arXiv subjects

Wenchao Yang

Publications and source records attributed to Wenchao Yang.

9 recordsLinked to original sources

arXiv2026

KAST-BAR: Knowledge-Anchored Semantically-Dynamic Topology Brain Autoregressive Modeling for Universal Neural Interpretation

While EEG foundation models have shown significant potential in universal neural decoding across tasks, their advancement remains constrained by the inadequacy modeling of complex spatiotemporal topology, as well as the inherent modality gap between low-level physiological signals and high-level textual semantics. To address these challenges, we propose a Knowledge-Anchored Semantically-Dynamic Topology Brain Autoregressive Model (KAST-BAR), which dynamically aligns physiological representations derived from multi-level brain topology with an expert-level semantic space. Specifically, we design a Dual-Stream Hierarchical Attention (DSHA) encoder that accurately captures the brain's intrinsic non-Euclidean topology by modeling local temporal dynamics with global spatial contexts. On this basis, a Knowledge-Anchored Semantic Profiler (KASP) is proposed to synthesize physically-grounded and instance-level textual profiles, which subsequently drive a Semantic Text-Aware Refiner (STAR) to dynamically reconstruct EEG representations using Latent Expert Queries. By conducting large-scale pre-training on 21 diverse datasets to build a foundation model, KAST-BAR effectively integrates expert-level medical knowledge into EEG signal representations, consistently achieving superior performance across six downstream tasks. Our code is available at https://github.com/KAST-BAR/KAST-BAR

cs.LG↗

arXiv2025

THD-BAR: Topology Hierarchical Derived Brain Autoregressive Modeling for EEG Generic Representations

Large-scale pre-trained models hold significant potential for learning universal EEG representations. However, most existing methods, particularly autoregressive (AR) frameworks, primarily rely on straightforward temporal sequencing of multi-channel EEG data, which fails to capture the rich physiological characteristics inherent to EEG signals. Moreover, their time-centered modeling approach also limits the effective representation of the dynamic spatial topology of brain activity. To address these challenges and fully exploit the potential of large-scale EEG models, we propose a novel Topology Hierarchical Derived Brain Autoregressive Modeling (THD-BAR) for EEG generic representations. The core innovation of THD-BAR lies in the introduction of the Brain Topology Hierarchy (BTH), which establishes a multi-scale spatial order for EEG channels. This hierarchical structure enables a redefinition of autoregressive learning as a "next-scale-time prediction" problem, effectively capturing both spatial and temporal dynamics. Based on BTH, we design a Topology-Hierarchical Vector Quantized-Variational Autoencoder (THVQ-VAE) for multi-scale tokenization and develop an enhanced Brain Autoregressive (BAR) module with specialized masking strategies for prediction. Through extensive large-scale pre-training on 17 datasets, followed by rigorous validation on 10 downstream datasets spanning 5 distinct tasks, THD-BAR consistently outperforms existing methods. These results highlight the superior generalization and modeling capabilities of our proposed approach.

eess.SP↗

arXiv2025

Oyster-I: Beyond Refusal -- Constructive Safety Alignment for Responsible Language Models

Large language models (LLMs) typically deploy safety mechanisms to prevent harmful content generation. Most current approaches focus narrowly on risks posed by malicious actors, often framing risks as adversarial events and relying on defensive refusals. However, in real-world settings, risks also come from non-malicious users seeking help while under psychological distress (e.g., self-harm intentions). In such cases, the model's response can strongly influence the user's next actions. Simple refusals may lead them to repeat, escalate, or move to unsafe platforms, creating worse outcomes. We introduce Constructive Safety Alignment (CSA), a human-centric paradigm that protects against malicious misuse while actively guiding vulnerable users toward safe and helpful results. Implemented in Oyster-I (Oy1), CSA combines game-theoretic anticipation of user reactions, fine-grained risk boundary discovery, and interpretable reasoning control, turning safety into a trust-building process. Oy1 achieves state-of-the-art safety among open models while retaining high general capabilities. On our Constructive Benchmark, it shows strong constructive engagement, close to GPT-5, and unmatched robustness on the Strata-Sword jailbreak dataset, nearing GPT-o1 levels. By shifting from refusal-first to guidance-first safety, CSA redefines the model-user relationship, aiming for systems that are not just safe, but meaningfully helpful. We release Oy1, code, and the benchmark to support responsible, user-centered AI.

cs.AI↗

arXiv2024

AVM-SLAM: Semantic Visual SLAM with Multi-Sensor Fusion in a Bird's Eye View for Automated Valet Parking

Accurate localization in challenging garage environments -- marked by poor lighting, sparse textures, repetitive structures, dynamic scenes, and the absence of GPS -- is crucial for automated valet parking (AVP) tasks. Addressing these challenges, our research introduces AVM-SLAM, a cutting-edge semantic visual SLAM architecture with multi-sensor fusion in a bird's eye view (BEV). This novel framework synergizes the capabilities of four fisheye cameras, wheel encoders, and an inertial measurement unit (IMU) to construct a robust SLAM system. Unique to our approach is the implementation of a flare removal technique within the BEV imagery, significantly enhancing road marking detection and semantic feature extraction by convolutional neural networks for superior mapping and localization. Our work also pioneers a semantic pre-qualification (SPQ) module, designed to adeptly handle the challenges posed by environments with repetitive textures, thereby enhancing loop detection and system robustness. To demonstrate the effectiveness and resilience of AVM-SLAM, we have released a specialized multi-sensor and high-resolution dataset of an underground garage, accessible at https://yale-cv.github.io/avm-slam_dataset, encouraging further exploration and validation of our approach within similar settings.

cs.RO↗

arXiv2021

The wake of a transversely oscillating circular cylinder in a flowing soap film at low Reynolds number

An inclined gravity-driven soap film channel was used to study the wake patterns formed behind a transversely oscillating cylinder at $Re =235 \pm 14$. The natural frequency of vortex shedding from a stationary cylinder, $f_{\text{St}}$, was used to identify the oscillation frequencies of interest. The (dimensionless) frequency, $f^*=f/{f_{\text{St}}}$, and amplitude, $A^*=A/D$, of the cylinder's motion was varied over a large portion of the fundamental synchronization region (i.e., for $f^* \approx 1$), and a `map' of wake patterns was constructed in ${(f^*, A^*)}$ space. Lock-on between the frequency of the cylinder's motion and the dominant frequency of the resulting vortex wake was observed for a large range of this parameter space, predominantly manifested as synchronized `2S' and `2P' wake modes. Synchronized `P+S', `2T', and `transitional' wakes were also found in smaller regions of parameter space. Unsynchronized `coalescing' and `perturbed von~\karman' wakes were observed as the oscillation frequency became sufficiently different from $f_{\text_{St}}$. The wake patterns and vortex formation processes found in this study, particularly for the 2P mode wakes, are more similar to those observed by \cite{Williamson1988} in three-dimensional experiments than those found by \cite{Leontini2006} in two-dimensional simulations, despite the physical constraint from the soap film that prevents three-dimensional effects in the wake.

physics.flu-dyn↗

arXiv2018

Achieving balanced open circuit voltage and short circuit current by tuning the interfacial energetics in organic bulk heterojunction solar cells: A drift-diffusion simulation

In organic bulk heterojunction solar cells, the donor/acceptor interfacial energy offset ($ΔE$) is found to provide the driving force for efficient charge separation which gives rise to high short circuit current density ($J_\mathrm{sc}$), but a high $ΔE$ inevitably undermines the open circuit voltage ($V_\mathrm{oc}$). In this paper, employing the device model method we calculated the steady state current density-voltage ($J-V$) and the $J_\mathrm{sc}-ΔE$ curves under two different charge separation mechanisms to investigate the optimum driving force required for achieving sizable $V_\mathrm{oc}$ and $J_\mathrm{sc}$ simultaneously. Under the Marcus charge transfer mechanism, with the increased $ΔE$ the Jsc increases rapidly for $ΔE\leq 0.2$ eV, and then maintains a nearly constant value before decreasing at the Marcus inverted region, which is due to the accumulation of undissociated excitons within their lifetime and is beneficial for obtaining a sizable $J_\mathrm{sc}$ under a $ΔE$ much smaller than the reorganization energy $λ$. For the coherent charge transfer mechanism in which the driving force act as the energy window of accessible charge separated states, with two typical types of density of states for the charge transfer excitons, it is shown that the highest $J_\mathrm{sc}$ can also be achieved under a small $ΔE$ of 0.2\,eV if the high-lying delocalized states are harvested in high proportion. This work demonstrates the existence of the optimum driving force of 0.2\,eV and provides some guidelines for engineering the interfacial energetics to achieve the high balanced $J_\mathrm{sc}$ and $V_\mathrm{oc}$.

cond-mat.mtrl-sci↗

arXiv2016

Energetic disorder induced leakage current in organic bulk heterojunction solar cells: comprehending the ultra-high open circuit voltage loss at low temperatures

In organic bulk heterojunction solar cells, the open circuit voltage ($V_\mathrm{oc}$) suffers from an ultra-high loss at low temperatures. In this work we investigate the origin of the loss through calculating the $V_\mathrm{oc}-T$ plots with the device model method systematically and comparing it with experimentally observed ones. When the energetic disorder is incorporated into the model by considering the disorder-suppressed and temperature-dependent charge carrier mobilities, it is found that for nonselective contacts the $V_\mathrm{oc}$ reduces drastically under the low temperature regime, while for selective contacts the $V_\mathrm{oc}$ keeps increasing with the decreasing temperature. The main reason is revealed that as the temperature decreases, the reduced mobilities give rise to low charge extraction efficiency and small bimolecular recombination rate for the photogenerated charge carriers, so that in the former case they can be extracted from the wrong electrode to form a leakage current which counteracts the photocurrent and increases quickly with voltage, leading to the anomalous reduction of $V_\mathrm{oc}$. In addition, it is revealed that the charge generation rate is slow-varying with temperature and does not induce significant $V_\mathrm{oc}$ loss. This work also provides a comprehensive picture for the $V_\mathrm{oc}$ behavior under varying device working conditions.

cond-mat.mtrl-sci↗

arXiv2016

A Computationally Efficient Framework for Automatic Inertial Sensor Calibration

The calibration of (low-cost) inertial sensors has become increasingly important over the past years since their use has grown exponentially in many applications going from unmanned aerial vehicle navigation to 3D-animation. However, this calibration procedure is often quite problematic since the signals issued from these sensors have a complex spectral structure and the methods available to estimate the parameters of these models are either unstable, computationally intensive and/or statistically inconsistent. This paper presents a new software platform for inertial sensor calibration based on the Generalized Method of Wavelet Moments which provides a computationally efficient, flexible, user-friendly and statistically sound tool to estimate and select from a wide range of complex models. The software is developed within the open-source statistical software R and is based on C++ language allowing it to achieve high computational performance.

stat.AP↗

arXiv2016

The gmwm R package: a comprehensive tool for time series analysis from state-space models to robustness

The gmwm R package for inference on time series models is mainly based on the quantity called wavelet variance which is derived from a wavelet decomposition of a time series. This quantity provides a means to summarize and graphically represent the features of time series in order to identify possible models. Moreover, it is used as a moment condition for model estimation through the generalized method of wavelet moments. Based on the latter method, this package not only provides an alternative method to estimate classical ARMA models but also delivers a general framework for the robust estimation of many time series models as well as a quick and efficient estimation of many linear state-space models.

stat.CO↗