Search

Your search keyword '"Li, Yaxi"' showing total 437 results
Start Over Author "Li, Yaxi" Publication Year Range Last 10 years
437 results on '"Li, Yaxi"'

1. SLM Meets LLM: Balancing Latency, Interpretability and Consistency in Hallucination Detection

Author

Hu, Mengya, Xu, Rui, Lei, Deren, Li, Yaxi, Wang, Mingyu, Ching, Emily, Kamal, Eslam, and Deng, Alex

Subjects
Computer Science - Computation and Language, Computer Science - Artificial Intelligence, Computer Science - Machine Learning
Abstract

Large language models (LLMs) are highly capable but face latency challenges in real-time applications, such as conducting online hallucination detection. To overcome this issue, we propose a novel framework that leverages a small language model (SLM) classifier for initial detection, followed by a LLM as constrained reasoner to generate detailed explanations for detected hallucinated content. This study optimizes the real-time interpretable hallucination detection by introducing effective prompting techniques that align LLM-generated explanations with SLM decisions. Empirical experiment results demonstrate its effectiveness, thereby enhancing the overall user experience., Comment: preprint under review

Published
2024

5. Chain of Natural Language Inference for Reducing Large Language Model Ungrounded Hallucinations

Author

Lei, Deren, Li, Yaxi, Hu, Mengya, Wang, Mingyu, Yun, Vincent, Ching, Emily, and Kamal, Eslam

Subjects
Computer Science - Computation and Language, Computer Science - Artificial Intelligence
Abstract

Large language models (LLMs) can generate fluent natural language texts when given relevant documents as background context. This ability has attracted considerable interest in developing industry applications of LLMs. However, LLMs are prone to generate hallucinations that are not supported by the provided sources. In this paper, we propose a hierarchical framework to detect and mitigate such ungrounded hallucination. Our framework uses Chain of Natural Language Inference (CoNLI) for hallucination detection and hallucination reduction via post-editing. Our approach achieves state-of-the-art performance on hallucination detection and enhances text quality through rewrite, using LLMs without any fine-tuning or domain-specific prompt engineering. We show that this simple plug-and-play framework can serve as an effective choice for hallucination detection and reduction, achieving competitive performance across various contexts., Comment: The source code is available at https://github.com/microsoft/CoNLI_hallucination

Published
2023

17. Faslpr gene dosage tunes the extent of lymphoproliferation and T cell differentiation in lupus

Author

Bohat, Ritu, Liang, Xiaofang, Chen, Yanping, Xu, Chunyu, Zheng, Ningbo, Guerrero, Ashley, Hou, Jiakai, Jaffery, Roshni, Egan, Nicholas A., Li, Yaxi, Tang, Yitao, Unsal, Esra, Robles, Adolfo, Chen, Si, Major, Angela M., Elldakli, Hadil, Chung, Sang-Hyuk, Liang, Han, Hicks, M. John, Du, Yong, Lin, Jamie S., Chen, Xiqun, Mohan, Chandra, and Peng, Weiyi

Published
2024
Full Text
View/download PDF

38. Shear‐Strained Pd Single‐Atom Electrocatalysts for Nitrate Reduction to Ammonia.

Author

Liu, Yunliang, Zhuang, Zechao, Liu, Yixian, Liu, Naiyun, Li, Yaxi, Cheng, Yuanyuan, Yu, Jingwen, Yu, Ruohan, Wang, Dingsheng, and Li, Haitao

Subjects
ATOMIC hydrogen, SHEAR strain, ACTIVATION energy, SHEARING force, COPPER, DENITRIFICATION
Abstract

Electrochemical nitrate reduction method (NitRR) is a low‐carbon, environmentally friendly, and efficient method for synthesizing ammonia, which has received widespread attention in recent years. Copper‐based catalysts have a leading edge in nitrate reduction due to their good adsorption of *NO3. However, the formation of active hydrogen (*H) on Cu surfaces is difficult and insufficient, resulting in a large amount of the by‐product NO2−. In this work, Pd single atoms suspended on the interlayer unsaturated bonds of CuO atoms formed due to dislocations (Pd−CuO) were prepared by low temperature treatment, and the Pd single atoms located on the dislocations were subjected to shear stress and the dynamic effect of support formation to promote the conversion of nitrate into ammonia. The catalysis had an ammonia yield of 4.2 mol. gcat−1. h−1, and a Faraday efficiency of 90 % for ammonia production at −0.5 V vs. RHE. Electrochemical in situ characterization and theoretical calculations indicate that the dynamic effects of Pd single atoms and carriers under shear stress obviously promote the production of active hydrogen, reduce the reaction energy barrier of the decision‐making step for nitrate conversion to ammonia, further promote ammonia generation. [ABSTRACT FROM AUTHOR]

Published
2024
Full Text
View/download PDF

Catalog

Books, media, physical & digital resources
See catalog results