arXiv subjects

Yifei Zheng

Publications and source records attributed to Yifei Zheng.

7 recordsLinked to original sources

arXiv2026

Listen, See and Track: Spatio-Temporal Audio-Visual Sound Event Reasoning for Omni-Modal Language Models

Understanding dynamic sound sources requires jointly determining what produces a sound, where the source is located, and how it moves over time. Yet existing audio-language models often represent clips as global acoustic events, while vision-language models lack the spatial audio cues needed to localize and track individual sources. To evaluate this missing capability, we introduce ST-OmniQA, a spatio-temporal audio-visual question-answering benchmark built from panoramic videos paired with synchronized first-order Ambisonics (FOA) audio of moving sound sources. It contains 40K videos and 400K question-answer pairs organized into four capability levels covering sound-event recognition, direction of arrival, source distance, motion trajectories, and temporally grounded audio-visual reasoning. Building on this benchmark, we propose ST-Omni-R1, which integrates FOA-derived semantic and trajectory representations with panoramic visual context and is trained through progressive curriculum learning and reasoning-tree reinforcement learning. ST-Omni-R1 achieves 77.83\% average semantic accuracy across the four levels, compared with 37.28\% for the best evaluated baseline. Results on three public spatial-audio benchmarks further indicate that its learned spatial and motion representations transfer beyond ST-OmniQA.

cs.AI↗

arXiv2026

L-SDPPO: Policy Optimization of Spiking Diffusion Policy for Intra-vehicular Robotic Manipulation

Intra-vehicular robots in spacecraft help reduce astronaut workload and improve mission efficiency. Recent research focuses on using deep learning methods to achieve the acute control required for operations in these complex environments. However, objects exhibit unpredictable, unconstrained drift without gravitational damping. These factors demand robustness against complex multimodal action distributions. Diffusion policies (DP) can model these complex actions, but their iterative sampling process consumes too much energy for the limited power budgets of spacecraft. We therefore propose a low-energy intra-vehicular robotic manipulation framework, L-SDPPO, in which the Spiking Diffusion Policy (SDP) is optimized with a reinforcement learning (RL) algorithm. Furthermore, to address the insufficient perception of dynamic spatiotemporal features in microgravity, we propose the statedependent latency injection (SDLI) mechanism, which mimics biological neural delays to dynamically regulate the timing of input information. Evaluation on five representative intra-vehicular daily tasks (e.g., hatch opening and precision container capping) shows that our method consistently achieves higher success rates and lower energy consumption, compared to the state-of-the-art robotic manipulation methods. These results demonstrate our method is a viable intra-vehicular robotic manipulation method.

cs.RO↗

arXiv2026

Near Real-time Full-wave Inverse Design of Electromagnetic Devices

Inverse design enables automating the discovery and optimization of devices achieving performance significantly exceeding that of traditional human-engineered designs. However, existing methodologies to inverse-design electromagnetic devices require computationally expensive and time-consuming full-wave electromagnetic simulation at each iteration or generation of large datasets for training neural-network surrogate models. This work introduces the Precomputed Numerical Green Function method, an approach for ultrafast electromagnetic inverse design. The static components of the design are incorporated into a numerical Green function obtained from a single fully-parallelized precomputation step, reducing the cost of evaluating candidate designs during optimization to only being proportional to the size of the region under modification. A low-rank matrix update technique is introduced that further decreases the cost of the method to milliseconds per iteration without any approximations or compromises in accuracy. This method is shown to have linear time complexity, reducing the total runtime for an inverse design by several orders of magnitude compared to using conventional electromagnetics solvers. The design examples considered demonstrate speedups of up to 16,000x, shortening the design process from multiple days to weeks down to minutes. The approach enables practical and ultrafast design of complex structures that are prohibitively time-consuming for prior inverse design methods.

physics.comp-ph↗

arXiv2025

ActionFlow: A Pipelined Action Acceleration for Vision Language Models on Edge

Vision-Language-Action (VLA) models have emerged as a unified paradigm for robotic perception and control, enabling emergent generalization and long-horizon task execution. However, their deployment in dynamic, real-world environments is severely hin dered by high inference latency. While smooth robotic interaction requires control frequencies of 20 to 30 Hz, current VLA models typi cally operate at only 3-5 Hz on edge devices due to the memory bound nature of autoregressive decoding. Existing optimizations often require extensive retraining or compromise model accuracy. To bridge this gap, we introduce ActionFlow, a system-level inference framework tailored for resource-constrained edge plat forms. At the core of ActionFlow is a Cross-Request Pipelin ing strategy, a novel scheduler that redefines VLA inference as a macro-pipeline of micro-requests. The strategy intelligently batches memory-bound Decode phases with compute-bound Prefill phases across continuous time steps to maximize hardware utilization. Furthermore, to support this scheduling, we propose a Cross Request State Packed Forward operator and a Unified KV Ring Buffer, which fuse fragmented memory operations into efficient dense computations. Experimental results demonstrate that ActionFlow achieves a 2.55x improvement in FPS on the OpenVLA-7B model without retraining, enabling real-time dy namic manipulation on edge hardware. Our work is available at https://anonymous.4open.science/r/ActionFlow-1D47.

cs.AI↗

arXiv2025

A Review of Generative AI in Computer Science Education: Challenges and Opportunities in Accuracy, Authenticity, and Assessment

This paper surveys the use of Generative AI tools, such as ChatGPT and Claude, in computer science education, focusing on key aspects of accuracy, authenticity, and assessment. Through a literature review, we highlight both the challenges and opportunities these AI tools present. While Generative AI improves efficiency and supports creative student work, it raises concerns such as AI hallucinations, error propagation, bias, and blurred lines between AI-assisted and student-authored content. Human oversight is crucial for addressing these concerns. Existing literature recommends adopting hybrid assessment models that combine AI with human evaluation, developing bias detection frameworks, and promoting AI literacy for both students and educators. Our findings suggest that the successful integration of AI requires a balanced approach, considering ethical, pedagogical, and technical factors. Future research may explore enhancing AI accuracy, preserving academic integrity, and developing adaptive models that balance creativity with precision.

cs.CY↗

arXiv2025

Full Scaling Automation for Sustainable Development of Green Data Centers

The rapid rise in cloud computing has resulted in an alarming increase in data centers' carbon emissions, which now accounts for >3% of global greenhouse gas emissions, necessitating immediate steps to combat their mounting strain on the global climate. An important focus of this effort is to improve resource utilization in order to save electricity usage. Our proposed Full Scaling Automation (FSA) mechanism is an effective method of dynamically adapting resources to accommodate changing workloads in large-scale cloud computing clusters, enabling the clusters in data centers to maintain their desired CPU utilization target and thus improve energy efficiency. FSA harnesses the power of deep representation learning to accurately predict the future workload of each service and automatically stabilize the corresponding target CPU usage level, unlike the previous autoscaling methods, such as Autopilot or FIRM, that need to adjust computing resources with statistical models and expert knowledge. Our approach achieves significant performance improvement compared to the existing work in real-world datasets. We also deployed FSA on large-scale cloud computing clusters in industrial data centers, and according to the certification of the China Environmental United Certification Center (CEC), a reduction of 947 tons of carbon dioxide, equivalent to a saving of 1538,000 kWh of electricity, was achieved during the Double 11 shopping festival of 2022, marking a critical step for our company's strategic goal towards carbon neutrality by 2030.

cs.DC↗

arXiv2021

Banach Intermediate Spaces for Gaussian Fréchet Spaces

In this article, we show that every centered Gaussian measure on an infinite dimensional separable Fréchet space $X$ over $\mathbb R$ admits some full measure Banach intermediate space between $X$ and its Cameron-Martin space. We provide a way of generating such spaces and, by showing a partial converse, give a characterization of Banach intermediate spaces. Finally, we show an example of constructing an $α$-Hölder intermediate space in the space of continuous functions, $\mathcal C_0[0, 1]$ with the classical Wiener measure.

math.FA↗