Yuwen Tan

I'm a second-year PhD student in the Computer Science department at Boston University, where I am advised by Prof. Boqing Gong.
My current research interests focus on visual understanding in multimodal models and video generation.
Previously, I received my master's degree from Huazhong University of Science and Technology, where my research focused on continual learning and machine unlearning under the supervision of Prof. Xiang Xiang.

Email  /  CV  /  Scholar  /  Github

Research

I'm interested in computer vision, deep learning, and generative AI. Most of my research is about vision language models and continual learning. Some papers are highlighted.

We establish natural language camera movement understanding as a standalone task and introduce ACaM, an extensive benchmark and instruction-tuning framework spanning real and synthetic videos.

This paper reveals that many state-of-the-art large language models (LLMs) lack hierarchical knowledge about the visual world, failing to recognize even well-established biological taxonomies.

In this position paper, we propose to lift data-tracing machine unlearning to knowledge-tracing for foundation models (FMs). We support this position based on practical needs and insights from cognitive studies.

The proposed method eliminates the need for constructing an adapter pool and avoids retaining any image samples. Experimental results on five benchmarks demonstrate the effectiveness of our method which achieves the SOTA performance.

We formulate coarse-to-fine few-shot class-incremental learning (C2FSCIL) and propose Knowe, a simple and effective strategy for learning fine-grained classes without forgetting coarse-grained knowledge.

Work Experience

Miscellanea