Publications | ENCODE Lab

2025

NeurIPS’25

HoliTom: Holistic Token Merging for Fast Video Large Language Models

Kele Shao, Keda Tao, Can Qin, Haoxuan You, Yang Sui, and Huan Wang^†

NeurIPS, 2025

@inproceedings{shao2025holitom,
  title = {HoliTom: Holistic Token Merging for Fast Video Large Language Models},
  author = {Shao, Kele and Tao, Keda and Qin, Can and You, Haoxuan and Sui, Yang and Wang, Huan},
  booktitle = {NeurIPS},
  year = {2025},
}

NeurIPS’25

Poison as Cure: Visual Noise for Mitigating Object Hallucinations in LVMs

Kejia Zhang, Keda Tao, Jiasheng Tang, and Huan Wang^†

NeurIPS, 2025

arXiv Bib Code Website

@inproceedings{zhang2025poison,
  title = {Poison as Cure: Visual Noise for Mitigating Object Hallucinations in LVMs},
  author = {Zhang, Kejia and Tao, Keda and Tang, Jiasheng and Wang, Huan},
  booktitle = {NeurIPS},
  year = {2025},
}

arXiv’25/07

When Tokens Talk Too Much: A Survey of Multimodal Long-Context Token Compression across Images, Videos, and Audios

Kele Shao^*, Keda Tao^* , Kejia Zhang, Sicheng Feng, Mu Cai, Yuzhang Shang, Haoxuan You, Can Qin, Yang Sui, and Huan Wang^†

arXiv preprint arXiv:2507.20198, 2025

arXiv Bib Paper Repo Database

@article{shao2025tokens,
  title = {When Tokens Talk Too Much: A Survey of Multimodal Long-Context Token Compression across Images, Videos, and Audios},
  author = {Shao, Kele and Tao, Keda and Zhang, Kejia and Feng, Sicheng and Cai, Mu and Shang, Yuzhang and You, Haoxuan and Qin, Can and Sui, Yang and Wang, Huan},
  journal = {arXiv preprint arXiv:2507.20198},
  year = {2025},
  database = {https://oasis-paddleboat-fc1.notion.site/when-tokens-talk-too-much-database},
  paper_repo = {https://github.com/cokeshao/Awesome-Multimodal-Token-Compression},
}

arXiv’25/05
Can MLLMs Guide Me Home? A Benchmark Study on Fine-Grained Visual Reasoning from Transit Maps

Sicheng Feng^* , Song Wang^*, Shuyi Ouyang, Lingdong Kong, Zikai Song, Jianke Zhu, Huan Wang^† , and Xinchao Wang

arXiv preprint arXiv:2505.18675, 2025

Abs arXiv Bib Code Website Dataset 量子位

Multimodal large language models (MLLMs) have recently achieved significant progress in visual tasks, including semantic scene understanding and text-image alignment, with reasoning variants enhancing performance on complex tasks involving mathematics and logic. However, their capacity for reasoning tasks involving fine-grained visual understanding remains insufficiently evaluated. To address this gap, we introduce ReasonMap, a benchmark designed to assess the fine-grained visual understanding and spatial reasoning abilities of MLLMs. ReasonMap encompasses high-resolution transit maps from 30 cities across 13 countries and includes 1,008 question-answer pairs spanning two question types and three templates. Furthermore, we design a two-level evaluation pipeline that properly assesses answer correctness and quality. Comprehensive evaluations of 15 popular MLLMs, including both base and reasoning variants, reveal a counterintuitive pattern: among open-source models, base models outperform reasoning ones, while the opposite trend is observed in closed-source models. Additionally, performance generally degrades when visual inputs are masked, indicating that while MLLMs can leverage prior knowledge to answer some questions, fine-grained visual reasoning tasks still require genuine visual perception for strong performance. Our benchmark study offers new insights into visual reasoning and contributes to investigating the gap between open-source and closed-source models.
@article{feng2025canmllms, title = {Can MLLMs Guide Me Home? A Benchmark Study on Fine-Grained Visual Reasoning from Transit Maps}, author = {Feng, Sicheng and Wang, Song and Ouyang, Shuyi and Kong, Lingdong and Song, Zikai and Zhu, Jianke and Wang, Huan and Wang, Xinchao}, journal = {arXiv preprint arXiv:2505.18675}, year = {2025}, dataset = {https://huggingface.co/datasets/FSCCS/ReasonMap}, qbitai = {https://mp.weixin.qq.com/s/sPJLQtHgl5DZghWLWa_H3Q}, }

CVPR’25

DyCoke: Dynamic Compression of Tokens for Fast Video Large Language Models

Keda Tao, Can Qin, Haoxuan You, Yang Sui, and Huan Wang

CVPR, 2025

arXiv Bib Code

@inproceedings{tao2025dycoke,
  title = {DyCoke: Dynamic Compression of Tokens for Fast Video Large Language Models},
  author = {Tao, Keda and Qin, Can and You, Haoxuan and Sui, Yang and Wang, Huan},
  booktitle = {CVPR},
  year = {2025},
}

ICLR’25

Accessing Vision Foundation Models at ImageNet-level Costs

Yitian Zhang, Xu Ma, Yue Bai, Huan Wang, and Yun Fu

ICLR, 2025

arXiv Bib Code

@inproceedings{zhang2024accessing,
  title = {Accessing Vision Foundation Models at ImageNet-level Costs},
  author = {Zhang, Yitian and Ma, Xu and Bai, Yue and Wang, Huan and Fu, Yun},
  booktitle = {ICLR},
  year = {2025},
}

2024

arXiv’24/11

Is Oracle Pruning the True Oracle?

Sicheng Feng, Keda Tao, and Huan Wang

arXiv preprint arXiv:2412.00143, 2024

arXiv Bib Code Website

ACM MM’24

Towards Real-time Video Compressive Sensing on Mobile Devices

Miao Cao , Lishun Wang, Huan Wang , Guoqing Wang, and Xin Yuan

ACM MM, 2024

arXiv Bib Code

@inproceedings{cao2024towards,
  title = {Towards Real-time Video Compressive Sensing on Mobile Devices},
  author = {Cao, Miao and Wang, Lishun and Wang, Huan and Wang, Guoqing and Yuan, Xin},
  booktitle = {ACM MM},
  year = {2024},
}

ECCV’24 Oral

A Simple Low-bit Quantization Framework for Video Snapshot Compressive Imaging

Miao Cao , Lishun Wang, Huan Wang, and Xin Yuan

ECCV, 2024

arXiv Bib Code

@inproceedings{cao2024simple,
  title = {A Simple Low-bit Quantization Framework for Video Snapshot Compressive Imaging},
  author = {Cao, Miao and Wang, Lishun and Wang, Huan and Yuan, Xin},
  booktitle = {ECCV},
  year = {2024},
}