About Me

I am currently a Postdoctoral Research Fellow in MBZUAI and CMU, working with Prof. Kun Zhang . I am broadly interested in computer vision and machine learning. My current research focuses on causal representation learning, attention learning, prompt learning, and video understanding.

Prior to that, I received my Ph.D degree at Tsinghua University in 2021, advised by Prof. Jie Zhou and Prof. Jiwen Lu ; and obtained B.Eng. at Tsinghua University in 2016.

Feel free to drop me an e-mail guangyichen1994(at)gmail.com , if you are interested in our research and want to discuss relevant research topics or potential collaborations.

News

  • 2024-10: 3 papers on casaul representation learning and 1 paper on causal discovery are accepted by NeurIPS'2024.
  • 2024-09: 1 paper on understanding LLMs's representation is accepted by EMNLP'2024.
  • 2024-09: I give a talk at swarma(集智) about causal representation learning [slides].
  • 2024-07: 1 paper on efficient generation of LLMs is accepted by ECCV'2024.
  • 2024-07: Call for Papers for our Causal Representation Learning Workshop@NeurIPS 2024: Welcome to submit your paper at Openreview!
  • 2024-07: I give a talk at Pacific Causal Inference Conference (PCIC) about causal representation learning for video understanding [slides].
  • 2024-06: I co-orginize the Causal Representation Learning Workshop at ICDM 2024. See you at Abu Dhabi in Deceber.
  • 2024-06: I give a talk at Valse to discuss prompt learning and visual context [slides], thanks Prof. Tianfei Zhou for the invitation!
  • 2024-05: 2 papers on causal representation learning are accepted by ICML'2024.
  • 2024-02: 1 paper on video understanding is accepted by CVPR'2024.
  • 2024-01: 3 papers on visaul reasonsing and causality are accepted by ICLR'2024.
  • 2024-01: 1 paper on graph representation learning is accepted by WWW'2024.
  • 2023-12: 1 paper on prompt learning is accepted by AAAI'2024 as Oral.
  • 2023-09: 3 paper (1 Spotlight) on causal representation learning, domain adaptation, and trajectory prediction are accepted by NeurIPS'2023.
  • 2023-07: 1 paper on visual reasoning is accepted by ICCV'2023.
  • 2023-04: 1 paper on graph representation learning is accepted by ICML'2023.
  • 2023-03: Our paper using a causal perspective to understand MAE is selected as Highlight (top 2.5%) at CVPR'2023.
  • 2023-02: 3 papers on causality, trajectory prediction, and prompt learning are accepted by CVPR'2023.
  • 2023-01: 2 papers on prompt learning (PLOT as Spotlight) and video understanding are accepted by ICLR'2023.
  • Research Topics

    sq-sample26

    Tem-adapter: Adapting Image-Text Pretraining for Video Question Answer

    We present Tem-Adapter, a method that improves VQA by leveraging image-based knowledge and introducing temporal and semantic aligners.
    sq-sample26

    Unsupervised Sampling Promoting for Stochastic Human Trajectory Prediction

    We propose to formualte the sampling process with Baysesian optimization to promote stochasitic human trajectory prediction in an unsupervised manner.
    sq-sample26

    Understanding Masked Autoencoders via Hierarchical Latent Variable Models

    We propose a causal perspective to understand the underlying mechanism of MAE to identify latent variables in the hierarchical model.
    sq-sample26

    GAIN: On the Generalization of Instructional Action Understanding

    We present a benchmark, named GAIN, to analyze the generalizability of instructional action understanding models.
    sq-sample26

    Stochastic Trajectory Prediction via Motion Indeterminacy Diffusion

    We propose a new framework to formulate the trajectory prediction task as a reverse process of motion indeterminacy diffusion.
    sq-sample26

    Spatial-Temporal Attention-aware Learning for Video-based Person Re-identification

    We propose a spatial-temporal attention to jointly discover the salient clues in both spatial and temporal domain.
    sq-sample26

    Probabilistic Temporal Modeling for Unintentional Action Localization

    We propose a probabilistic framework with dense predictions to allow the uncertainty of annotations for unintentional action localization.
    sq-sample26

    Unintentional Action Localization via Counterfactual Examples

    We propose to disentangle the effects of content and intention clues by building a counterfactual video pool and learning hidden causal processes contrastively.
    sq-sample26

    FineDiving: A Fine-grained Dataset for Procedure-aware Action Quality Assessment

    We construct a new fine-grained dataset for the explainable action quality assessment, named FineDiving.
    sq-sample26

    Human Trajectory Prediction via Counterfactual Analysis

    We propose a counterfactual analysis method for human trajectory prediction to investigate the causality between the predictions and inputs, and alleviate the negative effects brought by environment bias.
    sq-sample26

    Personalized Trajectory Prediction via Distribution Discrimination

    We present a distribution discrimination (DisDis) method to predict personalized motion patterns by distinguishing the potential distributions in a self-supervised manner.
    sq-sample26

    Counterfactual Attention Learning for Fine-Grained Visual Categorization and Re-identification

    We propose to learn the attention with counterfactual causality, which provides a tool to measure the attention quality and a powerful supervisory signal to guide the learning process.
    sq-sample26

    Person Re-identification via Attention Pyramid

    We propose attention pyramid networks by the "split-attend-merge-stack" principle to jointly learn the attentions under different scales.
    sq-sample26

    Temporal Coherence or Temporal Motion: Which is More Critical for Video-based Person Re-identification?

    We show temporal coherence plays a more critical role than temporal motion for video-based person ReID and develop an adversarial feature augmentation to highlight temporal coherence.
    sq-sample26

    Deep Credible Metric Learning for Unsupervised Domain Adaptation Person Re-identification

    We propose to adaptively and progressively mine credible training samples to avoid the damage from the noise of predicted pseudo labels for unsupervised domain adaptation person ReID.
    sq-sample26

    Self-Critical Attention Learning for Person Re-Identification

    We present a self-critical attention learning method which applies a critic module to examine and supervise the attention model.
    sq-sample26

    Prompt Learning with Optimal Transport for Vision-Language Models

    we propose to learn multiple comprehensive prompts with optimal transport to adapt the pretrained vision-laguage model.
    sq-sample26

    Adversarial Alignment for Source Free Object Detection

    Adversarial Alignment proposes a variance-based criterion of detection to build pseudo supervisions for source free adaptative object detection.
    sq-sample26

    DenseCLIP: Language-Guided Dense Prediction with Context-Aware Prompting

    DenseCLIP is a new framework for dense prediction by implicitly and explicitly adapting the pre-trained knowledge from CLIP.
    sq-sample26

    Temporally Disentangled Representation Learning

    A framework to recover time-delayed latent causal variables and identify their relations from sequential data under stationary environments or different distribution shifts.
    sq-sample26

    Partial Disentanglement for Domain Adaptation

    A framework with partial identifiablity to minimize unnecessary influences of domain shift with minimal changes of causal mechanisms.

    Publications

    Some selected recent publications. Please see Google Scholar for details.

    Honors and Awards

  • Jiang Zhen Scholarship, Tsinghua University, 2020
  • 2nd place in Semi-Supervised Recognition Challenge at FGVC7, CVPR, 2020
  • Samsung Scholarship, 2019
  • Academic Excellence Scholarship, Tsinghua University, 2015
  • Tsinghua Scholarship, Tsinghua University, 2014
  • National Encouragement Scholarship, Ministry of Education of P.R. China, 2014
  • National Encouragement Scholarship, Ministry of Education of P.R. China, 2013
  • Teaching

  • TA Analog Electronic Technology Foundation, Tsinghua University, 2018.
  • TA Analog Electronic Technology Foundation, Tsinghua University, 2019.
  • TA Numerical Analysis and Algorithm, Tsinghua University, 2019.
  • TA Probabilistic and Statistical Inference (ML703), MBZUAI,2021, Fall.
  • TA Probabilistic and Statistical Inference (ML703), MBZUAI,2022, (Spring, Fall).
  • TA Advanced Probabilistic and Statistical Inference (ML803), MBZUAI,2023, (Spring, Fall).
  • TA Advanced Probabilistic and Statistical Inference (ML803), MBZUAI,2024, Spring.
  • Academic Services

  • Publicity Chair: for CLeaR 2023.
  • Co-organizer: for the ICME 2019 workshop: The Third Workshop on Human Identification in Multimedia (HIM'19) [website]
  • Conference Reviewer / Program Committee Member: CVPR, ICCV, ECCV, NeurIPS, ICML, ICLR, AAAI, and so on.
  • Journal Reviewer: TPAMI, TIP, IJCV, TNNLS, TMM, TCSVT, and so on.