Xuming Lin

Xuming Lin

Auto-generated profile, based on publicly available information

Xuming Lin

Ant Group, Hangzhou, Zhejiang, China

Research Focus & Metrics

Academic Career Length

4Years (2023-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
18 August 2023 in ACM Transactions on Information Systems
Verified
Journal Article
Open access
LJLiqiang Jing
Liqiang Jing
XSXuemeng Song
Xuemeng Song
XLXuming Lin
Xuming Lin
ZZZhongzhou Zhao
Zhongzhou Zhao
WZWei Zhou
Wei Zhou
LNLiqiang Nie
Liqiang Nie
Existing data-to-text generation efforts mainly focus on generating a coherent text from non-linguistic input data, such as tables and attribute–value pairs, but overlook that different application scenarios may require texts of different styles. Inspired by this, we define a new task, namely stylized data-to-text generation, whose aim is to generate coherent text for the given non-linguistic data according to a specific style. This task is non-trivial, due to three challenges: the logic of the generated text, unstructured style reference and biased training samples. To address these challenges, we propose a novel stylized data-to-text generation model, named StyleD2T, comprising three components: logic planning-enhanced data embedding, mask-based style embedding, and unbiased stylized text generation. In the first component, we introduce a graph-guided logic planner for attribute organization to ensure the logic of generated text. In the second component, we devise feature-level mask-based style embedding to extract the essential style signal from the given unstructured style reference. In the last one, pseudo triplet augmentation is utilized to achieve unbiased text generation, and a multi-condition based confidence assignment function is designed to ensure the quality of pseudo samples. Extensive experiments on a newly collected dataset from Taobao 1 have been conducted, and the results show the superiority of our model over existing methods.

9 November 2026
Verified
Conference Paper
Open access
JLJunyu Lu
Junyu Lu
XLXuming Lin
Xuming Lin
LWLuoxuan Weng
Luoxuan Weng
ZWZhen Wen
Zhen Wen
YWYiyao Wang
Yiyao Wang
WCWei Chen
Wei Chen
As the capabilities of large language models (LLMs) continue to advance, they are increasingly deployed for complex decision-making tasks. However, users' initial queries frequently exhibit intent ambiguity, necessitating multi-turn interactions to ultimately deliver satisfactory solutions. In this context, the text-only modality reveals dual limitations: for information presentation, plain text cannot efficiently deliver key information due to lengthy and unstructured responses; for information collection, users face high expressive barriers and cognitive load when attempting to precisely articulate multidimensional preferences. To address these challenges, we introduce the generative user interface (GenUI) paradigm into multi-turn decision-making scenarios. Specifically, we propose a dual-agent architecture that introduces an intermediary UI agent (IUA) between the user and the decision making agent (DMA). The IUA acts as a dynamic bridge, translating the DMA's logical deductions into interactive GenUI components for seamless user communication. To ensure accurate synchronization between these agents, we introduce a novel GenUI decision exchange (GUIDE) protocol, which enables the DMA to precisely transmit complex decision-making states and rationale to the IUA. Furthermore, to effectively power our proposed architecture, we develop a tailored three-stage data synthesis pipeline, which generates high-quality, domain-specific data to simultaneously train and optimize both the DMA and IUA. Comprehensive automated evaluations and extensive ablation studies validate the robustness of our architecture design and training strategy. A user study demonstrates the superiority of the GenUI modality in reducing users' cognitive load and improving their overall experience in decision-making tasks.