|
Boshen Xu AboutMy research interests include video understanding (e.g., action recognition, vision-language pretraining, etc.) and embodied AI (e.g., 3D object assembly, 3D HOI reconstruction, etc.). I'm currently focusing on egocentric vision and related topics that benefit VR/AR/Embodied AI. I am a second-year PhD student at Renmin University of China (RUC) under the supervision of Professor Qin Jin at AIM3 Lab. Prior to joining RUC, I got my bachelor degree from School of Computer Science and Engineering, University of Electronic Science and Technology of China (UESTC). Email / CV / Github / Google Scholar |
|
Publications* denotes equal contributions. |
|
EgoDTM: Towards 3D-Aware Egocentric Video-Language Pretraining
Boshen Xu, Yuting Mei, Xinbi Liu, Sipeng Zheng, Qin Jin arxiv, 2025 code / arxiv We introduce EgoDTM, an Egocentric Depth- and Text-aware Model that bridges the gap between 2D visual understanding and 3D spatial awareness. |
|
TimeZero: Temporal Video Grounding with Reasoning-Guided LVLM
Ye Wang*, Boshen Xu*, Zihao Yue, Zihan Xiao , Ziheng Wang, Liang Zhang , Dingyi Yang, Wenxuan Wang, Qin Jin arxiv, 2025 code / paper We propose TimeZero, a reasoning-guided LVLM for temporal video grounding that extends inference through reinforcement learning to reason about video-language relationships. Achieves state-of-the-art performance on Charades-STA benchmark. |
|
Do Egocentric Video-Language Models Truly Understand Hand-Object Interactions?
Boshen Xu, Ziheng Wang*, Yang Du* , Zhinan Song, Sipeng Zheng, Qin Jin ICLR, 2025 code / paper We propose EgoNCE++, an asymmetric contrastive learning pretraining objective to solve the EgoVLMs' weakness in distinguishing HOI combinations with word variations. |
|
SPAFormer: Sequential 3D Part Assembly with Transformers
Boshen Xu, Sipeng Zheng, Qin Jin 3DV, 2025 project page / code / arxiv We present SPAFormer, a transformer-based framework that leverages assembly sequences constraints with three part encodings to address the combinatorial explosion challenge in 3D-PA task. |
|
Unveiling Visual Biases in Audio-Visual Localization Benchmarks
Liangyu Chen, Zihao Yue, Boshen Xu, Qin Jin ECCV AVGenL Workshop, 2024 arxiv We reveal that current audio-visual source localization benchmarks (VGG-SS, Epic-Sounding-Object) are easily hacked by vision-only models, therefore calling for a benchmark that requires more audio cues. |
|
Prompt-Oriented View-Agnostic Learning for Egocentric Hand-Object Interaction in the Multi-View World
Boshen Xu, Sipeng Zheng, Qin Jin ACM MM, 2023 project page / code / arxiv We propose POV, a view adaptation framework that enables transfer learning from multi-view third-person videos to egocentric videos. |
|
Open-Category Human-Object Interaction Pre-Training via Language Modeling Framework
Sipeng Zheng, Boshen Xu, Qin Jin CVPR, 2023 We introduce OpenCat, a language modeling framework that reformulates HOI prediction as sequence generation. |
AwardsServices |