Publications

Beyond Invariance: Test-Time Label-Shift Adaptation for Distributions with “Spurious” Correlations

Published in ICML 2023 (in preparation), 2022

Instead of try to eliminate spurious correlations through invariance, we exploit them instead via Test-Time Label-Shift Adaptation (TTLSA).

Download here

Individualized and Global Feature Attributions for Gradient Boosted Trees in the Presence of l2 Regularization

Published in AISTATS 2022 (submitted), 2022

Connecting local and global feature attribution with TreeInner.

Download here

Action2Motion: Conditioned Generation of 3D Human Motions

Published in ACM Multimedia, 2020

This paper considers an inverse problem of action recognition: given a prescribed action type, we aim to generate plausible human motion sequences in 3D.

Download here

Qingyao Sun

Publications

Beyond Invariance: Test-Time Label-Shift Adaptation for Distributions with “Spurious” Correlations

Individualized and Global Feature Attributions for Gradient Boosted Trees in the Presence of l2 Regularization

Action2Motion: Conditioned Generation of 3D Human Motions