arXiv subjects

Chao Wang

Publications and source records attributed to Chao Wang.

At least 235 records · Page 13Linked to original sources

arXiv2025

Mitigating Spurious Correlations Between Question and Answer via Chain-of-Thought Correctness Perception Distillation

Large language models (LLMs) excel at reasoning tasks but are expensive to deploy. Thus small language models (SLMs) are fine-tuned on CoT data generated by LLMs to copy LLMs' abilities. However, these CoT data may include noisy rationales that either fail to substantiate the answers or contribute no additional information to support answer prediction, which leads SLMs to capture spurious correlations between questions and answers and compromise the quality of reasoning. In this work, we propose Chain-of-Thought Correctness Perception Distillation (CoPeD), which aims to improve the reasoning quality of the student model from the perspectives of task setting and data utilization. Firstly, we introduce a correctness-aware task setting that encourages the student model to predict answers based on correct rationales and revise them when they are incorrect. This setting improves the faithfulness of reasoning and allows the model to learn from its mistakes. Then, we propose a Correctness-Aware Weighted loss, which dynamically adjusts the contribution of each training instance based on the combined loss of the rationale and the answer. This strategy encourages the model to focus more on samples where the rationale offers stronger support for the correct answer. Experiments have shown that CoPeD is effective on both in-distribution (IND) and out-of-distribution (OOD) benchmark reasoning datasets.

cs.CL↗

arXiv2025

VIKSER: Visual Knowledge-Driven Self-Reinforcing Reasoning Framework

Visual reasoning refers to the task of solving questions about visual information. Current visual reasoning methods typically employ pre-trained vision-language model (VLM) strategies or deep neural network approaches. However, existing efforts are constrained by limited reasoning interpretability, while hindering by the phenomenon of underspecification in the question text. Additionally, the absence of fine-grained visual knowledge limits the precise understanding of subject behavior in visual reasoning tasks. To address these issues, we propose VIKSER (Visual Knowledge-Driven Self-Reinforcing Reasoning Framework). Specifically, VIKSER, trained using knowledge distilled from large language models, extracts fine-grained visual knowledge with the assistance of visual relationship detection techniques. Subsequently, VIKSER utilizes fine-grained visual knowledge to paraphrase the question with underspecification. Additionally, we design a novel prompting method called Chain-of-Evidence (CoE), which leverages the power of "evidence for reasoning" to endow VIKSER with interpretable reasoning capabilities. Meanwhile, the integration of self-reflection technology empowers VIKSER with the ability to learn and improve from its mistakes. Experiments conducted on widely used datasets demonstrate that VIKSER achieves new state-of-the-art (SOTA) results in relevant tasks. Moreover, VIKSER achieves performance on par with leading proprietary models, such as the latest ChatGPT-5.

cs.CV↗

arXiv2025

Distributed Safety-Critical MPC for Multi-Agent Formation Control and Obstacle Avoidance

For nonlinear multi-agent systems with high relative degrees, achieving formation control and obstacle avoidance in a distributed manner remains a significant challenge. To address this issue, we propose a novel distributed safety-critical model predictive control (DSMPC) algorithm that incorporates discrete-time high-order control barrier functions (DHCBFs) to enforce safety constraints, alongside discrete-time control Lyapunov functions (DCLFs) to establish terminal constraints. To facilitate distributed implementation, we develop estimated neighbor states for formulating DHCBFs and DCLFs, while also devising a bound constraint to limit estimation errors and ensure convergence. Additionally, we provide theoretical guarantees regarding the feasibility and stability of the proposed DSMPC algorithm based on a mild assumption. The effectiveness of the proposed method is evidenced by the simulation results, demonstrating improved performance and reduced computation time compared to existing approaches.

eess.SY↗

arXiv2025

DiffusionGS: Generative Search with Query Conditioned Diffusion in Kuaishou

Personalized search ranking systems are critical for driving engagement and revenue in modern e-commerce and short-video platforms. While existing methods excel at estimating users' broad interests based on the filtered historical behaviors, they typically under-exploit explicit alignment between a user's real-time intent (represented by the user query) and their past actions. In this paper, we propose DiffusionGS, a novel and scalable approach powered by generative models. Our key insight is that user queries can serve as explicit intent anchors to facilitate the extraction of users' immediate interests from long-term, noisy historical behaviors. Specifically, we formulate interest extraction as a conditional denoising task, where the user's query guides a conditional diffusion process to produce a robust, user intent-aware representation from their behavioral sequence. We propose the User-aware Denoising Layer (UDL) to incorporate user-specific profiles into the optimization of attention distribution on the user's past actions. By reframing queries as intent priors and leveraging diffusion-based denoising, our method provides a powerful mechanism for capturing dynamic user interest shifts. Extensive offline and online experiments demonstrate the superiority of DiffusionGS over state-of-the-art methods.

cs.IR↗

arXiv2025

VQL: An End-to-End Context-Aware Vector Quantization Attention for Ultra-Long User Behavior Modeling

In large-scale recommender systems, ultra-long user behavior sequences encode rich signals of evolving interests. Extending sequence length generally improves accuracy, but directly modeling such sequences in production is infeasible due to latency and memory constraints. Existing solutions fall into two categories: (1) top-k retrieval, which truncates the sequence and may discard most attention mass when L >> k; and (2) encoder-based compression, which preserves coverage but often over-compresses and fails to incorporate key context such as temporal gaps or target-aware signals. Neither class achieves a good balance of low-loss compression, context awareness, and efficiency. We propose VQL, a context-aware Vector Quantization Attention framework for ultra-long behavior modeling, with three innovations. (1) Key-only quantization: only attention keys are quantized, while values remain intact; we prove that softmax normalization yields an error bound independent of sequence length, and a codebook loss directly supervises quantization quality. This also enables L-free inference via offline caches. (2) Multi-scale quantization: attention heads are partitioned into groups, each with its own small codebook, which reduces quantization error while keeping cache size fixed. (3) Efficient context injection: static features (e.g., item category, modality) are directly integrated, and relative position is modeled via a separable temporal kernel. All context is injected without enlarging the codebook, so cached representations remain query-independent. Experiments on three large-scale datasets (KuaiRand-1K, KuaiRec, TMALL) show that VQL consistently outperforms strong baselines, achieving higher accuracy while reducing inference latency, establishing a new state of the art in balancing accuracy and efficiency for ultra-long sequence recommendation.

cs.IR↗

arXiv2025

Probabilistic Inference for Datalog with Correlated Inputs

Probabilistic extensions of logic programming languages, such as ProbLog, integrate logical reasoning with probabilistic inference to evaluate probabilities of output relations; however, prior work does not account for potential statistical correlations among input facts. This paper introduces Praline, a new extension to Datalog designed for precise probabilistic inference in the presence of (partially known) input correlations. We formulate the inference task as a constrained optimization problem, where the solution yields sound and precise probability bounds for output facts. However, due to the complexity of the resulting optimization problem, this approach alone often does not scale to large programs. To address scalability, we propose a more efficient $δ$-exact inference algorithm that leverages constraint solving, static analysis, and iterative refinement. Our empirical evaluation on challenging real-world benchmarks, including side-channel analysis, demonstrates that our method not only scales effectively but also delivers tight probability bounds.

cs.PL↗

arXiv2025

TransLLM: A Unified Multi-Task Foundation Framework for Urban Transportation via Learnable Prompting

Urban transportation systems encounter diverse challenges across multiple tasks, such as traffic forecasting, electric vehicle (EV) charging demand prediction, and taxi dispatch. Existing approaches suffer from two key limitations: small-scale deep learning models are task-specific and data-hungry, limiting their generalizability across diverse scenarios, while large language models (LLMs), despite offering flexibility through natural language interfaces, struggle with structured spatiotemporal data and numerical reasoning in transportation domains. To address these limitations, we propose TransLLM, a unified foundation framework that integrates spatiotemporal modeling with large language models through learnable prompt composition. Our approach features a lightweight spatiotemporal encoder that captures complex dependencies via dilated temporal convolutions and dual-adjacency graph attention networks, seamlessly interfacing with LLMs through structured embeddings. A novel instance-level prompt routing mechanism, trained via reinforcement learning, dynamically personalizes prompts based on input characteristics, moving beyond fixed task-specific templates. The framework operates by encoding spatiotemporal patterns into contextual representations, dynamically composing personalized prompts to guide LLM reasoning, and projecting the resulting representations through specialized output layers to generate task-specific predictions. Experiments across seven datasets and three tasks demonstrate the exceptional effectiveness of TransLLM in both supervised and zero-shot settings. Compared to ten baseline models, it delivers competitive performance on both regression and planning problems, showing strong generalization and cross-task adaptability. Our code is available at https://github.com/BiYunying/TransLLM.

cs.CL↗

arXiv2025

Quantum State Preparation by Improved MPS Method

Efficient encoding of classical information plays a fundamental role in numerous practical quantum algorithms. However, the preparation of an arbitrary amplitude-encoded state has been proven to be time-consuming, and its deployment on current noisy devices can be challenging. In this work, we propose an improved Matrix Product State(MPS) method preparation protocol with an exponential reduction on the circuit depth, as well as topological adaptability. By refined utilization of the disentangling principle, we also reduce approximately 33% two-qubit gate count. To validate our method, we study various families of functions and distributions with provably bounded MPS rank. Numerical experiments show that our method significantly reduces circuit depth while achieving higher fidelity for states arising in financial and other applications.

quant-ph↗

arXiv2025

The invariance and non-decreasing expectation of an evolutionary path characteristic under weak selection

Fisher's fundamental theorem of natural selection states that the rate of change in a population's mean fitness equals its additive genetic variance in fitness. This implies that mean fitness should not decline in a constant environment, thereby positioning it as an indicator of evolutionary progression. However, this theorem has been shown to lack universality. Here, we derive the Fokker-Planck equation that describes the stochastic frequency dynamics of two phenotypes in a large population under weak selection and genetic drift, and develop a path integral formulation that characterizes the probability density of phenotypic frequency. Our formulation identifies that, under both selection and genetic drift, the ratio of the probability density of adaptive traits (e.g., phenotypic frequency) to that under neutrality represents a time-invariant evolutionary path characteristic. This ratio quantifies the cumulative effect of directional selection on the evolutionary process compared to genetic drift. Importantly, the expected value of this ratio does not decline over time. In the presence of fitness variance, the effect of directional selection on expected phenotypic changes accumulates over time, diverging progressively from paths shaped solely by genetic drift. The expectation of this time-invariant ratio thus offers a robust and informative alternative to mean fitness as a measure of progression in stochastic evolutionary dynamics.

q-bio.PE↗

arXiv2025

SeamlessFlow: A Trainer Agent Isolation RL Framework Achieving Bubble-Free Pipelines via Tag Scheduling

We introduce SeamlessFlow, a server based reinforcement learning (RL) framework that addresses two core challenges in industrial scale RL: (1) decoupling RL training from the complex execution flow of agents; (2) maximizing GPU utilization with minimal idle time while preserving the stability and scalability required for large-scale deployments. First, SeamlessFlow introduces a data plane that decouples the RL trainer from diverse, complex agent implementations while sustaining high throughput. A central trajectory manager maintains complete interaction histories and supports partial rollout, allowing rollout to pause for weight updates and resume seamlessly, keeping agents unaware of service interruptions. Second, we propose a tag driven scheduling paradigm that abstracts hardware into capability tagged resources, unifying colocated and disaggregated architectures. Based on this, SeamlessFlow introduces a spatiotemporal multiplexing pipeline that dynamically reassigns idle training nodes to rollout in a train rollout separated setup, eliminating pipeline bubbles and fully exploiting heterogeneous cluster resources. By combining these innovations, SeamlessFlow delivers both stability and high performance, making it well suited for multi agent, long horizon, and other complex RL tasks.

cs.LG↗

arXiv2025

Physics-informed Kolmogorov-Arnold Network with Chebyshev Polynomials for Fluid Mechanics

Solving partial differential equations (PDEs) is essential in scientific forecasting and fluid dynamics. Traditional approaches often incur expensive computational costs and trade-offs in efficiency and accuracy. Recent deep neural networks have improved the accuracy but require high-quality training data. Physics-informed neural networks (PINNs) effectively integrate physical laws to reduce the data reliance in limited sample scenarios. A novel machine-learning framework, Chebyshev physics-informed Kolmogorov--Arnold network (ChebPIKAN), is proposed to integrate the robust architectures of Kolmogorov--Arnold networks (KAN) with physical constraints to enhance the calculation accuracy of PDEs for fluid mechanics. We study the fundamentals of KAN, take advantage of the orthogonality of Chebyshev polynomial basis functions in spline fitting, and integrate physics-informed loss functions that are tailored to specific PDEs in fluid dynamics, including Allen--Cahn equation, nonlinear Burgers equation, Helmholtz equations, Kovasznay flow, cylinder wake flow, and cavity flow. Extensive experiments demonstrate that the proposed ChebPIKAN model significantly outperforms the standard KAN architecture in solving various PDEs by effectively embedding essential physical information. These results indicate that augmenting KAN with physical constraints can alleviate the overfitting issues of KAN and improve the extrapolation performance. Consequently, this study highlights the potential of ChebPIKAN as a powerful tool in computational fluid dynamics and propose a path toward fast and reliable predictions in fluid mechanics and beyond.

physics.flu-dyn↗

arXiv2025

Multi-layer 5D Optical Data Storage: Mathematical Modeling and Deep Learning-Based Reconstruction of Birefringent Parameters

Five-dimensional (5D) optical data storage has emerged as a promising technology for ultra-high-density, long-term data archiving. However, its practical realization is hindered by noise and interference during data readout. In this work, we develop a high-precision mathematical model for multi-layer 5D optical storage, grounded in the Jones matrix framework, to accurately capture polarization transformations induced by stacked birefringent nanostructures. Building on this model, we propose a 20-frame FiLM-conditioned U-Net algorithm to reconstruct birefringence parameters--specifically, slow-axis orientation and retardance magnitude-directly from measured intensity patterns. Trained on both ideal and noisy datasets, the network demonstrates robust reconstruction performance under challenging measurement conditions. Compared with conventional frame-based retrieval approaches, our method achieves over an order-of-magnitude improvement in reconstruction accuracy. The proposed model and algorithm can be readily integrated into existing 5D optical readout systems, offering both a solid theoretical foundation and practical tools for precise data recovery.

physics.optics↗

arXiv2025

Prefill-Decode Aggregation or Disaggregation? Unifying Both for Goodput-Optimized LLM Serving

An ongoing debate considers whether prefill-decode (PD) aggregation or disaggregation is superior for serving large language models (LLMs). This has driven optimizations for both approaches, each showing distinct advantages. This paper compares PD aggregation and disaggregation, showing that each excels under different service-level objectives (SLOs): aggregation is optimal for tight time-to-first-token (TTFT) and relaxed time-per-output-token (TPOT), while disaggregation excels for strict TPOT and relaxed TTFT. However, under balanced TTFT and TPOT SLOs, neither approach delivers optimal goodput. This paper proposes TaiChi, an LLM serving system that unifies PD disaggregation and aggregation for optimal goodput under any combination of TTFT and TPOT SLOs. TaiChi uses a unified disaggregation-aggregation architecture with differentiated-capability GPU instances: prefill-heavy (fast prefill, high-interference decode) and decode-heavy (low-interference decode, slow prefill). Three configurable sliders control the ratio between these instances and their chunk sizes. TaiChi adapts to various SLO regimes by adjusting sliders. When TTFT constraints are tight, TaiChi resembles a PD aggregation configuration; when TPOT dominates, it adapts toward PD disaggregation. Crucially, under balanced SLOs, TaiChi enables a hybrid mode for superior goodput. The key innovation behind this hybrid mode is latency shifting: selectively reallocating GPU resources from requests that meet SLOs to those at risk of violation, maximizing the number of SLO-satisfied requests. This fine-grained latency shifting is orchestrated by two scheduling mechanisms: flowing decode scheduling to control TPOTs and length-aware prefill scheduling to manage TTFTs, which jointly optimize request assignment. Our experiments show TaiChi improves goodput by up to 77% over state-of-the-art systems under balanced TTFT and TPOT SLOs.

cs.DC↗

arXiv2025

Physics-informed Fourier Basis Neural Network for Fluid Mechanics

Solving partial differential equations (PDEs) is an important yet challenging task in fluid mechanics. In this study, we embed an improved Fourier series into neural networks and propose a physics-informed Fourier basis neural network (FBNN) by incorporating physical information to solve canonical PDEs in fluid mechanics. The results demonstrated that the proposed framework exhibits a strong nonlinear fitting capability and exceptional periodic modeling performance. In particular, our model shows significant advantages for the Burgers equation with discontinuous solutions and Helmholtz equation with strong periodicity. By directly introducing sparse distributed data to reconstruct the entire flow field, we further intuitively validated the direct superiority of FBNN over conventional artificial neural networks (ANN) as well as the benefits of incorporating physical information into the network. By adjusting the activation functions of networks and comparing with an ANN and conventional physics-informed neural network, we proved that performance of the proposed FBNN architecture is not highly sensitive to the choice of activation functions. The nonlinear fitting capability of FBNN avoids excessive reliance on activation functions, thereby mitigating the risk of suboptimal outcomes or training failures stemming from unsuitable activation function choices.hese results highlightthe potential of PIFBNN as a powerful tool in computational fluid dynamics.

physics.flu-dyn↗

arXiv2025

I2CR: Intra- and Inter-modal Collaborative Reflections for Multimodal Entity Linking

Multimodal entity linking plays a crucial role in a wide range of applications. Recent advances in large language model-based methods have become the dominant paradigm for this task, effectively leveraging both textual and visual modalities to enhance performance. Despite their success, these methods still face two challenges, including unnecessary incorporation of image data in certain scenarios and the reliance only on a one-time extraction of visual features, which can undermine their effectiveness and accuracy. To address these challenges, we propose a novel LLM-based framework for the multimodal entity linking task, called Intra- and Inter-modal Collaborative Reflections. This framework prioritizes leveraging text information to address the task. When text alone is insufficient to link the correct entity through intra- and inter-modality evaluations, it employs a multi-round iterative strategy that integrates key visual clues from various aspects of the image to support reasoning and enhance matching accuracy. Extensive experiments on three widely used public datasets demonstrate that our framework consistently outperforms current state-of-the-art methods in the task, achieving improvements of 3.2%, 5.1%, and 1.6%, respectively. Our code is available at https://github.com/ziyan-xiaoyu/I2CR/.

cs.CV↗

arXiv2025

Towards Temporal-Aware Multi-Modal Retrieval Augmented Generation in Finance

Finance decision-making often relies on in-depth data analysis across various data sources, including financial tables, news articles, stock prices, etc. In this work, we introduce FinTMMBench, the first comprehensive benchmark for evaluating temporal-aware multi-modal Retrieval-Augmented Generation (RAG) systems in finance. Built from heterologous data of NASDAQ 100 companies, FinTMMBench offers three significant advantages. 1) Multi-modal Corpus: It encompasses a hybrid of financial tables, news articles, daily stock prices, and visual technical charts as the corpus. 2) Temporal-aware Questions: Each question requires the retrieval and interpretation of its relevant data over a specific time period, including daily, weekly, monthly, quarterly, and annual periods. 3) Diverse Financial Analysis Tasks: The questions involve 10 different financial analysis tasks designed by domain experts, including information extraction, trend analysis, sentiment analysis and event detection, etc. We further propose a novel TMMHybridRAG method, which first leverages LLMs to convert data from other modalities (e.g., tabular, visual and time-series data) into textual format and then incorporates temporal information in each node when constructing graphs and dense indexes. Its effectiveness has been validated in extensive experiments, but notable gaps remain, highlighting the challenges presented by our FinTMMBench.

q-fin.CP↗

arXiv2025

Technical Report of TeleChat2, TeleChat2.5 and T1

We introduce the latest series of TeleChat models: \textbf{TeleChat2}, \textbf{TeleChat2.5}, and \textbf{T1}, offering a significant upgrade over their predecessor, TeleChat. Despite minimal changes to the model architecture, the new series achieves substantial performance gains through enhanced training strategies in both pre-training and post-training stages. The series begins with \textbf{TeleChat2}, which undergoes pretraining on 10 trillion high-quality and diverse tokens. This is followed by Supervised Fine-Tuning (SFT) and Direct Preference Optimization (DPO) to further enhance its capabilities. \textbf{TeleChat2.5} and \textbf{T1} expand the pipeline by incorporating a continual pretraining phase with domain-specific datasets, combined with reinforcement learning (RL) to improve performance in code generation and mathematical reasoning tasks. The \textbf{T1} variant is designed for complex reasoning, supporting long Chain-of-Thought (CoT) reasoning and demonstrating substantial improvements in mathematics and coding. In contrast, \textbf{TeleChat2.5} prioritizes speed, delivering rapid inference. Both flagship models of \textbf{T1} and \textbf{TeleChat2.5} are dense Transformer-based architectures with 115B parameters, showcasing significant advancements in reasoning and general task performance compared to the original TeleChat. Notably, \textbf{T1-115B} outperform proprietary models such as OpenAI's o1-mini and GPT-4o. We publicly release \textbf{TeleChat2}, \textbf{TeleChat2.5} and \textbf{T1}, including post-trained versions with 35B and 115B parameters, to empower developers and researchers with state-of-the-art language models tailored for diverse applications.

cs.CL↗

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↗