Sovereign 8s.is mirror of standardgalactic/Token-Warping-MLLM
Find a file
Repository files (latest commit first)
Filename Latest commit message Latest commit date
2026-03-31 15:57:20 +09:00
assets init: new repository with readme including tl;dr and code release todos 2026-03-31 15:57:20 +09:00
.DS_Store init: new repository with readme including tl;dr and code release todos 2026-03-31 15:57:20 +09:00
README.md init: new repository with readme including tl;dr and code release todos 2026-03-31 15:57:20 +09:00

Token Warping Helps MLLMs Look from Nearby Viewpoints

Phillip Y. Lee*, Chanho Park*, Mingue Park, Seungwoo Yoo, Juil Koo, Minhyuk Sung
KAIST
CVPR 2026
(* denotes equal contribution)

TL;DR

We propose a training-free method that warps image tokens (instead of pixels) to help MLLMs reason about how a scene looks from nearby viewpoints. Our backward token fetching approach achieves state-of-the-art performance while being robust to depth estimation noise.

Code Release

Coming soon! We are currently preparing the codebase for public release. Stay tuned.

  • Inference code
  • ViewBench benchmark data
  • Evaluation scripts

Citation

If you find our work useful, please consider citing:

@inproceedings{lee2026tokenwarping,
  title={Token Warping Helps MLLMs Look from Nearby Viewpoints},
  author={Lee, Phillip Y. and Park, Chanho and Park, Mingue and Yoo, Seungwoo and Koo, Juil and Sung, Minhyuk},
  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
  year={2026}
}

Acknowledgements

We thank Daehyeon Choi and Sangwoo Youn for their valuable discussions. This work was supported by the National Research Foundation of Korea (NRF), the Institute of Information & Communications Technology Planning & Evaluation (IITP), the Industrial Technology Innovation Program, and the National Supercomputing Center, funded by the Korean government (MSIT/MOTIE).