arXiv subjects

Pengchao Tian

Publications and source records attributed to Pengchao Tian.

3 recordsLinked to original sources

arXiv2026

Efficient Causal Structure Learning via Modular Subgraph Integration

Learning causal structures from observational data remains a fundamental yet computationally intensive task, particularly in high-dimensional settings where existing methods face challenges such as the super-exponential growth of the search space and increasing computational demands. To address this, we introduce VISTA (Voting-based Integration of Subgraph Topologies for Acyclicity), a modular framework that decomposes the global causal structure learning problem into local subgraphs based on Markov Blankets. The global integration is achieved through a weighted voting mechanism that penalizes low-support edges via exponential decay, filters unreliable ones with an adaptive threshold, and ensures acyclicity using a Feedback Arc Set (FAS) algorithm. The framework is model-agnostic, imposing no assumptions on the inductive biases of base learners, is compatible with arbitrary data settings without requiring specific structural forms, and fully supports parallelization. We also theoretically establish finite-sample error bounds for VISTA, and prove its asymptotic consistency under mild conditions. Extensive experiments on both synthetic and real datasets consistently demonstrate the effectiveness of VISTA, yielding notable improvements in both accuracy and efficiency over a wide range of base learners.

stat.ML↗

arXiv2025

Improving VQA Reliability: A Dual-Assessment Approach with Self-Reflection and Cross-Model Verification

Vision-language models (VLMs) have demonstrated significant potential in Visual Question Answering (VQA). However, the susceptibility of VLMs to hallucinations can lead to overconfident yet incorrect answers, severely undermining answer reliability. To address this, we propose Dual-Assessment for VLM Reliability (DAVR), a novel framework that integrates Self-Reflection and Cross-Model Verification for comprehensive uncertainty estimation. The DAVR framework features a dual-pathway architecture: one pathway leverages dual selector modules to assess response reliability by fusing VLM latent features with QA embeddings, while the other deploys external reference models for factual cross-checking to mitigate hallucinations. Evaluated in the Reliable VQA Challenge at ICCV-CLVL 2025, DAVR achieves a leading $\Phi_{100}$ score of 39.64 and a 100-AUC of 97.22, securing first place and demonstrating its effectiveness in enhancing the trustworthiness of VLM responses.

cs.CV↗

arXiv2024

Two-Stage Optimization for Efficient V2G Coordination in Distribution Power System

With the growing popularity of electric vehicles (EVs), maintaining power grid stability has become a significant challenge. To address this issue, EV scheduling control strategies have been developed to manage vehicle-to-grid (V2G) in coordination with the optimal power flow. In existing studies, such coordination optimization is formulated as a mixed-integer nonlinear programming (MINP), which is computationally challenging due to the binary EV charging and discharging variables. To address this challenge, we develop an efficient two-stage optimization method for this mixed-integer nonlinear coordination problem. This method first employs an efficient technique called the difference of convex (DC) to relax the integrality and reformulate MINP into a series of path-following continuous programming. Although the DC approach shows promising efficiency for solving MINP, it cannot guarantee the feasibility of the solutions. Consequently, we propose a trust region optimization method in stage two that constructs a trust region around DC's solution and then searches for the best feasible solution within this region. Our simulation results demonstrate that, compared to the open-source optimization solver SCIP, our proposed method significantly enhances computational efficiency while achieving near optimality.

cs.CE↗