Source author record

Xiantao Zhang

Xiantao Zhang appears in the imported research catalog. Authorship, coauthor and topic links are available while profile ownership is still unclaimed.

ResearcherUnclaimed source record

Computation and Language Computer Vision Information Retrieval physics.flu-dyn

Catalog footprint

What is connected

3works

4topics

4close collaborators

Actions

Connect this record

Open graph Browse works

Inspect adjacent papers, topics, institutions and collaborators without losing the researcher page.

Building this map preview

BZPEER is loading the nearby papers, people, topics and institutions for this page.

preprint2026arXiv

SpecEdit: Training-Free Acceleration for Diffusion based Image Editing via Semantic Locking

Diffusion-based image editing offers strong semantic controllability, but remains computationally expensive due to iterative high-resolution denoising over all spatial tokens. Dynamic-resolution sampling reduces this cost by performing early steps at reduced resolution. However, existing approaches prioritize upsampling using low-level heuristics such as edge detection or channel variance, which are weakly aligned with editing semantics and may lead to structural inconsistency. Moreover, spatial regions are often upsampled without verifying whether semantic modification is actually required, resulting in redundant high-resolution computation and accumulated errors. Therefore, we propose SpecEdit, a training-free dynamic-resolution framework tailored for diffusion-based image editing. SpecEdit follows a draft-and-verify scheme: a low-resolution draft first estimates the semantic outcome, after which token-level discrepancies are used to identify edit-relevant tokens for high-resolution denoising, while the remaining tokens stay at a coarse resolution. Experiments on Qwen-Image-Edit and FLUX.1-Kontext-dev demonstrate up to 10x and 7x acceleration, while maintaining strong quality. SpecEdit is complementary to step distillation and other acceleration techniques, achieving up to 13x speedup when combined with existing methods. Our code is in supplementary material and will be released on GitHub.

preprint2025arXiv

Roles of MLLMs in Visually Rich Document Retrieval for RAG: A Survey

Visually rich documents (VRDs) challenge retrieval-augmented generation (RAG) with layout-dependent semantics, brittle OCR, and evidence spread across complex figures and structured tables. This survey examines how Multimodal Large Language Models (MLLMs) are being used to make VRD retrieval practical for RAG. We organize the literature into three roles: Modality-Unifying Captioners, Multimodal Embedders, and End-to-End Representers. We compare these roles along retrieval granularity, information fidelity, latency and index size, and compatibility with reranking and grounding. We also outline key trade-offs and offer some practical guidance on when to favor each role. Finally, we identify promising directions for future research, including adaptive retrieval units, model size reduction, and the development of evaluation methods.

preprint2022arXiv

A discrete-module-finite-element hydroelasticity method in analyzing dynamic response of floating flexible structures

A discrete-module-finite element (DMFE) based hydroelasticity method has been proposed and well developed. Firstly, a freely floating flexible structure is discretized into several macro-submodules in two horizontal directions to perform a multi-rigid-body hydrodynamic analysis. Each macro-submodule is then abstracted to a lumped mass at the center of gravity that bears the external forces including inertia force, hydrodynamic force and hydrostatic force. Apart from external forces, all lumped masses are also subjected to structural forces that reflect the structural deformation features of the original flexible structure. The key to calculating the structural forces is derivation of the equivalent overall structural stiffness matrix with respect to the displacements of all lumped masses, which is tackled following the finite element procedure. More specifically, each macro-submodule is discretized into a number of microelements to derive the corresponding structural stiffness matrix, which is manipulated to a new one including only the nodes at the position of the lumped masses and surrounding boundaries by using the substructure approach, and subsequently the target overall stiffness matrix is obtained by combining together all macro-submodules. Finally, based on equivalence between external and structural forces, the DMFE method establishes the hydroelastic equation on all lumped masses with their displacements as unknown variables. Solving the equation gives the displacement response of all lumped masses. Displacement and structural force responses are first calculated on the interfaces of every two adjacent macro-submodules, after which at any given position of the flexible structure, the recovery of displacement is based on the structural stiffness matrix of the corresponding macro-submodule and the recovery of structural force uses the spline interpolation scheme.