Xiaoyi Wang

ORCID: 0000-0003-0533-5080
Publications
Citations
Views
---
Saved
---
About
Contact & Profiles
Research Areas
  • Topic Modeling
  • Natural Language Processing Techniques
  • Speech and dialogue systems
  • CRISPR and Genetic Engineering
  • Meat and Animal Product Quality
  • Recommender Systems and Techniques
  • Genetic and phenotypic traits in livestock
  • Sentiment Analysis and Opinion Mining
  • Renal and related cancers
  • Educational Technology and Assessment
  • Machine Learning in Healthcare
  • Pluripotent Stem Cells Research
  • Advanced Graph Neural Networks
  • Natural Products and Biological Research

Yunnan Agricultural University
2024

Capital Normal University
2023-2024

Albert Einstein College of Medicine
2015

Despite their impressive capacities, Large language models (LLMs) often struggle with the hallucination issue of generating inaccurate or fabricated content even when they possess correct knowledge. In this paper, we extend exploration correlation between hidden-state prediction changes and output factuality into a deeper, token-wise level. Based on insights , propose cross-layer Entropy eNhanced Decoding (END), decoding method that mitigates hallucinations without requiring extra training....

10.48550/arxiv.2502.03199 preprint EN arXiv (Cornell University) 2025-02-05

Backfat thickness (BT) and intramuscular fat (IMF) content are closely appertained to meat production quality in pig production. Deposition subcutaneous adipose (SA) IMF concerns different genes regulatory mechanisms. And larger studies with rigorous design should be carried explore the molecular regulation of deposition tissues. The purpose this study is gain a better understanding mechanisms underlying differences among tissues identify tissue-specific involved regulating deposition....

10.1016/j.heliyon.2024.e31311 article EN cc-by-nc-nd Heliyon 2024-05-01

Document-level relational extraction requires reading, memorization, and reasoning to discover relevant factual information in multiple sentences. It is difficult for the current hierarchical network graph methods fully capture structural behind document make natural from context. Different previous methods, this article reconstructs relation task into a machine reading comprehension task. Each pair of entities relationships characterized by question template, translated identifying answers...

10.1145/3666042 article EN ACM Transactions on Asian and Low-Resource Language Information Processing 2024-06-01

Hepatocytes derived from human somatic cells would be useful in regenerative medicine, drug development and cell-based disease models. Several types of have been reprogrammed to induced pluripotent (iPSCs) then differentiated hepatocyte-like (iHeps). However, the method for generating such renal epithelial shed urine has not described systematically. Thus, we tested whether these cell-derived iPS will show ability differentiate into iHeps may potential engraft mouse liver. 250 – 500 ml fresh...

10.1055/s-0035-1567975 article EN Zeitschrift für Gastroenterologie 2015-12-14

Abstract Conversational recommender system (CRS) needs to be seamlessly integrated between the two modules of recommendation and dialog, aiming recommend high-quality items users through multiple rounds interactive dialogs. Items can typically refer goods, movies, news, etc. Through this form express their preferences in real time, fully understand user’s thoughts corresponding items. Although mainstream dialog systems have improved performance some extent, there are still key issues, such...

10.1017/s1351324923000451 article EN cc-by Natural Language Engineering 2023-09-08

Pre-trained models acquire knowledge from vast amounts of unannotated and unstructured data through self-supervised learning. However, they suffer limitations such as inadequate performance limited reasoning capabilities due to the lack external guidance. To address these limitations, integrating structured graphs into pretrained enables them both general semantic free text real-world behind text, thereby effectively addressing downstream knowledge-driven tasks. This paper introduces...

10.1109/iceace60673.2023.10442824 article EN 2023-12-29

Automatic essay scoring (AES) refers to the use of computers automatically score essays without human intervention. Whether discourse structure is reasonable an important consideration dimension in narrative scoring. At present, research on text at home and abroad still its infancy, there a lack open corpus for training. In addition, most evaluation models constructed past are trained by selecting features based manual experience. To solve above problems, this paper develops composition...

10.1109/iceace60673.2023.10442671 article EN 2023-12-29
Coming Soon ...