arXiv subjects

Ziyao Zhang

Publications and source records attributed to Ziyao Zhang.

At least 19 recordsLinked to original sources

arXiv2026

Biologically Inspired Predictive Gating Signed-Dyadic Coding for Mandarin Keyword Spotting

In continuous spiking keyword spotting, temporally correlated acoustic frames can repeatedly trigger memory access and neural computation at a dense input interface, even when hidden activity is sparse. Predictive processing and sensory adaptation suggest retaining predictable input locally and responding selectively to deviations. Inspired by this mechanism, we propose Predictive Gating Signed-Dyadic Coding (PG). PG maintains a local prediction for each input channel and transmits an event only when the residual exceeds a learned threshold. Each event carries a sign and a finite power-of-two magnitude, and the receiver reconstructs it with fixed-point shifts and additions. In Mandarin keyword spotting, INPUT-PG reaches 90.00% accuracy while reducing mean SRAM reads by 33.87% and inference cycles by 32.77%. PG provides a hardware interface for research on low-power continuous speech processing and event-based sensory systems.

cs.SD↗

arXiv2026

Power-of-Two Leaky Integrate-and-Fire Neurons for Event-Driven Vision-Language-Action Inference

Vision-language-action (VLA) models have demonstrated strong capabilities in robotic manipulation, yet their reliance on dense, continuous-valued computation presents a fundamental challenge for energy-constrained deployment. Inspired by the sparse, event-driven signaling and temporal integration mechanisms of biological neurons, a promising alternative is to introduce spiking computation into VLA models. However, enabling such computation without restructuring pretrained policies or compromising closed-loop control remains an open challenge. To address this challenge, we introduce the Power-of-Two Leaky Integrate-and-Fire (PoT-LIF) neuron, which emits signed spikes with power-of-two amplitudes. By combining magnitude-coded spike events with residual-preserving membrane dynamics, PoT-LIF enables continuous information to be progressively represented through discrete spikes. The Heartbeat formulation further localizes this process within weighted projections, bridging biologically inspired spiking computation and pretrained VLA architectures without modifying their external inference interfaces. We integrate PoT-LIF into the vision projection, vision-language projection, and action projection of $π_{0.5}$ and evaluate both operating modes over 2,000 paired closed-loop LIBERO episodes per configuration. Standard-mode PoT-LIF VLA achieves success rates of 96.90% and 96.35% at $T = 1$ and $T = 2$, respectively, while Heartbeat mode achieves 97.25% and 96.90% at $T = 2$ and $T = 4$, compared with 96.70% for the original continuous-activation $π_{0.5}$ baseline. At $T = 1$ with a 60.01% firing rate, PoT-LIF uses approximately 1/7.67 of the dense-reference arithmetic energy, corresponding to a modeled reduction of 86.95%.

cs.AI↗

arXiv2026

GrowPage: On-Demand KV Budgeting for Efficient LLM Reasoning Serving

Long-output reasoning has made the key--value (KV) cache a critical memory bottleneck for efficient LLM serving. Existing KV compression methods usually rely on a predefined per-request budget and adjust only which KV states are retained, leaving the total capacity fixed throughout decoding. However, reasoning workloads exhibit substantial demand variation: different requests require different KV capacities, and the attention demand of an individual request evolves during generation. We introduce \textbf{GrowPage}, an on-demand KV budgeting framework that treats KV capacity as a runtime resource. GrowPage maintains lightweight dual-timescale query summaries to capture recent and long-term attention behaviors, and uses their relative attention working sets to estimate demand evolution. At each capacity boundary, GrowPage either compresses KV states within the current allocation or acquires an additional physical page when broader demand emerges. By integrating with PagedAttention's page-level memory abstraction, GrowPage preserves continuous batching and prefix caching. Experiments on reasoning benchmarks across multiple models show that GrowPage achieves a superior performance--throughput trade-off over existing approaches.

cs.AI↗

arXiv2026

SWE-TRACE: Optimizing Long-Horizon SWE Agents Through Rubric Process Reward Models and Heuristic Test-Time Scaling

Resolving real-world software engineering (SWE) issues with autonomous agents requires complex, long-horizon reasoning. Current pipelines are bottlenecked by unoptimized demonstration data, sparse execution rewards, and computationally prohibitive inference scaling, which collectively exacerbate token bloat, reward hacking, and policy degradation. We present SWE-TRACE (Trajectory Reduction and Agentic Criteria Evaluation), a unified framework optimizing the SWE agent lifecycle across data curation, reinforcement learning (RL), and test-time inference. First, we introduce an LLM multi-task cascading method, utilizing stepwise oracle verification to distill a 60K-instance Supervised Fine-Tuning (SFT) corpus strictly biased toward token-efficient, shortest-path trajectories. Second, to overcome the instability of sparse outcome rewards, we design a MemoryAugmented Agentic RL pipeline featuring a Rubric-Based Process Reward Model (PRM). An auxiliary Rubric-Agent provides dense, fine-grained heuristic feedback on intermediate steps, guiding the model through long-horizon tasks. Finally, we bridge training and inference by repurposing the PRM for heuristic-guided Test-Time Scaling (TTS). By dynamically evaluating and pruning action candidates at each step, SWE-TRACE achieves superior search efficiency without the latency overhead of standard parallel sampling. Extensive experiments on standard SWE benchmarks demonstrate that SWE-TRACE significantly advances the state-of-the-art, maximizing resolution rates while drastically reducing both token consumption and inference latency.

cs.SE↗

arXiv2026

Adapting AI to the Moment: Understanding the Dynamics of Parent-AI Collaboration Modes in Real-Time Conversations with Children

Parent-AI collaboration to support real-time conversations with children is challenging due to the sensitivity and open-ended nature of such interactions. Existing systems often simplify collaboration into static modes, providing limited support for adapting AI to continuously evolving conversational contexts. To address this gap, we systematically investigate the dynamics of parent-AI collaboration modes in real-time conversations with children. We conducted a co-design study with eight parents and developed COMPASS, a research probe that enables flexible combinations of parental support functions during conversations. Using COMPASS, we conducted a lab-based study with 21 parent-child pairs. We show that parent-AI collaboration unfolds through evolving modes that adapt systematically to contextual factors. We further identify three types of parental strategies--parent-oriented, child-oriented, and relationship-oriented--that shape how parents engage with AI. These findings advance the understanding of dynamic human-AI collaboration in relational, high-stakes settings and inform the design of flexible, context-adaptive parental support systems.

cs.HC↗

arXiv2026

PACEE: Parent-Centered AI Scaffolding for Emotion Education in Early Childhood Conversations

Emotion education is critical for children aged 3 to 6. However, existing technologies largely focus on children's direct interaction with AI, overlooking the central role of parents in guiding early emotional development at home. To address this gap, we conducted co-design sessions with five kindergarten teachers and five parents to identify key parental challenges and opportunities for AI support in family emotion education. Based on these insights, we developed PACEE, an LLM-based assistant designed to support parents in guiding children's emotional development through conversations, rather than directly interacting with children. PACEE provides parent-centered AI scaffolding that supports parents in real-time conversation through personalized guidance, post-hoc reflection through trackable feedback, and understanding children's emotional states through modeling. We evaluated PACEE with 16 families. Results show that PACEE enhances parent-child engagement, fosters deeper emotional communication, and improves parents' expertise and overall experience in guiding their children. Our findings extend emotion coaching practices to the context of generative AI and offer design insights for building AI systems that support parent-centered family education.

cs.HC↗

arXiv2026

ApET: Approximation-Error Guided Token Compression for Efficient VLMs

Recent Vision-Language Models (VLMs) have demonstrated remarkable multimodal understanding capabilities, yet the redundant visual tokens incur prohibitive computational overhead and degrade inference efficiency. Prior studies typically relies on [CLS] attention or text-vision cross-attention to identify and discard redundant visual tokens. Despite promising results, such solutions are prone to introduce positional bias and, more critically, are incompatible with efficient attention kernels such as FlashAttention, limiting their practical deployment for VLM acceleration. In this paper, we step away from attention dependencies and revisit visual token compression from an information-theoretic perspective, aiming to maximally preserve visual information without any attention involvement. We present ApET, an Approximation-Error guided Token compression framework. ApET first reconstructs the original visual tokens with a small set of basis tokens via linear approximation, then leverages the approximation error to identify and drop the least informative tokens. Extensive experiments across multiple VLMs and benchmarks demonstrate that ApET retains 95.2% of the original performance on image-understanding tasks and even attains 100.4% on video-understanding tasks, while compressing the token budgets by 88.9% and 87.5%, respectively. Thanks to its attention-free design, ApET seamlessly integrates with FlashAttention, enabling further inference acceleration and making VLM deployment more practical. Code is available at https://github.com/MaQianKun0/ApET.

cs.CV↗

arXiv2026

Multimodal Priors-Augmented Text-Driven 3D Human-Object Interaction Generation

We address the challenging task of text-driven 3D human-object interaction (HOI) motion generation. Existing methods primarily rely on a direct text-to-HOI mapping, which suffers from three key limitations due to the significant cross-modality gap: (Q1) sub-optimal human motion, (Q2) unnatural object motion, and (Q3) weak interaction between humans and objects. To address these challenges, we propose MP-HOI, a novel framework grounded in four core insights: (1) Multimodal Data Priors: We leverage multimodal data (text, image, pose/object) from large multimodal models as priors to guide HOI generation, which tackles Q1 and Q2 in data modeling. (2) Enhanced Object Representation: We improve existing object representations by incorporating geometric keypoints, contact features, and dynamic properties, enabling expressive object representations, which tackles Q2 in data representation. (3) Multimodal-Aware Mixture-of-Experts (MoE) Model: We propose a modality-aware MoE model for effective multimodal feature fusion paradigm, which tackles Q1 and Q2 in feature fusion. (4) Cascaded Diffusion with Interaction Supervision: We design a cascaded diffusion framework that progressively refines human-object interaction features under dedicated supervision, which tackles Q3 in interaction refinement. Comprehensive experiments demonstrate that MP-HOI outperforms existing approaches in generating high-fidelity and fine-grained HOI motions.

cs.CV↗

arXiv2026

RealSec-bench: A Benchmark for Evaluating Secure Code Generation in Real-World Repositories

Large Language Models (LLMs) have demonstrated remarkable capabilities in code generation, but their proficiency in producing secure code remains a critical, under-explored area. Existing benchmarks often fall short by relying on synthetic vulnerabilities or evaluating functional correctness in isolation, failing to capture the complex interplay between functionality and security found in real-world software. To address this gap, we introduce RealSec-bench, a new benchmark for secure code generation meticulously constructed from real-world, high-risk Java repositories. Our methodology employs a multi-stage pipeline that combines systematic SAST scanning with CodeQL, LLM-based false positive elimination, and rigorous human expert validation. The resulting benchmark contains 105 instances grounded in real-word repository contexts, spanning 19 Common Weakness Enumeration (CWE) types and exhibiting a wide diversity of data flow complexities, including vulnerabilities with up to 34-hop inter-procedural dependencies. Using RealSec-bench, we conduct an extensive empirical study on 5 popular LLMs. We introduce a novel composite metric, SecurePass@K, to assess both functional correctness and security simultaneously. We find that while Retrieval-Augmented Generation (RAG) techniques can improve functional correctness, they provide negligible benefits to security. Furthermore, explicitly prompting models with general security guidelines often leads to compilation failures, harming functional correctness without reliably preventing vulnerabilities. Our work highlights the gap between functional and secure code generation in current LLMs.

cs.CR↗

arXiv2026

Numerical simulations of simultaneous pair-drop impacts and their energetics

We present three-dimensional direct numerical simulations of the simultaneous impact of two identical drops on an hydrophobic substrate, varying the relative strength of capillary and viscous effects respectively through Weber and Reynolds numbers of impact. The interaction between the two drops is characterized by the appearance of a lamella arising from the collision of the two droplets' spreading rims. We examine the width, the height, and the general morphological evolution of the central sheet; the numerical data is validated against prior experiments and used to guide the development of an energetic model for the maximum elevation of the central sheet. In particular, the rise of the central sheet resembles the spreading behaviour single-drop impacts, especially at high Weber and Reynolds numbers. This fact can be used to estimate scalings in the capillary- and viscous-dominated regimes, which can be used to collapse the trajectories. These insights provide a route for a more complete understanding of the dynamics for the central rising sheet, and anticipate the detailed study of its fragmentation characteristics

physics.flu-dyn↗

arXiv2025

Unseen Speaker and Language Adaptation for Lightweight Text-To-Speech with Adapters

In this paper we investigate cross-lingual Text-To-Speech (TTS) synthesis through the lens of adapters, in the context of lightweight TTS systems. In particular, we compare the tasks of unseen speaker and language adaptation with the goal of synthesising a target voice in a target language, in which the target voice has no recordings therein. Results from objective evaluations demonstrate the effectiveness of adapters in learning language-specific and speaker-specific information, allowing pre-trained models to learn unseen speaker identities or languages, while avoiding catastrophic forgetting of the original model's speaker or language information. Additionally, to measure how native the generated voices are in terms of accent, we propose and validate an objective metric inspired by mispronunciation detection techniques in second-language (L2) learners. The paper also provides insights into the impact of adapter placement, configuration and the number of speakers used.

eess.AS↗

arXiv2025

Training-free Token Reduction for Vision Mamba

Vision Mamba has emerged as a strong competitor to Vision Transformers (ViTs) due to its ability to efficiently capture long-range dependencies with linear computational complexity. While token reduction, an effective compression technique in ViTs, has rarely been explored in Vision Mamba. Exploring Vision Mamba's efficiency is essential for enabling broader applications. However, we find that directly applying existing token reduction techniques for ViTs to Vision Mamba leads to significant performance degradation. This is primarily because Mamba is a sequence model without attention mechanisms, whereas most token reduction techniques for ViTs rely on attention mechanisms for importance measurement and overlook the order of compressed tokens. In this paper, we investigate a Mamba structure-aware importance score to evaluate token importance in a simple and effective manner. Building on this score, we further propose MTR, a training-free \textbf{M}amba \textbf{T}oken \textbf{R}eduction framework. Without the need for training or additional tuning parameters, our method can be seamlessly integrated as a plug-and-play component across various Mamba models. Extensive experiments demonstrate that our approach significantly reduces computational workload while minimizing performance impact across various tasks and multiple backbones. Notably, MTR reduces FLOPs by approximately 40\% on the Vim-B backbone, with only a 1.6\% drop in ImageNet performance without retraining.

cs.CV↗

arXiv2025

3D surface profiling via photonic integrated geometric sensor

Measurements of microscale surface patterns are essential for process and quality control in industries across semiconductors, micro-machining, and biomedicines. However, the development of miniaturized and intelligent profiling systems remains a longstanding challenge, primarily due to the complexity and bulkiness of existing benchtop systems required to scan large-area samples. A real-time, in-situ, and fast detection alternative is therefore highly desirable for predicting surface topography on the fly. In this paper, we present an ultracompact geometric profiler based on photonic integrated circuits, which directly encodes the optical reflectance of the sample and decodes it with a neural network. This platform is free of complex interferometric configurations and avoids time-consuming nonlinear fitting algorithms. We show that a silicon programmable circuit can generate pseudo-random kernels to project input data into higher dimensions, enabling efficient feature extraction via a lightweight one-dimensional convolutional neural network. Our device is capable of high-fidelity, fast-scanning-rate thickness identification for both smoothly varying samples and intricate 3D printed emblem structures, paving the way for a new class of compact geometric sensors.

physics.optics↗

arXiv2025

Tri-layer SiN-on-Si 8x8 Optical Switches with Thermo-optic and Electro-optic Actuators

We present two spatial-multiplexed switch-and-select (S&S) 8x8 optical switches incorporating a tri-layer SiN-on-Si platform, one equipped with thermo-optic (T-O) and the other electro-optic (E-O) switching elements. To the best of our knowledge, the electro-optic switch fabric is the first-of-its-kind device assembled in such a multi-layer platform. The shuffle between the multiplexer and demultiplexer array is established via a tri-layer Si-SiN-SiN structure, creating a three-dimensional crossing-free photonic shuffle network. At the same time, the implementation of the S&S topology can effectively suppress the first-order crosstalk. The measured on-chip losses for the T-O switch range from 2.1 to 11.5 dB, with a 5.2 dB average, while the E-O device exhibits losses between 8.7 to 19.6 dB, with a 15.1 dB average. Both switches demonstrate ultra-low crosstalk, with measured ranges of 38.9 to 50.8 dB and 42.8 to 51.9 dB, for the T-O and E-O devices respectively. The switching times are 17.6 us for the T-O switch and 5.9 ns with the E-O actuated one. These performance metrics highlight the potential of these switches for next-generation data center applications.

physics.optics↗

arXiv2025

Fg-T2M++: LLMs-Augmented Fine-Grained Text Driven Human Motion Generation

We address the challenging problem of fine-grained text-driven human motion generation. Existing works generate imprecise motions that fail to accurately capture relationships specified in text due to: (1) lack of effective text parsing for detailed semantic cues regarding body parts, (2) not fully modeling linguistic structures between words to comprehend text comprehensively. To tackle these limitations, we propose a novel fine-grained framework Fg-T2M++ that consists of: (1) an LLMs semantic parsing module to extract body part descriptions and semantics from text, (2) a hyperbolic text representation module to encode relational information between text units by embedding the syntactic dependency graph into hyperbolic space, and (3) a multi-modal fusion module to hierarchically fuse text and motion features. Extensive experiments on HumanML3D and KIT-ML datasets demonstrate that Fg-T2M++ outperforms SOTA methods, validating its ability to accurately generate motions adhering to comprehensive text semantics.

cs.CV↗

arXiv2025

LLM Hallucinations in Practical Code Generation: Phenomena, Mechanism, and Mitigation

Code generation aims to automatically generate code from input requirements, significantly enhancing development efficiency. Recent large language models (LLMs) based approaches have shown promising results and revolutionized code generation task. Despite the promising performance, LLMs often generate contents with hallucinations, especially for the code generation scenario requiring the handling of complex contextual dependencies in practical development process. Although previous study has analyzed hallucinations in LLM-powered code generation, the study is limited to standalone function generation. In this paper, we conduct an empirical study to study the phenomena, mechanism, and mitigation of LLM hallucinations within more practical and complex development contexts in repository-level generation scenario. First, we manually examine the code generation results from six mainstream LLMs to establish a hallucination taxonomy of LLM-generated code. Next, we elaborate on the phenomenon of hallucinations, analyze their distribution across different models. We then analyze causes of hallucinations and identify four potential factors contributing to hallucinations. Finally, we propose an RAG-based mitigation method, which demonstrates consistent effectiveness in all studied LLMs. The replication package including code, data, and experimental results is available at https://github.com/DeepSoftwareAnalytics/LLMCodingHallucination

cs.SE↗

arXiv2025

Dilated space-and-wavelength selective crosspoint optical switch

Photonic integrated switches that are both space and wavelength selective are a highly promising technology for data-intensive applications as they benefit from multi-dimensional manipulation of optical signals. However, scaling these switches normally poses stringent challenges such as increased fabrication complexity and control difficulties, due to the growing number of switching elements. In this work, we propose a new type of dilated crosspoint topology, which efficiently handles both space and wavelength selective switching, while reducing the required switching element count by an order of magnitude compared to reported designs. To the best of our knowledge, our design requires the fewest switching elements for an equivalent routing paths number and it fully cancels the first-order in-band crosstalk. We demonstrate such an ultra-compact space-and-wavelength-selective switch (SWSS) at a scale of $4\times 4\times 4λ$ on the silicon-on-insulator (SOI) platform. Experimental results reveal that the switch achieves an insertion loss ranging from 2.3 dB to 8.6 dB and crosstalk levels in between -35.3 dB and -59.7 dB. The add-drop microring-resonators (MRRs) are equipped with micro-heaters, exhibiting a rise and fall time of 46 $μ$s and 0.33 $μ$s, respectively. These performance characteristics highlight the switch's ultra-low element count and crosstalk with low insertion loss, making it a promising candidate for advanced data center applications.

physics.optics↗

arXiv2024

Mindalogue: LLM-Powered Nonlinear Interaction for Effective Learning and Task Exploration

Current generative AI models like ChatGPT, Claude, and Gemini are widely used for knowledge dissemination, task decomposition, and creative thinking. However, their linear interaction methods often force users to repeatedly compare and copy contextual information when handling complex tasks, increasing cognitive load and operational costs. Moreover, the ambiguity in model responses requires users to refine and simplify the information further. To address these issues, we developed "Mindalogue", a system using a non-linear interaction model based on "nodes + canvas" to enhance user efficiency and freedom while generating structured responses. A formative study with 11 users informed the design of Mindalogue, which was then evaluated through a study with 16 participants. The results showed that Mindalogue significantly reduced task steps and improved users' comprehension of complex information. This study highlights the potential of non-linear interaction in improving AI tool efficiency and user experience in the HCI field.

cs.HC↗