arXiv subjects

Yongzhe Zhang

Publications and source records attributed to Yongzhe Zhang.

10 recordsLinked to original sources

arXiv2026

MISO: Model-Internal-State-Guided Optimization for Ranking Models

Ranking models are repeatedly refined within established model families, yet the choice of which component to scale, replace, or retire is often guided by expensive trial-and-error. We present Model Internal State Optimization (MISO), a systems workflow that uses model internal states (MIS), including parameters, activations, gradients, and normalization statistics, to prioritize such local optimization decisions. MISO extracts MIS from a trained ranking model, aggregates them into ranking, alignment, and comparison signals, and converts those signals into a small set of interpretable candidate edits. Because MIS are re-extracted after each retraining cycle, MISO naturally supports an adaptive optimization workflow that tracks evolving model behavior as data distributions and system requirements shift over time. In an ads ranking case study, MISO improves normalized entropy while requiring substantially fewer validation runs than expert-driven and black-box scaling workflows, offering a practical middle ground between manual tuning and opaque automated search.

cs.IR↗

arXiv2025

LP-DETR: Layer-wise Progressive Relations for Object Detection

This paper presents LP-DETR (Layer-wise Progressive DETR), a novel approach that enhances DETR-based object detection through multi-scale relation modeling. Our method introduces learnable spatial relationships between object queries through a relation-aware self-attention mechanism, which adaptively learns to balance different scales of relations (local, medium and global) across decoder layers. This progressive design enables the model to effectively capture evolving spatial dependencies throughout the detection pipeline. Extensive experiments on COCO 2017 dataset demonstrate that our method improves both convergence speed and detection accuracy compared to standard self-attention module. The proposed method achieves competitive results, reaching 52.3\% AP with 12 epochs and 52.5\% AP with 24 epochs using ResNet-50 backbone, and further improving to 58.0\% AP with Swin-L backbone. Furthermore, our analysis reveals an interesting pattern: the model naturally learns to prioritize local spatial relations in early decoder layers while gradually shifting attention to broader contexts in deeper layers, providing valuable insights for future research in object detection.

cs.CV↗

arXiv2025

ChallengeMe: An Adversarial Learning-enabled Text Summarization Framework

The astonishing performance of large language models (LLMs) and their remarkable achievements in production and daily life have led to their widespread application in collaborative tasks. However, current large models face challenges such as hallucination and lack of specificity in content generation in vertical domain tasks. Inspired by the contrast and classification mechanisms in human cognitive processes, this paper constructs an adversarial learning-based prompt framework named ChallengeMe, which includes three cascaded solutions: generation prompts, evaluation prompts, and feedback optimization. In this process, we designed seven core optimization dimensions and set the threshold for adversarial learning. The results of mixed case studies on the text summarization task show that the proposed framework can generate more accurate and fluent text summaries compared to the current advanced mainstream LLMs.

cs.CL↗

arXiv2024

CoVis: A Collaborative Framework for Fine-grained Graphic Visual Understanding

Graphic visual content helps in promoting information communication and inspiration divergence. However, the interpretation of visual content currently relies mainly on humans' personal knowledge background, thereby affecting the quality and efficiency of information acquisition and understanding. To improve the quality and efficiency of visual information transmission and avoid the limitation of the observer due to the information cocoon, we propose CoVis, a collaborative framework for fine-grained visual understanding. By designing and implementing a cascaded dual-layer segmentation network coupled with a large-language-model (LLM) based content generator, the framework extracts as much knowledge as possible from an image. Then, it generates visual analytics for images, assisting observers in comprehending imagery from a more holistic perspective. Quantitative experiments and qualitative experiments based on 32 human participants indicate that the CoVis has better performance than current methods in feature extraction and can generate more comprehensive and detailed visual descriptions than current general-purpose large models.

cs.CV↗

arXiv2023

Nonconvex ancient solutions to Curve Shortening Flow

We construct an ancient solution to planar curve shortening. The solution is at all times compact and embedded. For $t\ll0$ it is approximated by the rotating Yin-Yang soliton, truncated at a finite angle $α(t) = -t$, and closed off by a small copy of the Grim Reaper translating soliton.

math.DG↗

arXiv2020

Superconvexity of the Heat Kernel on Hyperbolic Space

We prove a conjecture of Bernstein that the superconvexity of the heat kernel on hyperbolic space holds in all dimensions and, hence, there is an analog of Huisken's monotonicity formula for mean curvature flow in hyperbolic space of all dimensions.

math.DG↗

arXiv2020

FastSV: A Distributed-Memory Connected Component Algorithm with Fast Convergence

This paper presents a new distributed-memory algorithm called FastSV for finding connected components in an undirected graph. Our algorithm simplifies the classic Shiloach-Vishkin algorithm and employs several novel and efficient hooking strategies for faster convergence. We map different steps of FastSV to linear algebraic operations and implement them with the help of scalable graph libraries. FastSV uses sparse operations to avoid redundant work and optimized MPI communication to avoid bottlenecks. The resultant algorithm shows high-performance and scalability as it can find the connected components of a hyperlink graph with over 134B edges in 30 seconds using 262K cores on a Cray XC40 supercomputer. FastSV outperforms the state-of-the-art algorithm by an average speedup of 2.21x (max 4.27x) on a variety of real-world graphs.

cs.DS↗

arXiv2018

Composing Optimization Techniques for Vertex-Centric Graph Processing via Communication Channels

Pregel's vertex-centric model allows us to implement many interesting graph algorithms, where optimization plays an important role in making it practically useful. Although many optimizations have been developed for dealing with different performance issues, it is hard to compose them together to optimize complex algorithms, where we have to deal with multiple performance issues at the same time. In this paper, we propose a new approach to composing optimizations, by making use of the \emph{channel} interface, as a replacement of Pregel's message passing and aggregator mechanism, which can better structure the communication in Pregel algorithms. We demonstrate that it is convenient to optimize a Pregel program by simply using a proper channel from the channel library or composing them to deal with multiple performance issues. We intensively evaluate the approach through many nontrivial examples. By adopting the channel interface, our system achieves an all-around performance gain for various graph algorithms. In particular, the composition of different optimizations makes the S-V algorithm 2.20x faster than the current best implementation.

cs.DC↗

arXiv2017

Palgol: A High-Level DSL for Vertex-Centric Graph Processing with Remote Data Access

Pregel is a popular distributed computing model for dealing with large-scale graphs. However, it can be tricky to implement graph algorithms correctly and efficiently in Pregel's vertex-centric model, especially when the algorithm has multiple computation stages, complicated data dependencies, or even communication over dynamic internal data structures. Some domain-specific languages (DSLs) have been proposed to provide more intuitive ways to implement graph algorithms, but due to the lack of support for remote access --- reading or writing attributes of other vertices through references --- they cannot handle the above mentioned dynamic communication, causing a class of Pregel algorithms with fast convergence impossible to implement. To address this problem, we design and implement Palgol, a more declarative and powerful DSL which supports remote access. In particular, programmers can use a more declarative syntax called chain access to naturally specify dynamic communication as if directly reading data on arbitrary remote vertices. By analyzing the logic patterns of chain access, we provide a novel algorithm for compiling Palgol programs to efficient Pregel code. We demonstrate the power of Palgol by using it to implement several practical Pregel algorithms, and the evaluation result shows that the efficiency of Palgol is comparable with that of hand-written code.

cs.DC↗

arXiv2014

Metal-Dielectric-Graphene Sandwich for Surface Enhanced Raman Spectroscopy

Raman intensity of Rhodamine B (RhB) is enhanced by inserting a thin high \k{appa} dielectric layer which reduces the surface plasmon damping at the gold-graphene interface. The results indicate that the Raman intensity increases sharply by plasmonic resonance enhancement while maintaining efficient fluorescence quenching with optimized dielectric layer thickness.

physics.optics↗