Picture for Weixin Liang

Weixin Liang

On the nonlinear correlation of ML performance between data subpopulations

May 04, 2023
Weixin Liang, Yining Mao, Yongchan Kwon, Xinyu Yang, James Zou

Figure 1 for On the nonlinear correlation of ML performance between data subpopulations

Figure 2 for On the nonlinear correlation of ML performance between data subpopulations

Figure 3 for On the nonlinear correlation of ML performance between data subpopulations

Figure 4 for On the nonlinear correlation of ML performance between data subpopulations

* Accepted to the main conference of ICML 2023

Access Paper or Ask Questions

SEAL : Interactive Tool for Systematic Error Analysis and Labeling

Oct 11, 2022
Nazneen Rajani, Weixin Liang, Lingjiao Chen, Meg Mitchell, James Zou

Figure 1 for SEAL : Interactive Tool for Systematic Error Analysis and Labeling

Figure 2 for SEAL : Interactive Tool for Systematic Error Analysis and Labeling

Figure 3 for SEAL : Interactive Tool for Systematic Error Analysis and Labeling

Figure 4 for SEAL : Interactive Tool for Systematic Error Analysis and Labeling

* Accepted at EMNLP 2022 demo track

Access Paper or Ask Questions

GSCLIP : A Framework for Explaining Distribution Shifts in Natural Language

Jun 30, 2022
Zhiying Zhu, Weixin Liang, James Zou

Figure 1 for GSCLIP : A Framework for Explaining Distribution Shifts in Natural Language

Figure 2 for GSCLIP : A Framework for Explaining Distribution Shifts in Natural Language

Figure 3 for GSCLIP : A Framework for Explaining Distribution Shifts in Natural Language

* Accepted by ICML 2022 DataPerf

Access Paper or Ask Questions

Mind the Gap: Understanding the Modality Gap in Multi-modal Contrastive Representation Learning

Mar 03, 2022
Weixin Liang, Yuhui Zhang, Yongchan Kwon, Serena Yeung, James Zou

Figure 1 for Mind the Gap: Understanding the Modality Gap in Multi-modal Contrastive Representation Learning

Figure 2 for Mind the Gap: Understanding the Modality Gap in Multi-modal Contrastive Representation Learning

Figure 3 for Mind the Gap: Understanding the Modality Gap in Multi-modal Contrastive Representation Learning

Figure 4 for Mind the Gap: Understanding the Modality Gap in Multi-modal Contrastive Representation Learning

* Our code and data are available at https://modalitygap.readthedocs.io/

Access Paper or Ask Questions

MetaShift: A Dataset of Datasets for Evaluating Contextual Distribution Shifts and Training Conflicts

Feb 14, 2022
Weixin Liang, James Zou

Figure 1 for MetaShift: A Dataset of Datasets for Evaluating Contextual Distribution Shifts and Training Conflicts

Figure 2 for MetaShift: A Dataset of Datasets for Evaluating Contextual Distribution Shifts and Training Conflicts

Figure 3 for MetaShift: A Dataset of Datasets for Evaluating Contextual Distribution Shifts and Training Conflicts

Figure 4 for MetaShift: A Dataset of Datasets for Evaluating Contextual Distribution Shifts and Training Conflicts

* ICLR 2022. Code & data available at https://github.com/Weixin-Liang/MetaShift

Access Paper or Ask Questions