Teng Wang 王腾

I am a Ph.D. candidate (2020-Now) at the Department of Computer Science, The University of Hong Kong (HKU), fortunately supervised by Prof. Ping Luo and Prof. Feng Zheng. Before that, I obtained my B.E. and M.E. degrees from Sun Yat-sen University (SYSU) under the supervision of Prof. Huicheng Zheng. I was a research intern at Tencent AI Lab and Tencent Data Platform.

My recent research interests lie in vision-language multimodal learning and video understanding.

Email  /  CV  /  Google Scholar  /  Github

(* indicates equal contribution)

Improve the generalization by overcoming the hallucination problem in LLMs-based visual models.

Wikipedia knowledge enhanced prompt tuning for CLIP-based few-shot classification.

Caption Anything generates descriptive captions for any object within an image, offering a range of language styles to accommodate diverse user preferences. It supports visual controls (mouse click) and language controls (length, sentiment, factuality, and language).

Learning video-language representation by solving a bidirectional set prediction problem. We won the first place in both video grounding and video captioning tracks of PIC Challenge 2022 .

We accelerate the convergence of language modeling by predicting 100% tokens like Auto-Regressive (AR) meanwhile achieving competitive performance with Masked Language Modeling (MLM).

Vision-language learning from stand-alone image and text corpora by cross-modal augmentation.

The first DETR-style parallel decoding paradigm for dense video captioning.

Hierarchical visual representation across scene, event, and frame levels in untrimmed videos.