arXiv subjects

Dongqi Zheng

Publications and source records attributed to Dongqi Zheng.

6 recordsLinked to original sources

arXiv2026

When Few Steps Are Enough: Training-Free Acceleration of Identity-Preserved Generation

Identity-preserved image generation is typically built on many-step diffusion backbones, making personalized generation expensive at deployment time. We show that this cost is often unnecessary for identity-conditioned FLUX generation. A frozen InfuseNet identity adapter trained with dev transfers directly to the distilled schnell backbone without retraining. This two-line replacement -- changing the backbone path and disabling classifier-free guidance -- reduces latency by 5.9x while improving ArcFace identity similarity by +0.028 and lpips by -0.016 over the standard 28-step dev baseline. To explain why this works, we analyze the denoising trajectory and find that identity fidelity enters an early effective regime, often within 4-8 steps, while later steps primarily refine visual detail, sharpness, and contrast. Adapter ablations confirm that identity formation depends on the identity adapter, while attention-stream norm probes suggest that the relative conditioning contribution decreases as sampling proceeds. Preliminary style-adapter and object-adapter sweeps on SDXL and SD1.5 show similar diminishing returns after intermediate steps. These results position distilled backbone replacement as a simple, training-free strategy for improving the efficiency-fidelity tradeoff of identity-preserved generation.

cs.CV↗

arXiv2025

ARS: Adaptive Reasoning Suppression for Efficient Large Reasoning Language Models

Large Reasoning Language Models (LRLMs or LRMs) demonstrate remarkable capabilities in complex reasoning tasks, but suffer from significant computational inefficiencies due to overthinking phenomena. Existing efficient reasoning methods face the challenge of balancing reasoning quality with inference cost reduction. We propose \textbf{Adaptive Reasoning Suppression (ARS)}, a novel training-free approach that dynamically suppresses redundant reasoning steps while preserving accuracy through adaptive certainty monitoring. ARS introduces a multi-checkpoint certainty estimation mechanism with progressive suppression thresholds, achieving superior efficiency compared to static suppression methods. Our extensive evaluation across mathematical reasoning benchmarks using multiple model architectures demonstrates that ARS achieves up to 53%, 46.1%, and 57.9% in token, latency and energy reduction, while maintaining or improving accuracy.

cs.AI↗

arXiv2025

CAFL-L: Constraint-Aware Federated Learning with Lagrangian Dual Optimization for On-Device Language Models

We introduce Constraint-Aware Federated Learning with Lagrangian Dual Optimization (CAFL-L), a principled extension of FedAvg that explicitly incorporates device-level resource constraints including energy, communication, memory, and thermal budgets. CAFL-L employs Lagrangian dual optimization to dynamically adapt training hyperparameters -- freezing depth, local steps, batch size, and communication compression -- while preserving training stability through token-budget preservation via gradient accumulation. Experiments on a character-level language model demonstrate that CAFL-L achieves superior constraint satisfaction compared to standard FedAvg (reducing memory usage by 20% and communication by 95%) while maintaining competitive validation performance, making it practical for deployment on resource-constrained edge devices.

cs.LG↗

arXiv2025

A Real-Time On-Device Defect Detection Framework for Laser Power-Meter Sensors via Unsupervised Learning

We present an automated vision-based system for defect detection and classification of laser power meter sensor coatings. Our approach addresses the critical challenge of identifying coating defects such as thermal damage and scratches that can compromise laser energy measurement accuracy in medical and industrial applications. The system employs an unsupervised anomaly detection framework that trains exclusively on ``good'' sensor images to learn normal coating distribution patterns, enabling detection of both known and novel defect types without requiring extensive labeled defect datasets. Our methodology consists of three key components: (1) a robust preprocessing pipeline using Laplacian edge detection and K-means clustering to segment the area of interest, (2) synthetic data augmentation via StyleGAN2, and (3) a UFlow-based neural network architecture for multi-scale feature extraction and anomaly map generation. Experimental evaluation on 366 real sensor images demonstrates $93.8\%$ accuracy on defective samples and $89.3\%$ accuracy on good samples, with image-level AUROC of 0.957 and pixel-level AUROC of 0.961. The system provides potential annual cost savings through automated quality control and processing times of 0.5 seconds per image in on-device implementation.

cs.CV↗

arXiv2025

Diffusion Models on the Edge: Challenges, Optimizations, and Applications

Diffusion models have shown remarkable capabilities in generating high-fidelity data across modalities such as images, audio, and video. However, their computational intensity makes deployment on edge devices a significant challenge. This survey explores the foundational concepts of diffusion models, identifies key constraints of edge platforms, and synthesizes recent advancements in model compression, sampling efficiency, and hardware-software co-design to make diffusion models viable on edge devices. We also review promising applications and suggest future research directions.

cs.DC↗

arXiv2020

Why In2O3 Can Make 0.7 nm Atomic Layer Thin Transistors?

In this work, we demonstrate enhancement-mode field-effect transistors by atomic-layer-deposited (ALD) amorphous In2O3 channel with thickness down to 0.7 nm. Thickness is found to be critical on the materials and electron transport of In2O3. Controllable thickness of In2O3 at atomic scale enables the design of sufficient 2D carrier density in the In2O3 channel integrated with the conventional dielectric. The threshold voltage and channel carrier density are found to be considerably tuned by channel thickness. Such phenomenon is understood by the trap neutral level (TNL) model where the Fermi-level tends to align deeply inside the conduction band of In2O3 and can be modulated to the bandgap in atomic layer thin In2O3 due to quantum confinement effect, which is confirmed by density function theory (DFT) calculation. The demonstration of enhancement-mode amorphous In2O3 transistors suggests In2O3 is a competitive channel material for back-end-of-line (BEOL) compatible transistors and monolithic 3D integration applications.

cond-mat.mtrl-sci↗