attention mechanism. Finally, the Positional Depth Encoder (PDE) combines the 2D image features and 3D position embeddings to generate the 3D depth-aware features for query decoding. Additionally, hybrid depth supervision is adopted for complementary depth learning from both metric and distribution aspects. Extensive experiments conducted on the nuScenes dataset demonstrate the effectiveness and superiority of our proposed method.

cs.CV↗

arXiv2025

AirScape: An Aerial Generative World Model with Motion Controllability

How to enable agents to predict the outcomes of their own motion intentions in three-dimensional space has been a fundamental problem in embodied intelligence. To explore general spatial imagination capability, we present AirScape, the first world model designed for six-degree-of-freedom aerial agents. AirScape predicts future observation sequences based on current visual inputs and motion intentions. Specifically, we construct a dataset for aerial world model training and testing, which consists of 11k video-intention pairs. This dataset includes first-person-view videos capturing diverse drone actions across a wide range of scenarios, with over 1,000 hours spent annotating the corresponding motion intentions. Then we develop a two-phase schedule to train a foundation model--initially devoid of embodied spatial knowledge--into a world model that is controllable by motion intentions and adheres to physical spatio-temporal constraints. Experimental results demonstrate that AirScape significantly outperforms existing foundation models in 3D spatial imagination capabilities, especially with over a 50% improvement in metrics reflecting motion alignment. The project is available at: https://embodiedcity.github.io/AirScape/.

cs.RO↗

arXiv2025

TokenSelect: Efficient Long-Context Inference and Length Extrapolation for LLMs via Dynamic Token-Level KV Cache Selection

Rapid advances in Large Language Models (LLMs) have spurred demand for processing extended context sequences in contemporary applications. However, this progress faces two challenges: performance degradation due to sequence lengths out-of-distribution, and excessively long inference times caused by the quadratic computational complexity of attention. These issues limit LLMs in long-context scenarios. In this paper, we propose Dynamic Token-Level KV Cache Selection (TokenSelect), a training-free method for efficient and accurate long-context inference. TokenSelect builds upon the observation of non-contiguous attention sparsity, using QK dot products to measure per-head KV Cache criticality at token-level. By per-head soft voting mechanism, TokenSelect selectively involves a few critical KV cache tokens in attention calculation without sacrificing accuracy. To further accelerate TokenSelect, we design the Selection Cache based on observations of consecutive Query similarity and implemented the efficient Paged Dot Product Kernel, significantly reducing the selection overhead. A comprehensive evaluation of TokenSelect demonstrates up to $23.84\times$ speedup in attention computation and up to $2.28\times$ acceleration in end-to-end latency, while providing superior performance compared to state-of-the-art long-context inference methods.

cs.CL↗

arXiv2025

Limitations of strong coupling in non-Markovian quantum thermometry

We investigate quantum thermometry using a single-qubit probe embedded in a non-Markovian environment, employing the numerically exact hierarchical equations of motion (HEOM) to overcome the limitations of Born-Markov approximations. Through a systematic analysis of the dynamical and steady-state behavior of the quantum signal-to-noise ratio (QSNR) for temperature estimation, we identify several key findings that challenge the conventional expectation that strong coupling necessarily enhances thermometric performance. In non-equilibrium dynamical thermometry, weak system-environment coupling generally yields the optimal QSNR, whereas in the steady-state regime, strong coupling enhances sensitivity only in the ultra-low-temperature limit, while weak coupling significantly improves precision at moderately low temperatures. To optimize performance across coupling regimes, we develop a hybrid computational framework that integrates HEOM with quantum-enhanced particle swarm optimization, enabling precise quantum dynamical control under varying coupling strengths. Our results reveal fundamental constraints and opportunities in quantum thermometry, offering practical strategies for the design of high-performance quantum thermometers operating in realistic open quantum systems.

quant-ph↗

arXiv2025

LongScape: Advancing Long-Horizon Embodied World Models with Context-Aware MoE

Video-based world models hold significant potential for generating high-quality embodied manipulation data. However, current video generation methods struggle to achieve stable long-horizon generation: classical diffusion-based approaches often suffer from temporal inconsistency and visual drift over multiple rollouts, while autoregressive methods tend to compromise on visual detail. To solve this, we introduce LongScape, a hybrid framework that adaptively combines intra-chunk diffusion denoising with inter-chunk autoregressive causal generation. Our core innovation is an action-guided, variable-length chunking mechanism that partitions video based on the semantic context of robotic actions. This ensures each chunk represents a complete, coherent action, enabling the model to flexibly generate diverse dynamics. We further introduce a Context-aware Mixture-of-Experts (CMoE) framework that adaptively activates specialized experts for each chunk during generation, guaranteeing high visual quality and seamless chunk transitions. Extensive experimental results demonstrate that our method achieves stable and consistent long-horizon generation over extended rollouts. Our code is available at: https://github.com/tsinghua-fib-lab/Longscape.

cs.CV↗