arXiv subjects

Weiyu Sun

Publications and source records attributed to Weiyu Sun.

11 recordsLinked to original sources

arXiv2026

Hardware Acceleration of Block-Diffusion LLM for Edge Devices

Single-stream (batch-one) edge inference cannot amortize weight traffic across requests. Full-attention diffusion LLMs recompute the entire sequence at every step; native block diffusion makes completed blocks immutable and exactly cacheable, yet refinement still streams prefix KV and FFN weights. We co-design WIFiV-LPDDR, a wide-I/O LPDDR system for precision-tagged reads, BRQ-KV for a canonical low-rank-plus-INT8-residual prefix with query-dependent per-entry precision, and DAT-FFN for drift-mapped canonical replacement, adjacent-stage-corrected low-bit delta, or cached-state carry while keeping live activations unquantized. Both map to an input-stationary mixed-precision systolic array. For the evaluated 1.5B/7B models on modeled Jetson-class platforms, the full stack provides arithmetic-mean energy-reduction factors of 3.79x/3.96x and arithmetic-mean latency speedups of 2.88x/4.44x at the reported DAT-FFN settings; every corresponding compressed model-benchmark score drops by less than one absolute percentage point from its baseline.

cs.AR↗

arXiv2026

Direct Measurement of Exciton Dispersion in the Long-Wavelength Limit

Exciton dispersion, which governs the propagation, scattering and radiative decay of electron-hole pairs, is essential to optoelectronics and quantum materials. In two-dimensional systems, weakened dielectric screening and long-range electron-hole exchange are predicted to induce nonanalytic exciton dispersion in the long-wavelength limit. However, direct quantitative characterization of its dimensional evolution remains lacking, especially in the ultralow-q regime (q < 0.02 $\r{A}^{-1}$). Here we employ defocus-engineered momentum-resolved electron energy-loss spectroscopy in scanning transmission electron microscopy, achieving an ultrahigh momentum resolution of 0.0002 $\r{A}^{-1}$. Using freestanding hBN as a prototypical platform, we resolve layer-dependent exciton dispersion and quantify its characteristic crossover momentum and group velocity in the long-wavelength limit. With increasing thickness, the nonanalytic linear-dispersion regime is progressively compressed, manifested by a reduction in characteristic crossover momentum q_c from $1.82 \times 10^{-1} \r{A}^{-1}$ in the monolayer to $3.0 \times 10^{-1} \r{A}^{-1}$ in 25 layers. Meanwhile, the low-q group velocity increases from $2.0 \times 10^{-3} c$ to $2.9 \times 10^{-2} c$, before the dispersion ultimately approaches the bulk-like parabolic limit. We further examine how the exciton band structure of monolayer hBN responds to its surrounding environment, including temperature, adjacent graphene layers, and interlayer twist in BN/graphene heterostructures. These findings uncover the fundamental physics of low-dimensional excitons, deliver valuable guidance for modulating exciton transport, diffusion and quasiparticle coupling in layered quantum materials, and establish a powerful experimental route to explore low-dimensional exciton physics.

cond-mat.mes-hall↗

arXiv2026

EDU-CIRCUIT-HW: Evaluating Multimodal Large Language Models on Real-World University-Level STEM Student Handwritten Solutions

Multimodal Large Language Models (MLLMs) hold significant promise for revolutionizing traditional education and reducing teachers' workload. However, accurately interpreting unconstrained STEM student handwritten solutions with intertwined mathematical formulas, diagrams, and textual reasoning poses a significant challenge due to the lack of authentic and domain-specific benchmarks. Additionally, current evaluation paradigms predominantly rely on the outcomes of downstream tasks (e.g., auto-grading), which often probe only a subset of the recognized content, thereby failing to capture the MLLMs' understanding of complex handwritten logic as a whole. To bridge this gap, we release EDU-CIRCUIT-HW, a dataset consisting of 1,300+ authentic student handwritten solutions from a university-level STEM course. Utilizing the expert-verified verbatim transcriptions and grading reports of student solutions, we simultaneously evaluate various MLLMs' upstream recognition fidelity and downstream auto-grading performance. Our evaluation uncovers an astonishing scale of latent failures within MLLM-recognized student handwritten content, highlighting the models' insufficient reliability for auto-grading and other understanding-oriented applications in high-stakes educational settings. As a potential solution, we present a case study demonstrating that leveraging identified error patterns to preemptively detect and correct recognition errors, while requiring only minimal human intervention (e.g., routing 3.3% of assignments to human graders and the remainder to the GPT-5.1 grader), can effectively enhance the robustness of the deployed AI-enabled grading system. Code and dataset are available in this GitHub repo: https://gt-learning-innovation.github.io/CIRCUIT_EDU_HW_ACL.

cs.CV↗

arXiv2025

Enhancing Large Language Model-Based Systems for End-to-End Circuit Analysis Problem Solving

LLMs have demonstrated strong performance in data-rich domains such as programming, yet their reliability in engineering tasks remains limited. Circuit analysis--requiring multimodal understanding and precise mathematical reasoning--highlights these challenges. Although Gemini 2.5 Pro shows improved capabilities in diagram interpretation and analog-circuit reasoning, it still struggles to consistently produce correct solutions when given both textual problem descriptions and circuit diagrams. Meanwhile, engineering education demands scalable AI tools capable of generating accurate solutions for applications such as automated homework feedback. This paper presents an enhanced end-to-end circuit problem-solving framework built upon Gemini. We first conduct a systematic benchmark on undergraduate circuit problems and identify two key failure modes: 1) circuit-recognition hallucinations, particularly incorrect source polarity detection, and 2) reasoning-process hallucinations, such as incorrect current direction assumptions. To address recognition errors, we integrate a fine-tuned YOLO detector and OpenCV-based processing to isolate voltage and current sources, enabling Gemini to accurately re-identify source polarities from cropped images. To mitigate reasoning errors, we introduce an ngspice-driven verification loop, in which simulation discrepancies trigger iterative solution refinement with optional HITL feedback. Experimental results demonstrate that the proposed pipeline achieves 97.59% accuracy, substantially outperforming Gemini's baseline of 79.52%. Furthermore, on four variations of hand-drawn circuit diagrams, accuracy improves from 56.06%--71.21% to 93.94%--95.45% with statistically significant gains. These results highlight the robustness, scalability, and practical applicability of the proposed framework for engineering education and real-world circuit analysis tasks.

cs.CY↗

arXiv2024

BEEAR: Embedding-based Adversarial Removal of Safety Backdoors in Instruction-tuned Language Models

Safety backdoor attacks in large language models (LLMs) enable the stealthy triggering of unsafe behaviors while evading detection during normal interactions. The high dimensionality of potential triggers in the token space and the diverse range of malicious behaviors make this a critical challenge. We present BEEAR, a mitigation approach leveraging the insight that backdoor triggers induce relatively uniform drifts in the model's embedding space. Our bi-level optimization method identifies universal embedding perturbations that elicit unwanted behaviors and adjusts the model parameters to reinforce safe behaviors against these perturbations. Experiments show BEEAR reduces the success rate of RLHF time backdoor attacks from >95% to <1% and from 47% to 0% for instruction-tuning time backdoors targeting malicious code generation, without compromising model utility. Requiring only defender-defined safe and unwanted behaviors, BEEAR represents a step towards practical defenses against safety backdoors in LLMs, providing a foundation for further advancements in AI safety and security.

cs.CR↗

arXiv2024

Boosting Single Positive Multi-label Classification with Generalized Robust Loss

Multi-label learning (MLL) requires comprehensive multi-semantic annotations that is hard to fully obtain, thus often resulting in missing labels scenarios. In this paper, we investigate Single Positive Multi-label Learning (SPML), where each image is associated with merely one positive label. Existing SPML methods only focus on designing losses using mechanisms such as hard pseudo-labeling and robust losses, mostly leading to unacceptable false negatives. To address this issue, we first propose a generalized loss framework based on expected risk minimization to provide soft pseudo labels, and point out that the former losses can be seamlessly converted into our framework. In particular, we design a novel robust loss based on our framework, which enjoys flexible coordination between false positives and false negatives, and can additionally deal with the imbalance between positive and negative samples. Extensive experiments show that our approach can significantly improve SPML performance and outperform the vast majority of state-of-the-art methods on all the four benchmarks.

cs.LG↗

arXiv2024

Resolve Domain Conflicts for Generalizable Remote Physiological Measurement

Remote photoplethysmography (rPPG) technology has become increasingly popular due to its non-invasive monitoring of various physiological indicators, making it widely applicable in multimedia interaction, healthcare, and emotion analysis. Existing rPPG methods utilize multiple datasets for training to enhance the generalizability of models. However, they often overlook the underlying conflict issues across different datasets, such as (1) label conflict resulting from different phase delays between physiological signal labels and face videos at the instance level, and (2) attribute conflict stemming from distribution shifts caused by head movements, illumination changes, skin types, etc. To address this, we introduce the DOmain-HArmonious framework (DOHA). Specifically, we first propose a harmonious phase strategy to eliminate uncertain phase delays and preserve the temporal variation of physiological signals. Next, we design a harmonious hyperplane optimization that reduces irrelevant attribute shifts and encourages the model's optimization towards a global solution that fits more valid scenarios. Our experiments demonstrate that DOHA significantly improves the performance of existing methods under multiple protocols. Our code is available at https://github.com/SWY666/rPPG-DOHA.

cs.CV↗

arXiv2024

Backdoor Contrastive Learning via Bi-level Trigger Optimization

Contrastive Learning (CL) has attracted enormous attention due to its remarkable capability in unsupervised representation learning. However, recent works have revealed the vulnerability of CL to backdoor attacks: the feature extractor could be misled to embed backdoored data close to an attack target class, thus fooling the downstream predictor to misclassify it as the target. Existing attacks usually adopt a fixed trigger pattern and poison the training set with trigger-injected data, hoping for the feature extractor to learn the association between trigger and target class. However, we find that such fixed trigger design fails to effectively associate trigger-injected data with target class in the embedding space due to special CL mechanisms, leading to a limited attack success rate (ASR). This phenomenon motivates us to find a better backdoor trigger design tailored for CL framework. In this paper, we propose a bi-level optimization approach to achieve this goal, where the inner optimization simulates the CL dynamics of a surrogate victim, and the outer optimization enforces the backdoor trigger to stay close to the target throughout the surrogate CL procedure. Extensive experiments show that our attack can achieve a higher attack success rate (e.g., $99\%$ ASR on ImageNet-100) with a very low poisoning rate ($1\%$). Besides, our attack can effectively evade existing state-of-the-art defenses. Code is available at: https://github.com/SWY666/SSL-backdoor-BLTO.

cs.CR↗

arXiv2023

Self-similarity Prior Distillation for Unsupervised Remote Physiological Measurement

Remote photoplethysmography (rPPG) is a noninvasive technique that aims to capture subtle variations in facial pixels caused by changes in blood volume resulting from cardiac activities. Most existing unsupervised methods for rPPG tasks focus on the contrastive learning between samples while neglecting the inherent self-similar prior in physiological signals. In this paper, we propose a Self-Similarity Prior Distillation (SSPD) framework for unsupervised rPPG estimation, which capitalizes on the intrinsic self-similarity of cardiac activities. Specifically, we first introduce a physical-prior embedded augmentation technique to mitigate the effect of various types of noise. Then, we tailor a self-similarity-aware network to extract more reliable self-similar physiological features. Finally, we develop a hierarchical self-distillation paradigm to assist the network in disentangling self-similar physiological patterns from facial videos. Comprehensive experiments demonstrate that the unsupervised SSPD framework achieves comparable or even superior performance compared to the state-of-the-art supervised methods. Meanwhile, SSPD maintains the lowest inference time and computation cost among end-to-end models.

cs.CV↗

arXiv2023

Tackling the Non-IID Issue in Heterogeneous Federated Learning by Gradient Harmonization

Federated learning (FL) is a privacy-preserving paradigm for collaboratively training a global model from decentralized clients. However, the performance of FL is hindered by non-independent and identically distributed (non-IID) data and device heterogeneity. In this work, we revisit this key challenge through the lens of gradient conflicts on the server side. Specifically, we first investigate the gradient conflict phenomenon among multiple clients and reveal that stronger heterogeneity leads to more severe gradient conflicts. To tackle this issue, we propose FedGH, a simple yet effective method that mitigates local drifts through Gradient Harmonization. This technique projects one gradient vector onto the orthogonal plane of the other within conflicting client pairs. Extensive experiments demonstrate that FedGH consistently enhances multiple state-of-the-art FL baselines across diverse benchmarks and non-IID scenarios. Notably, FedGH yields more significant improvements in scenarios with stronger heterogeneity. As a plug-and-play module, FedGH can be seamlessly integrated into any FL framework without requiring hyperparameter tuning.

cs.LG↗

arXiv2022

BYHE: A Simple Framework for Boosting End-to-end Video-based Heart Rate Measurement Network

Heart rate measuring based on remote photoplethysmography (rPPG) plays an important role in health caring, which estimates heart rate from facial video in a non-contact, less-constrained way. End-to-end neural network is a main branch of rPPG-based heart rate estimation methods, whose trait is recovering rPPG signal containing sufficient heart rate message from original facial video directly. However, there exists some easily neglected problems on relevant datasets which thwarting the efficient training of end-to-end methods, such as uncertain temporal delay and indefinite envelope shape of label waves. Although many novel and powerful networks are proposed, hitherto there are no systematic research digging into these problems. In this paper, from perspective of common intrinsic rhythm periodical self-similarity results from cardiac activities, we propose a comprehensive methodology, Boost Your Heartbeat Estimation (BYHE), including new label representations, corresponding network adjustments and loss functions. BYHE can be easily grafted on current end-to-end network and boost its training efficiency. By applying our methodology, we can save tremendous time without conducting laborious handworks, such as label wave alignment which is necessary for previous end-to-end methods, and meanwhile enhance the utilization on datasets. According to our experiments, BYHE can leverage classical end-to-end network to reach competitive performance against those state-of-the-art methods on mostly used datasets. Such improvement indicates selecting perspicuous and efficient label representation is also a promising direction towards better remote physiological signal measurement.

cs.CV↗