SearcharxivSearch

arXiv subjects

Jianchen Wang

Publications and source records attributed to Jianchen Wang.

3 recordsLinked to original sources

FractureFields: Contact-Aware Binary Multi-Field Transfer for Fractured 3D Gaussian Simulation

Physics-integrated 3D Gaussian representations make it possible to simulate image-reconstructed assets directly as particles, but current Gaussia-MPM pipelines keep a single Eulerian velocity field even after fracture. When disconnected fragments share interpolation support, they still write to and read from the same grid nodes, producing cross-fragment momentum leakage that appears as residual adhesion and non-physical stretching. We present FractureFields, a topology-adaptive transfer for fractured 3D Gaussian objects. After a structural event assigns persistent fragment identities, FractureFields builds fragment-specific mass and momentum fields in a single P2G pass, advances each field independently, and performs a field-aware G2P update so particles only sample their own fragment's grid state. To handle re-contact, we add a momentum-conserving contact projection that applies equal and opposite normal impulses only when two fragment fields are approaching, preserving free separation otherwise. Experiments on reconstructed scenes and a controlled re-contact benchmark show that fragment-conditioned routing eliminates realized cross-fragment mixing by construction, while contact projection reduces interpenetration during collision without reintroducing residual coupling. Overall, we argue that post-fracture simulation should treat structural disconnection as a change in local dynamical state, not merely a change in constitutive stress.

cs.CV

Efficiently Quantifying and Mitigating Ripple Effects in Model Editing

Large Language Models have revolutionized numerous tasks with their remarkable efficacy. However, editing these models, crucial for rectifying outdated or erroneous information, often leads to a complex issue known as the ripple effect in the hidden space. While difficult to detect, this effect can significantly impede the efficacy of model editing tasks and deteriorate model performance. This paper addresses this scientific challenge by proposing a novel evaluation methodology, Graphical Impact Evaluation(GIE), which quantitatively evaluates the adaptations of the model and the subsequent impact of editing. Furthermore, we introduce the Selective Impact Revision(SIR), a model editing method designed to mitigate this ripple effect. Our comprehensive evaluations reveal that the ripple effect in the hidden space is a significant issue in all current model editing methods. However, our proposed methods, GIE and SIR, effectively identify and alleviate this issue, contributing to the advancement of LLM editing techniques.

cs.CL

Xiezhi: An Ever-Updating Benchmark for Holistic Domain Knowledge Evaluation

New Natural Langauge Process~(NLP) benchmarks are urgently needed to align with the rapid development of large language models (LLMs). We present Xiezhi, the most comprehensive evaluation suite designed to assess holistic domain knowledge. Xiezhi comprises multiple-choice questions across 516 diverse disciplines ranging from 13 different subjects with 249,587 questions and accompanied by Xiezhi-Specialty and Xiezhi-Interdiscipline, both with 15k questions. We conduct evaluation of the 47 cutting-edge LLMs on Xiezhi. Results indicate that LLMs exceed average performance of humans in science, engineering, agronomy, medicine, and art, but fall short in economics, jurisprudence, pedagogy, literature, history, and management. We anticipate Xiezhi will help analyze important strengths and shortcomings of LLMs, and the benchmark is released in~\url{https://github.com/MikeGu721/XiezhiBenchmark}.

cs.CL