Scholar
Wanpeng Zhang
Google Scholar ID: _IKNf9EAAAAJ
Ph.D. Candidate, Peking University
Citations & Impact
Citations
257
H-index
8
i10-index
8
Publications
18
Co-authors
12
list available
Publications
Resume (English only)
Academic Achievements
- “Being-H0: Vision-Language-Action Pretraining from Large-Scale Human Videos” (arXiv): Introduced the first dexterous Vision-Language-Action model pretrained from human videos; equal contribution
- “Unified Multimodal Understanding via Byte-Pair Visual Encoding” (ICCV’25 Highlight): Developed a complete training framework and the Being-VL-0.5 model
- “From Pixels to Tokens: Byte-Pair Encoding on Quantized Visual Modalities” (ICLR’25): Proposed a BPE tokenizer for images, enabling more effective multi-modal alignment in Transformers
- “VideoOrion: Tokenizing Object Dynamics in Videos” (ICCV’25): Encodes object dynamics using a two-branch architecture with object tokens
- “OpenMMEgo: Enhancing Egocentric Understanding for LMMs with Open Weights and Data” (NeurIPS’25): Enhanced egocentric video understanding via synthetic data and token compression
- “LLM-Based Explicit Models of Opponents for Multi-Agent Games” (NAACL’25): Proposed EMO for modeling opponents in multi-agent settings using LLMs
- “Tackling Non-Stationarity in Reinforcement Learning via Causal-Origin Representation” (ICML’24): Addressed non-stationarity through causal-origin representations
- “AdaRefiner: Refining Decisions of Language Models with Adaptive Feedback” (NAACL’24): Enabled co-learning between LLMs and RL agents via mutual feedback
- “Entity Divider with Language Grounding in Multi-Agent Reinforcement Learning” (ICML’23): Equal contribution
