arXiv subjects

Wei Wu

Publications and source records attributed to Wei Wu.

At least 145 records · Page 8Linked to original sources

arXiv2025

PromptCoT 2.0: Scaling Prompt Synthesis for Large Language Model Reasoning

Large language models (LLMs) are evolving from conversational systems into strong reasoners for tasks such as Olympiad mathematics and competitive programming. While scaling parameters and test-time computation has driven progress, a key bottleneck is the lack of high-quality training problems: human-curated datasets are costly and limited, while existing synthetic corpora are often too easy or narrow. PromptCoT 1.0 showed that injecting rationales into prompt synthesis increases problem difficulty. Building on this, we present PromptCoT 2.0, a scalable framework that replaces hand-crafted heuristics with an expectation-maximization (EM) loop, where rationales are iteratively refined to guide prompt construction. This produces problems that are both harder and more diverse than prior corpora. The synthetic prompts support two post-training regimes: (1) Self-Play, where strong models improve autonomously via verifiable feedback without stronger teachers; and (2) Supervised Fine-Tuning (SFT), where weaker models learn from teacher-distilled traces. Extensive experiments demonstrate the effectiveness of this approach. In self-play, applying PromptCoT 2.0 to Qwen3-30B-A3B-Thinking-2507 sets new state-of-the-art results at the 30B scale, with +4.4, +4.8, and +5.3 on AIME 24/25 and HMMT 25, +6.1 and +5.0 on LiveCodeBench v5/v6, and +35 Elo on Codeforces. In SFT, training Qwen2.5-7B-Instruct solely on synthetic prompts boosts accuracy to 73.1 (AIME 24), 65.6 (AIME 25), and 53.4 (LiveCodeBench v5), surpassing models trained on human or hybrid data. Analyses further confirm that PromptCoT 2.0 yields fundamentally harder and distributionally distinct problems. These results establish prompt synthesis as a new axis for scaling reasoning and position PromptCoT 2.0 as a scalable foundation for future open-source models. The implementation is available at https://github.com/inclusionAI/PromptCoT.

cs.LG↗

arXiv2025

A Semantic Segmentation Algorithm for Pleural Effusion Based on DBIF-AUNet

Pleural effusion semantic segmentation can significantly enhance the accuracy and timeliness of clinical diagnosis and treatment by precisely identifying disease severity and lesion areas. Currently, semantic segmentation of pleural effusion CT images faces multiple challenges. These include similar gray levels between effusion and surrounding tissues, blurred edges, and variable morphology. Existing methods often struggle with diverse image variations and complex edges, primarily because direct feature concatenation causes semantic gaps. To address these challenges, we propose the Dual-Branch Interactive Fusion Attention model (DBIF-AUNet). This model constructs a densely nested skip-connection network and innovatively refines the Dual-Domain Feature Disentanglement module (DDFD). The DDFD module orthogonally decouples the functions of dual-domain modules to achieve multi-scale feature complementarity and enhance characteristics at different levels. Concurrently, we design a Branch Interaction Attention Fusion module (BIAF) that works synergistically with the DDFD. This module dynamically weights and fuses global, local, and frequency band features, thereby improving segmentation robustness. Furthermore, we implement a nested deep supervision mechanism with hierarchical adaptive hybrid loss to effectively address class imbalance. Through validation on 1,622 pleural effusion CT images from Southwest Hospital, DBIF-AUNet achieved IoU and Dice scores of 80.1% and 89.0% respectively. These results outperform state-of-the-art medical image segmentation models U-Net++ and Swin-UNet by 5.7%/2.7% and 2.2%/1.5% respectively, demonstrating significant optimization in segmentation accuracy for complex pleural effusion CT images.

cs.CV↗

arXiv2025

Large Foundation Models for Trajectory Prediction in Autonomous Driving: A Comprehensive Survey

Trajectory prediction serves as a critical functionality in autonomous driving, enabling the anticipation of future motion paths for traffic participants such as vehicles and pedestrians, which is essential for driving safety. Although conventional deep learning methods have improved accuracy, they remain hindered by inherent limitations, including lack of interpretability, heavy reliance on large-scale annotated data, and weak generalization in long-tail scenarios. The rise of Large Foundation Models (LFMs) is transforming the research paradigm of trajectory prediction. This survey offers a systematic review of recent advances in LFMs, particularly Large Language Models (LLMs) and Multimodal Large Language Models (MLLMs) for trajectory prediction. By integrating linguistic and scene semantics, LFMs facilitate interpretable contextual reasoning, significantly enhancing prediction safety and generalization in complex environments. The article highlights three core methodologies: trajectory-language mapping, multimodal fusion, and constraint-based reasoning. It covers prediction tasks for both vehicles and pedestrians, evaluation metrics, and dataset analyses. Key challenges such as computational latency, data scarcity, and real-world robustness are discussed, along with future research directions including low-latency inference, causality-aware modeling, and motion foundation models.

cs.RO↗

arXiv2025

Scaling Video-Language Models to 10K Frames via Hierarchical Differential Distillation

Long-form video processing fundamentally challenges vision-language models (VLMs) due to the high computational costs of handling extended temporal sequences. Existing token pruning and feature merging methods often sacrifice critical temporal dependencies or dilute semantic information. We introduce differential distillation, a principled approach that systematically preserves task-relevant information while suppressing redundancy. Based on this principle, we develop ViLAMP, a hierarchical video-language model that processes hour-long videos at "mixed precision" through two key mechanisms: (1) differential keyframe selection that maximizes query relevance while maintaining temporal distinctiveness at the frame level and (2) differential feature merging that preserves query-salient features in non-keyframes at the patch level. Hence, ViLAMP retains full information in keyframes while reducing non-keyframes to their most salient features, resembling mixed-precision training. Extensive experiments demonstrate ViLAMP's superior performance across four video understanding benchmarks, particularly on long-form content. Notably, ViLAMP can process ultra-long videos (up to 10K frames) on a single NVIDIA A100 GPU, achieving substantial computational efficiency while maintaining state-of-the-art performance. Code and model are available at https://github.com/steven-ccq/ViLAMP.

cs.CL↗

arXiv2025

Fast approximation algorithms for the 1-median problem on real-world large graphs

The 1-median problem (1MP) on undirected weighted graphs seeks to find a facility location minimizing the total weighted distance to all customer nodes. Although the 1MP can be solved exactly by computing the single-source shortest paths from each customer node, such approaches become computationally expensive on large-scale graphs with millions of nodes. In many real-world applications, such as recommendation systems based on large-scale knowledge graphs, the number of nodes (i.e., potential facility locations) is enormous, whereas the number of customer nodes is relatively small and spatially concentrated. In such cases, exhaustive graph exploration is not only inefficient but also unnecessary. Leveraging this observation, we propose three approximation algorithms that reduce computation by terminating Dijkstra's algorithm early. We provide theoretical analysis showing that one of the proposed algorithms guarantees an approximation ratio of 2, whereas the other two improve this ratio to 1.618. We demonstrate that the lower bound of the approximation ratio is 1.2 by presenting a specific instance. Moreover, we show that all proposed algorithms return optimal solutions when the number of customer nodes is less than or equal to three. Extensive experiments demonstrate that our algorithms significantly outperform baseline exact methods in runtime while maintaining near-optimal accuracy across all tested graph types. Notably, on grid graphs with 10 million nodes, our algorithms obtains all optimal solutions within 1 millisecond, whereas the baseline exact method requires over 70 seconds on average.

cs.DS↗

arXiv2025

OmniCache: A Trajectory-Oriented Global Perspective on Training-Free Cache Reuse for Diffusion Transformer Models

Diffusion models have emerged as a powerful paradigm for generative tasks such as image synthesis and video generation, with Transformer architectures further enhancing performance. However, the high computational cost of diffusion Transformers-stemming from a large number of sampling steps and complex per-step computations-presents significant challenges for real-time deployment. In this paper, we introduce OmniCache, a training-free acceleration method that exploits the global redundancy inherent in the denoising process. Unlike existing methods that determine caching strategies based on inter-step similarities and tend to prioritize reusing later sampling steps, our approach originates from the sampling perspective of DIT models. We systematically analyze the model's sampling trajectories and strategically distribute cache reuse across the entire sampling process. This global perspective enables more effective utilization of cached computations throughout the diffusion trajectory, rather than concentrating reuse within limited segments of the sampling procedure. In addition, during cache reuse, we dynamically estimate the corresponding noise and filter it out to reduce its impact on the sampling direction. Extensive experiments demonstrate that our approach accelerates the sampling process while maintaining competitive generative quality, offering a promising and practical solution for efficient deployment of diffusion-based generative models.

cs.CV↗

arXiv2025

VQualA 2025 Challenge on Face Image Quality Assessment: Methods and Results

Face images play a crucial role in numerous applications; however, real-world conditions frequently introduce degradations such as noise, blur, and compression artifacts, affecting overall image quality and hindering subsequent tasks. To address this challenge, we organized the VQualA 2025 Challenge on Face Image Quality Assessment (FIQA) as part of the ICCV 2025 Workshops. Participants created lightweight and efficient models (limited to 0.5 GFLOPs and 5 million parameters) for the prediction of Mean Opinion Scores (MOS) on face images with arbitrary resolutions and realistic degradations. Submissions underwent comprehensive evaluations through correlation metrics on a dataset of in-the-wild face images. This challenge attracted 127 participants, with 1519 final submissions. This report summarizes the methodologies and findings for advancing the development of practical FIQA approaches.

cs.CV↗

arXiv2025

SpectrumFM: Redefining Spectrum Cognition via Foundation Modeling

The enhancement of spectrum efficiency and the realization of secure spectrum utilization are critically dependent on spectrum cognition. However, existing spectrum cognition methods often exhibit limited generalization and suboptimal accuracy when deployed across diverse spectrum environments and tasks. To overcome these challenges, we propose a spectrum foundation model, termed SpectrumFM, which provides a new paradigm for spectrum cognition. An innovative spectrum encoder that exploits the convolutional neural networks and the multi-head self attention mechanisms is proposed to effectively capture both fine-grained local signal structures and high-level global dependencies in the spectrum data. To enhance its adaptability, two novel self-supervised learning tasks, namely masked reconstruction and next-slot signal prediction, are developed for pre-training SpectrumFM, enabling the model to learn rich and transferable representations. Furthermore, low-rank adaptation (LoRA) parameter-efficient fine-tuning is exploited to enable SpectrumFM to seamlessly adapt to various downstream spectrum cognition tasks, including spectrum sensing (SS), anomaly detection (AD), and wireless technology classification (WTC). Extensive experiments demonstrate the superiority of SpectrumFM over state-of-the-art methods. Specifically, it improves detection probability in the SS task by 30% at -4 dB signal-to-noise ratio (SNR), boosts the area under the curve (AUC) in the AD task by over 10%, and enhances WTC accuracy by 9.6%.

eess.SP↗

arXiv2025

Towards Constraint-aware Learning for Resource Allocation in NFV Networks

Virtual Network Embedding (VNE) is a fundamental resource allocation challenge that is associated with hard and multifaceted constraints in network function virtualization (NFV). Existing works for VNE struggle to handle such complex constraints, leading to compromised system performance and stability. In this paper, we propose a \textbf{CON}straint-\textbf{A}ware \textbf{L}earning framework, named \textbf{CONAL}, for efficient constraint handling in VNE. Concretely, we formulate the VNE problem as a constrained Markov decision process with violation tolerance, enabling precise assessments of both solution quality and constraint violations. To achieve the persistent zero violation to guarantee solutions' feasibility, we propose a reachability-guided optimization with an adaptive reachability budget method. This method also stabilizes policy optimization by appropriately handling scenarios with no feasible solutions. Furthermore, we propose a constraint-aware graph representation method to efficiently learn cross-graph relations and constrained path connectivity in VNE. Finally, extensive experimental results demonstrate the superiority of our proposed method over state-of-the-art baselines. Our code is available at \href{https://github.com/GeminiLight/conal-vne}{https://github.com/GeminiLight/conal-vne}.

cs.NI↗

arXiv2025

Magnetic Feshbach resonances in Ba$^+$+Li collisions due to strong spin-orbit coupling

We report a pronounced dependence of magnetic Feshbach resonance spectra on the initial hyperfine-Zeeman state of Li in ultracold $^{138}$Ba$^+$+$^6$Li collisions. The measured number and distribution of resonances differ significantly between the two lowest states despite their similar electron spin character. We address this puzzle by developing a comprehensive yet generic computational model calibrated against key statistical features in the experimental spectrum. We confirm that strong spin-orbit coupling induces essential changes in the distribution of resonances, leading to an increase in the number of resolvable resonances. Our model reproduces the statistics of the spectrum with the lowest Li state but struggles with the second-lowest state, where a significantly smaller number of resonances is experimentally observed.

physics.atom-ph↗

arXiv2025

Measuring Spiritual Values and Bias of Large Language Models

Large language models (LLMs) have become integral tool for users from various backgrounds. LLMs, trained on vast corpora, reflect the linguistic and cultural nuances embedded in their pre-training data. However, the values and perspectives inherent in this data can influence the behavior of LLMs, leading to potential biases. As a result, the use of LLMs in contexts involving spiritual or moral values necessitates careful consideration of these underlying biases. Our work starts with verification of our hypothesis by testing the spiritual values of popular LLMs. Experimental results show that LLMs' spiritual values are quite diverse, as opposed to the stereotype of atheists or secularists. We then investigate how different spiritual values affect LLMs in social-fairness scenarios e.g., hate speech identification). Our findings reveal that different spiritual values indeed lead to different sensitivity to different hate target groups. Furthermore, we propose to continue pre-training LLMs on spiritual texts, and empirical results demonstrate the effectiveness of this approach in mitigating spiritual bias.

cs.CL↗

arXiv2025

Unsupervised Morphological Tree Tokenizer

As a cornerstone in language modeling, tokenization involves segmenting text inputs into pre-defined atomic units. Conventional statistical tokenizers often disrupt constituent boundaries within words, thereby corrupting semantic information. To address this drawback, we introduce morphological structure guidance to tokenization and propose a deep model to induce character-level structures of words. Specifically, the deep model jointly encodes internal structures and representations of words with a mechanism named $\textit{MorphOverriding}$ to ensure the indecomposability of morphemes. By training the model with self-supervised objectives, our method is capable of inducing character-level structures that align with morphological rules without annotated training data. Based on the induced structures, our algorithm tokenizes words through vocabulary matching in a top-down manner. Empirical results indicate that the proposed method effectively retains complete morphemes and outperforms widely adopted methods such as BPE and WordPiece on both morphological segmentation tasks and language modeling tasks. Code is available at https://github.com/martianmartina/TreeTokenizer.

cs.CL↗

arXiv2025

Extended Inductive Reasoning for Personalized Preference Inference from Behavioral Signals

Large language models (LLMs) have demonstrated significant success in complex reasoning tasks such as math and coding. In contrast to these tasks where deductive reasoning predominates, inductive reasoning-the ability to derive general rules from incomplete evidence, remains underexplored. This paper investigates extended inductive reasoning in LLMs through the lens of personalized preference inference, a critical challenge in LLM alignment where current approaches struggle to capture diverse user preferences. The task demands strong inductive reasoning capabilities as user preferences are typically embedded implicitly across various interaction forms, requiring models to synthesize consistent preference patterns from scattered signals. We propose AlignXplore, a model that leverages extended reasoning chains to enable systematic preference inference from behavioral signals in users' interaction histories. Such explicit preference articulation enables efficient streaming inference: when new behavioral signals emerge, the model can directly build upon previously inferred preference descriptions rather than reprocessing historical signals from scratch, while also supporting iterative refinement to the inferred preferences. We develop AlignXplore by combining cold-start training based on synthetic data with subsequent online reinforcement learning. Through extensive experiments, we demonstrate that AlignXplore achieves substantial improvements over the backbone model by an average of 15.49\% on in-domain and out-of-domain benchmarks, while maintaining strong generalization ability across different input formats and downstream models. Further analyses establish best practices for preference inference learning through systematic comparison of reward modeling strategies, while revealing the emergence of human-like inductive reasoning patterns during training.

cs.CL↗

arXiv2025

Parallel spin wave for the Villain model

In this paper, we study the Villain model in $\mathbb{Z}^d$ in dimension $d\geq 3$. It is conjectured, that the parallel correlation function in the infinite volume Gibbs state, i.e., the map $$ x \mapsto \langle \cosθ(0) \cosθ(x) \rangle_{μ_{\mathrm{Vil}, β}} -\left( \langle \cosθ(0) \rangle_{μ_{\mathrm{Vil}, β}} \right)^2, $$ decays like $|x|^{-2(d-2)}$ as $|x| \to \infty$ at low temperature. The results of Bricmont, Fontaine, Lebowitz, Lieb, and Spencer (1981) show that for the related XY model, this correlation decays at least as fast as $|x|^{2-d}$. We prove the optimal upper and lower bounds for the Villain model in $d=3$, up to a logarithmic correction, and also improve the upper bound in general dimensions. Our proof builds upon the approach developed in our previous article, which in turn is inspired by a key observation of Fröhlich and Spencer (1982): in the low temperature regime, a combination of duality transformation and renormalisation allows certain properties of the Villain model to be analysed in terms of a (vector-valued) $\nabla φ$ interface model. This latter model can be investigated using the Helffer-Sjöstrand representation formula combined with tools of elliptic and parabolic regularity.

math.PR↗

arXiv2025

Knowledge Graph-Based Explainable and Generalized Zero-Shot Semantic Communications

Data-driven semantic communication is based on superficial statistical patterns, thereby lacking interpretability and generalization, especially for applications with the presence of unseen data. To address these challenges, we propose a novel knowledge graph-enhanced zero-shot semantic communication (KGZS-SC) network. Guided by the structured semantic information from a knowledge graph-based semantic knowledge base (KG-SKB), our scheme provides generalized semantic representations and enables reasoning for unseen cases. Specifically, the KG-SKB aligns the semantic features in a shared category semantics embedding space and enhances the generalization ability of the transmitter through aligned semantic features, thus reducing communication overhead by selectively transmitting compact visual semantics. At the receiver, zero-shot learning (ZSL) is leveraged to enable direct classification for unseen cases without the demand for retraining or additional computational overhead, thereby enhancing the adaptability and efficiency of the classification process in dynamic or resource-constrained environments. The simulation results conducted on the APY datasets show that the proposed KGZS-SC network exhibits robust generalization and significantly outperforms existing SC frameworks in classifying unseen categories across a range of SNR levels.

cs.LG↗

arXiv2025

RoboScape: Physics-informed Embodied World Model

World models have become indispensable tools for embodied intelligence, serving as powerful simulators capable of generating realistic robotic videos while addressing critical data scarcity challenges. However, current embodied world models exhibit limited physical awareness, particularly in modeling 3D geometry and motion dynamics, resulting in unrealistic video generation for contact-rich robotic scenarios. In this paper, we present RoboScape, a unified physics-informed world model that jointly learns RGB video generation and physics knowledge within an integrated framework. We introduce two key physics-informed joint training tasks: temporal depth prediction that enhances 3D geometric consistency in video rendering, and keypoint dynamics learning that implicitly encodes physical properties (e.g., object shape and material characteristics) while improving complex motion modeling. Extensive experiments demonstrate that RoboScape generates videos with superior visual fidelity and physical plausibility across diverse robotic scenarios. We further validate its practical utility through downstream applications including robotic policy training with generated data and policy evaluation. Our work provides new insights for building efficient physics-informed world models to advance embodied intelligence research. The code is available at: https://github.com/tsinghua-fib-lab/RoboScape.

cs.CV↗

arXiv2025

AeroLite-MDNet: Lightweight Multi-task Deviation Detection Network for UAV Landing

Unmanned aerial vehicles (UAVs) are increasingly employed in diverse applications such as land surveying, material transport, and environmental monitoring. Following missions like data collection or inspection, UAVs must land safely at docking stations for storage or recharging, which is an essential requirement for ensuring operational continuity. However, accurate landing remains challenging due to factors like GPS signal interference. To address this issue, we propose a deviation warning system for UAV landings, powered by a novel vision-based model called AeroLite-MDNet. This model integrates a multiscale fusion module for robust cross-scale object detection and incorporates a segmentation branch for efficient orientation estimation. We introduce a new evaluation metric, Average Warning Delay (AWD), to quantify the system's sensitivity to landing deviations. Furthermore, we contribute a new dataset, UAVLandData, which captures real-world landing deviation scenarios to support training and evaluation. Experimental results show that our system achieves an AWD of 0.7 seconds with a deviation detection accuracy of 98.6\%, demonstrating its effectiveness in enhancing UAV landing reliability. Code will be available at https://github.com/ITTTTTI/Maskyolo.git

cs.RO↗

arXiv2025

Block Tensor Decomposition: A dual grid scheme with formal O(N3) for THC decomposition of molecular systems

Accurate and fast treatment of electron-electron interactions remains a central challenge in electronic structure theory because post-Hartree-Fock methods often suffered from the computational cost for 4-index electron repulsion integrals (ERIs). Low-rank approaches such as tensor hyper-contraction (THC) and interpolative separable density fitting (ISDF) have been proposed for Hartree-Fock exchange and correlation's calculations. Their application to molecular systems remains inefficient due to the construction of THC kernel whose time scale increases as quartic with the number of basis functions. In this work, we present an algorithm named block tensor decomposition (BTD) based on a dual grid scheme that combines Hilbert sort and pivoted Cholesky decomposition to generate compact interpolative grids, allowing strict $O(N^3)$ scaling for THC/ISDF kernel construction. The key parameters in BTD are optimized via differential evolution, balancing efficiency and accuracy. Furthermore, we apply BTD in scaled opposite-spin MP2 (SOS-MP2), leveraging sparse mapping in real space to achieve quadratic scaling for electron correlation calculation and linear scaling for exchange calculation. This work advances low-scaling THC/ISDF methodologies for molecular systems, offering a robust framework for efficient and accurate electronic structure computations.

physics.chem-ph↗