Výsledky vyhledávání

Report

Enhancing Multiple Dimensions of Trustworthiness in LLMs via Sparse Activation Control

Autor: Xiao, Yuxin, Wan, Chaoqun, Zhang, Yonggang, Wang, Wenxiao, Lin, Binbin, He, Xiaofei, Shen, Xu, Ye, Jieping

As the development and application of Large Language Models (LLMs) continue to advance rapidly, enhancing their trustworthiness and aligning them with human preferences has become a critical area of research. Traditional methods rely heavily on exten

Externí odkaz: http://arxiv.org/abs/2411.02461

Zobrazit plný text záznamu

Report

SFTMix: Elevating Language Model Instruction Tuning with Mixup Recipe

Autor: Xiao, Yuxin, Zhang, Shujian, Zhou, Wenxuan, Ghassemi, Marzyeh, Zhao, Sanqiang

To induce desired behaviors in large language models (LLMs) for interaction-driven tasks, the instruction-tuning stage typically trains LLMs on instruction-response pairs using the next-token prediction (NTP) loss. Previous work aiming to improve ins

Externí odkaz: http://arxiv.org/abs/2410.05248

Zobrazit plný text záznamu

Report

In the Name of Fairness: Assessing the Bias in Clinical Record De-identification

Autor: Xiao, Yuxin, Lim, Shulammite, Pollard, Tom Joseph, Ghassemi, Marzyeh

Data sharing is crucial for open science and reproducible research, but the legal sharing of clinical data requires the removal of protected health information from electronic health records. This process, known as de-identification, is often achieve

Externí odkaz: http://arxiv.org/abs/2305.11348

Zobrazit plný text záznamu

Report

Uncertainty Quantification with Pre-trained Language Models: A Large-Scale Empirical Analysis

Autor: Xiao, Yuxin, Liang, Paul Pu, Bhatt, Umang, Neiswanger, Willie, Salakhutdinov, Ruslan, Morency, Louis-Philippe

Pre-trained language models (PLMs) have gained increasing popularity due to their compelling prediction performance in diverse natural language processing (NLP) tasks. When formulating a PLM-based prediction pipeline for NLP tasks, it is also crucial

Externí odkaz: http://arxiv.org/abs/2210.04714

Zobrazit plný text záznamu