arXiv subjects

Xinyu Zhang

Publications and source records attributed to Xinyu Zhang.

At least 181 records · Page 10Linked to original sources

arXiv2025

RefineCoder: Iterative Improving of Large Language Models via Adaptive Critique Refinement for Code Generation

Code generation has attracted increasing attention with the rise of Large Language Models (LLMs). Many studies have developed powerful code LLMs by synthesizing code-related instruction data and applying supervised fine-tuning. However, these methods are limited by teacher model distillation and ignore the potential of iterative refinement by self-generated code. In this paper, we propose Adaptive Critique Refinement (ACR), which enables the model to refine itself by self-generated code and external critique, rather than directly imitating the code responses of the teacher model. Concretely, ACR includes a composite scoring system with LLM-as-a-Judge to evaluate the quality of code responses and a selective critique strategy with LLM-as-a-Critic to critique self-generated low-quality code responses. We develop the RefineCoder series by iteratively applying ACR, achieving continuous performance improvement on multiple code generation benchmarks. Compared to the baselines of the same size, our proposed RefineCoder series can achieve comparable or even superior performance using less data.

cs.CL↗

arXiv2025

InverTwin: Solving Inverse Problems via Differentiable Radio Frequency Digital Twin

Digital twins (DTs), virtual simulated replicas of physical scenes, are transforming various industries. However, their potential in radio frequency (RF) sensing applications has been limited by the unidirectional nature of conventional RF simulators. In this paper, we present InverTwin, an optimization-driven framework that creates RF digital twins by enabling bidirectional interaction between virtual and physical realms. InverTwin overcomes the fundamental differentiability challenges of RF optimization problems through novel design components, including path-space differentiation to address discontinuity in complex simulation functions, and a radar surrogate model to mitigate local non-convexity caused by RF signal periodicity. These techniques enable smooth gradient propagation and robust optimization of the DT model. Our implementation and experiments demonstrate InverTwin's versatility and effectiveness in augmenting both data-driven and model-driven RF sensing systems for DT reconstruction.

eess.SP↗

arXiv2025

Progressive Bird's Eye View Perception for Safety-Critical Autonomous Driving: A Comprehensive Survey

Bird's-Eye-View (BEV) perception has become a foundational paradigm in autonomous driving, enabling unified spatial representations that support robust multi-sensor fusion and multi-agent collaboration. As autonomous vehicles transition from controlled environments to real-world deployment, ensuring the safety and reliability of BEV perception in complex scenarios - such as occlusions, adverse weather, and dynamic traffic - remains a critical challenge. This survey provides the first comprehensive review of BEV perception from a safety-critical perspective, systematically analyzing state-of-the-art frameworks and implementation strategies across three progressive stages: single-modality vehicle-side, multimodal vehicle-side, and multi-agent collaborative perception. Furthermore, we examine public datasets encompassing vehicle-side, roadside, and collaborative settings, evaluating their relevance to safety and robustness. We also identify key open-world challenges - including open-set recognition, large-scale unlabeled data, sensor degradation, and inter-agent communication latency - and outline future research directions, such as integration with end-to-end autonomous driving systems, embodied intelligence, and large language models.

cs.RO↗

arXiv2025

A Review and Classification of Model Uncertainty

Model uncertainty is a crucial issue in statistics, econometrics and machine learning, yet its definition remains ambiguous and is subject to various interpretations in the literature. So far, there has not been a universally accepted definition of model uncertainty. We review different understandings of model uncertainty and categorize them into three distinct types: uncertainty about the true model, model selection uncertainty, and model selection instability. We further offer interpretations and examples for a better illustration of these definitions. We also discuss the potential consequences of neglecting model uncertainty in the process of conducting statistical inference, and provide effective solutions to these problems. Our aim is to help researchers better understand the concept of model uncertainty and obtain valid statistical inference results on the premise of its existence.

stat.ME↗

arXiv2025

Mode-Aware Non-Linear Tucker Autoencoder for Tensor-based Unsupervised Learning

High-dimensional data, particularly in the form of high-order tensors, presents a major challenge in self-supervised learning. While MLP-based autoencoders (AE) are commonly employed, their dependence on flattening operations exacerbates the curse of dimensionality, leading to excessively large model sizes, high computational overhead, and challenging optimization for deep structural feature capture. Although existing tensor networks alleviate computational burdens through tensor decomposition techniques, most exhibit limited capability in learning non-linear relationships. To overcome these limitations, we introduce the Mode-Aware Non-linear Tucker Autoencoder (MA-NTAE). MA-NTAE generalized classical Tucker decomposition to a non-linear framework and employs a Pick-and-Unfold strategy, facilitating flexible per-mode encoding of high-order tensors via recursive unfold-encode-fold operations, effectively integrating tensor structural priors. Notably, MA-NTAE exhibits linear growth in computational complexity with tensor order and proportional growth with mode dimensions. Extensive experiments demonstrate MA-NTAE's performance advantages over standard AE and current tensor networks in compression and clustering tasks, which become increasingly pronounced for higher-order, higher-dimensional tensors.

cs.LG↗

arXiv2025

Sufficiency-principled Transfer Learning via Model Averaging

When the transferable set is unknowable, transfering informative knowledge as much as possible\textemdash a principle we refer to as \emph{sufficiency}, becomes crucial for enhancing transfer learning effectiveness. However, existing transfer learning methods not only overlook the sufficiency principle, but also rely on restrictive single-similarity assumptions (\eg individual or combinatorial similarity), leading to suboptimal performance. To address these limitations, we propose a sufficiency-principled transfer learning framework via unified model averaging algorithms, accommodating both individual and combinatorial similarities. Theoretically, we establish the asymptotic/high-probability optimality, enhanced convergence rate and asymptotic normality for multi-source linear regression models with a diverging number of parameters, achieving sufficiency, robustness to negative transfer, privacy protection and feasible statistical inference. Extensive simulations and an empirical data analysis of Beijing housing rental data demonstrate the promising superiority of our framework over conventional alternatives.

stat.ME↗

arXiv2025

P3: Prompts Promote Prompting

Current large language model (LLM) applications often employ multi-component prompts, comprising both system and user prompts, to guide model behaviors. While recent advancements have demonstrated the efficacy of automatically optimizing either the system or user prompt to boost performance, such unilateral approaches often yield suboptimal outcomes due to the interdependent nature of these components. In this work, we introduce P3, a novel self-improvement framework that concurrently optimizes both system and user prompts through an iterative process. The offline optimized prompts are further leveraged to promote online prompting by performing query-dependent prompt optimization. Extensive experiments on general tasks (e.g., Arena-hard and Alpaca-eval) and reasoning tasks (e.g., GSM8K and GPQA) demonstrate that P3 achieves superior performance in the realm of automatic prompt optimization. Our results highlight the effectiveness of a holistic optimization strategy in enhancing LLM performance across diverse domains.

cs.CL↗

arXiv2025

Entropy Loss: An Interpretability Amplifier of 3D Object Detection Network for Intelligent Driving

With the increasing complexity of the traffic environment, the significance of safety perception in intelligent driving is intensifying. Traditional methods in the field of intelligent driving perception rely on deep learning, which suffers from limited interpretability, often described as a "black box." This paper introduces a novel type of loss function, termed "Entropy Loss," along with an innovative training strategy. Entropy Loss is formulated based on the functionality of feature compression networks within the perception model. Drawing inspiration from communication systems, the information transmission process in a feature compression network is expected to demonstrate steady changes in information volume and a continuous decrease in information entropy. By modeling network layer outputs as continuous random variables, we construct a probabilistic model that quantifies changes in information volume. Entropy Loss is then derived based on these expectations, guiding the update of network parameters to enhance network interpretability. Our experiments indicate that the Entropy Loss training strategy accelerates the training process. Utilizing the same 60 training epochs, the accuracy of 3D object detection models using Entropy Loss on the KITTI test set improved by up to 4.47\% compared to models without Entropy Loss, underscoring the method's efficacy. The implementation code is available at https://github.com/yhbcode000/Eloss-Interpretability.

cs.CV↗

arXiv2025

Hear-Your-Click: Interactive Object-Specific Video-to-Audio Generation

Video-to-audio (V2A) generation shows great potential in fields such as film production. Despite significant advances, current V2A methods relying on global video information struggle with complex scenes and generating audio tailored to specific objects. To address these limitations, we introduce Hear-Your-Click, an interactive V2A framework enabling users to generate sounds for specific objects by clicking on the frame. To achieve this, we propose Object-aware Contrastive Audio-Visual Fine-tuning (OCAV) with a Mask-guided Visual Encoder (MVE) to obtain object-level visual features aligned with audio. Furthermore, we tailor two data augmentation strategies, Random Video Stitching (RVS) and Mask-guided Loudness Modulation (MLM), to enhance the model's sensitivity to segmented objects. To measure audio-visual correspondence, we designed a new evaluation metric, the CAV score. Extensive experiments demonstrate that our framework offers more precise control and improves generation performance across various metrics. Project Page: https://github.com/SynapGrid/Hear-Your-Click

cs.CV↗

arXiv2025

VST-Pose: A Velocity-Integrated Spatiotem-poral Attention Network for Human WiFi Pose Estimation

WiFi-based human pose estimation has emerged as a promising non-visual alternative approaches due to its pene-trability and privacy advantages. This paper presents VST-Pose, a novel deep learning framework for accurate and continuous pose estimation using WiFi channel state information. The proposed method introduces ViSTA-Former, a spatiotemporal attention backbone with dual-stream architecture that adopts a dual-stream architecture to separately capture temporal dependencies and structural relationships among body joints. To enhance sensitivity to subtle human motions, a velocity modeling branch is integrated into the framework, which learns short-term keypoint dis-placement patterns and improves fine-grained motion representation. We construct a 2D pose dataset specifically designed for smart home care scenarios and demonstrate that our method achieves 92.2% accuracy on the PCK@50 metric, outperforming existing methods by 8.3% in PCK@50 on the self-collected dataset. Further evaluation on the public MMFi dataset confirms the model's robustness and effectiveness in 3D pose estimation tasks. The proposed system provides a reliable and privacy-aware solution for continuous human motion analysis in indoor environments. Our codes are available in https://github.com/CarmenQing/VST-Pose.

cs.CV↗

arXiv2025

SkipVAR: Accelerating Visual Autoregressive Modeling via Adaptive Frequency-Aware Skipping

Recent studies on Visual Autoregressive (VAR) models have highlighted that high-frequency components, or later steps, in the generation process contribute disproportionately to inference latency. However, the underlying computational redundancy involved in these steps has yet to be thoroughly investigated. In this paper, we conduct an in-depth analysis of the VAR inference process and identify two primary sources of inefficiency: step redundancy and unconditional branch redundancy. To address step redundancy, we propose an automatic step-skipping strategy that selectively omits unnecessary generation steps to improve efficiency. For unconditional branch redundancy, we observe that the information gap between the conditional and unconditional branches is minimal. Leveraging this insight, we introduce unconditional branch replacement, a technique that bypasses the unconditional branch to reduce computational cost. Notably, we observe that the effectiveness of acceleration strategies varies significantly across different samples. Motivated by this, we propose SkipVAR, a sample-adaptive framework that leverages frequency information to dynamically select the most suitable acceleration strategy for each instance. To evaluate the role of high-frequency information, we introduce high-variation benchmark datasets that test model sensitivity to fine details. Extensive experiments show SkipVAR achieves over 0.88 average SSIM with up to 1.81x overall acceleration and 2.62x speedup on the GenEval benchmark, maintaining model quality. These results confirm the effectiveness of frequency-aware, training-free adaptive acceleration for scalable autoregressive image generation. Our code is available at https://github.com/fakerone-li/SkipVAR and has been publicly released.

cs.CV↗

arXiv2025

Failure Forecasting Boosts Robustness of Sim2Real Rhythmic Insertion Policies

This paper addresses the challenges of Rhythmic Insertion Tasks (RIT), where a robot must repeatedly perform high-precision insertions, such as screwing a nut into a bolt with a wrench. The inherent difficulty of RIT lies in achieving millimeter-level accuracy and maintaining consistent performance over multiple repetitions, particularly when factors like nut rotation and friction introduce additional complexity. We propose a sim-to-real framework that integrates a reinforcement learning-based insertion policy with a failure forecasting module. By representing the wrench's pose in the nut's coordinate frame rather than the robot's frame, our approach significantly enhances sim-to-real transferability. The insertion policy, trained in simulation, leverages real-time 6D pose tracking to execute precise alignment, insertion, and rotation maneuvers. Simultaneously, a neural network predicts potential execution failures, triggering a simple recovery mechanism that lifts the wrench and retries the insertion. Extensive experiments in both simulated and real-world environments demonstrate that our method not only achieves a high one-time success rate but also robustly maintains performance over long-horizon repetitive tasks.

cs.RO↗

arXiv2025

Cognitive Exoskeleton: Augmenting Human Cognition with an AI-Mediated Intelligent Visual Feedback

In this paper, we introduce an AI-mediated framework that can provide intelligent feedback to augment human cognition. Specifically, we leverage deep reinforcement learning (DRL) to provide adaptive time pressure feedback to improve user performance in a math arithmetic task. Time pressure feedback could either improve or deteriorate user performance by regulating user attention and anxiety. Adaptive time pressure feedback controlled by a DRL policy according to users' real-time performance could potentially solve this trade-off problem. However, the DRL training and hyperparameter tuning may require large amounts of data and iterative user studies. Therefore, we propose a dual-DRL framework that trains a regulation DRL agent to regulate user performance by interacting with another simulation DRL agent that mimics user cognition behaviors from an existing dataset. Our user study demonstrates the feasibility and effectiveness of the dual-DRL framework in augmenting user performance, in comparison to the baseline group.

cs.HC↗

arXiv2025

Orientation-Aware Sparse Tensor PCA for Efficient Unsupervised Feature Selection

Recently, introducing Tensor Decomposition (TD) techniques into unsupervised feature selection (UFS) has been an emerging research topic. A tensor structure is beneficial for mining the relations between different modes and helps relieve the computation burden. However, while existing methods exploit TD to preserve the data tensor structure, they do not consider the influence of data orientation and thus have difficulty in handling orientation-specific data such as time series. To solve the above problem, we utilize the orientation-dependent tensor-tensor product from Tensor Singular Value Decomposition based on *M-product (T-SVDM) and extend the one-dimensional Sparse Principal Component Analysis (SPCA) to a tensor form. The proposed sparse tensor PCA model can constrain sparsity at the specified mode and yield sparse tensor principal components, enhancing flexibility and accuracy in learning feature relations. To ensure fast convergence and a flexible description of feature correlation, we develop a convex version specially designed for general UFS tasks and propose an efficient slice-by-slice algorithm that performs dual optimization in the transform domain. Experimental results on real-world datasets demonstrate the effectiveness and remarkable computational efficiency of the proposed method for tensor data of diverse structures over the state-of-the-art. When transform axes align with feature distribution patterns, our method is promising for various applications. The codes related to our proposed methods and the experiments are available at https://github.com/zjj20212035/STPCA.git.

cs.LG↗

arXiv2025

HAC-LOCO: Learning Hierarchical Active Compliance Control for Quadruped Locomotion under Continuous External Disturbances

Despite recent remarkable achievements in quadruped control, it remains challenging to ensure robust and compliant locomotion in the presence of unforeseen external disturbances. Existing methods prioritize locomotion robustness over compliance, often leading to stiff, high-frequency motions, and energy inefficiency. This paper, therefore, presents a two-stage hierarchical learning framework that can learn to take active reactions to external force disturbances based on force estimation. In the first stage, a velocity-tracking policy is trained alongside an auto-encoder to distill historical proprioceptive features. A neural network-based estimator is learned through supervised learning, which estimates body velocity and external forces based on proprioceptive measurements. In the second stage, a compliance action module, inspired by impedance control, is learned based on the pre-trained encoder and policy. This module is employed to actively adjust velocity commands in response to external forces based on real-time force estimates. With the compliance action module, a quadruped robot can robustly handle minor disturbances while appropriately yielding to significant forces, thus striking a balance between robustness and compliance. Simulations and real-world experiments have demonstrated that our method has superior performance in terms of robustness, energy efficiency, and safety. Experiment comparison shows that our method outperforms the state-of-the-art RL-based locomotion controllers. Ablation studies are given to show the critical roles of the compliance action module.

cs.RO↗

arXiv2025

Seedream 3.0 Technical Report

We present Seedream 3.0, a high-performance Chinese-English bilingual image generation foundation model. We develop several technical improvements to address existing challenges in Seedream 2.0, including alignment with complicated prompts, fine-grained typography generation, suboptimal visual aesthetics and fidelity, and limited image resolutions. Specifically, the advancements of Seedream 3.0 stem from improvements across the entire pipeline, from data construction to model deployment. At the data stratum, we double the dataset using a defect-aware training paradigm and a dual-axis collaborative data-sampling framework. Furthermore, we adopt several effective techniques such as mixed-resolution training, cross-modality RoPE, representation alignment loss, and resolution-aware timestep sampling in the pre-training phase. During the post-training stage, we utilize diversified aesthetic captions in SFT, and a VLM-based reward model with scaling, thereby achieving outputs that well align with human preferences. Furthermore, Seedream 3.0 pioneers a novel acceleration paradigm. By employing consistent noise expectation and importance-aware timestep sampling, we achieve a 4 to 8 times speedup while maintaining image quality. Seedream 3.0 demonstrates significant improvements over Seedream 2.0: it enhances overall capabilities, in particular for text-rendering in complicated Chinese characters which is important to professional typography generation. In addition, it provides native high-resolution output (up to 2K), allowing it to generate images with high visual quality.

cs.CV↗

arXiv2025

Let Your Video Listen to Your Music!

Aligning the rhythm of visual motion in a video with a given music track is a practical need in multimedia production, yet remains an underexplored task in autonomous video editing. Effective alignment between motion and musical beats enhances viewer engagement and visual appeal, particularly in music videos, promotional content, and cinematic editing. Existing methods typically depend on labor-intensive manual cutting, speed adjustments, or heuristic-based editing techniques to achieve synchronization. While some generative models handle joint video and music generation, they often entangle the two modalities, limiting flexibility in aligning video to music beats while preserving the full visual content. In this paper, we propose a novel and efficient framework, termed MVAA (Music-Video Auto-Alignment), that automatically edits video to align with the rhythm of a given music track while preserving the original visual content. To enhance flexibility, we modularize the task into a two-step process in our MVAA: aligning motion keyframes with audio beats, followed by rhythm-aware video inpainting. Specifically, we first insert keyframes at timestamps aligned with musical beats, then use a frame-conditioned diffusion model to generate coherent intermediate frames, preserving the original video's semantic content. Since comprehensive test-time training can be time-consuming, we adopt a two-stage strategy: pretraining the inpainting module on a small video set to learn general motion priors, followed by rapid inference-time fine-tuning for video-specific adaptation. This hybrid approach enables adaptation within 10 minutes with one epoch on a single NVIDIA 4090 GPU using CogVideoX-5b-I2V as the backbone. Extensive experiments show that our approach can achieve high-quality beat alignment and visual smoothness.

cs.CV↗

arXiv2025

You Only Render Once: Enhancing Energy and Computation Efficiency of Mobile Virtual Reality

Mobile Virtual Reality (VR) is essential to achieving convenient and immersive human-computer interaction and realizing emerging applications such as Metaverse. However, existing VR technologies require two separate renderings of binocular images, causing a significant bottleneck for mobile devices with limited computing capability and power supply. This paper proposes an approach to rendering optimization for mobile VR called EffVR. By utilizing the per-pixel attribute, EffVR can generate binocular VR images from the monocular image through genuinely one rendering, saving half the computation over conventional approaches. Our evaluation indicates that, compared with the state-of-art, EffVRcan save 27% power consumption on average while achieving high binocular image quality (0.9679 SSIM and 34.09 PSNR) in mobile VR applications. Additionally, EffVR can increase the frame rate by 115.2%. These results corroborate EffVRsuperior computation/energy-saving performance, paving the road to a sustainable mobile VR. The source code, demo video, android app, and more are released anonymously at https://yoro-vr.github.io/

cs.GR↗