arXiv subjects

Haoyang Zhao

Publications and source records attributed to Haoyang Zhao.

3 recordsLinked to original sources

arXiv2026

PhysMind: From Video to Executable Worlds for Training-Free Physical Reasoning

Reliable physical reasoning from video requires understanding how objects move, interact, and respond to interventions. Existing vision-language models (VLMs) often struggle to interpret these dynamics and reason reliably about future and counterfactual outcomes. We introduce PhysMind, a training-free agentic framework that constructs one reusable, question-agnostic executable world per video. PhysMind recovers a temporally consistent dynamic scene through object segmentation, mesh reconstruction, and 6D pose tracking, then fits analytic continuous-time dynamics and latent physical parameters without unrolling a time-stepped simulator. Given a question, it inspects, continues, or edits the world and answers from the resulting trajectories and interactions. Relative to direct chain-of-thought (CoT) reasoning with the same VLM, PhysMind improves accuracy by 38.23 points on CLEVRER and 8.08 points on Physion++. On counterfactual questions, it exceeds the strongest evaluated VLM baseline, GPT-5.5, by 19.25 points.

cs.CV↗

arXiv2025

Leakage-Free Evaluation and Distribution-Robust Spatio-Temporal Graph Learning for Inductive Kriging

Inductive kriging estimates values at unobserved locations from sparse sensor data, enabling continuous field reconstruction when dense deployment is impractical. However, common 2 x 2 and 2 x 3 evaluation protocols can leak spatial information through model selection and obscure true out-of-distribution (OOD) behavior. We propose a leakage-free 3 x 3 partition that separates training, validation, and testing in both space and time, so that model fitting, checkpoint selection, and final reporting are performed on distinct spatio-temporal domains. Under this stricter setting, we introduce DRIK (Distribution-Robust Inductive Kriging), a framework with three task-specific mechanisms: Spatial Continuity Regularization (SCR) perturbs coordinates to reduce dependence on one discretized graph; Masked Flow Disambiguation (MFD) prunes ambiguous propagation from zero-padded masked nodes; and Structural Domain Expansion (SDE) uses validation-node topology without labels to reduce train-inference structural mismatch. Experiments on six spatio-temporal datasets show that DRIK consistently outperforms state-of-the-art baselines, reducing MAE by up to 12.48% and achieving lower test-to-validation MAE ratios under leakage-free evaluation. These results indicate that robust inductive kriging requires both leakage-free evaluation and mechanisms that explicitly address the structural shifts introduced by unseen nodes.

cs.LG↗

arXiv2015

Efficiency Enhancement in Organic Solar Cells by Incorporating Silica-coated Gold Nanorods at the Buffer/Active interface

The performance of organic solar cells (OSCs) can be greatly improved by incorporating silica-coated gold nanorods (Au@SiO2 NRs) at the interface between the hole transporting layer and the active layer due to the plasmonic effect. The silica shell impedes the aggregation effect of the Au NRs in ethanol solution as well as the server charge recombination on the surface of the Au NRs otherwise they would bring forward serious reduction in open circuit voltage when incorporating the Au NRs at the positions in contact with the active materials. As a result, while the high open circuit voltage being maintained, the optimized plasmonic OSCs possess an increased short circuit current, and correspondingly an elevated power conversion efficiency with the enhancement factor of ~11%. The origin of performance improvement in OSCs with the Au@SiO2 NRs was analyzed systematically using morphological, electrical, optical characterizations along with theoretical simulation. It is found that the broadband enhancement in absorption, which yields the broadband enhancement in exciton generation in the active layer, is the major factor contributing to the increase in the short circuit current density. Simulation results suggest that the excitation of the transverse and longitudinal surface plasmon resonances of individual NRs as well as their mutual coupling can generate strong electric field near the vicinity of the NRs, thereby an improved exciton generation profile in the active layer. The incorporation of Au@SiO2 NRs at the interface between the hole transporting layer and the active layer also improves hole extraction in the OSCs.

physics.optics↗