Qianqi Yan, Yichen Guo, Ching-Chen Kuo, Shan Jiang, Hang Yin, Yang Zhao, Xin Eric Wang
View original ↗Develop a lightweight library that tracks token-level source attribution during multimodal inference. This will help developers debug hallucination and improve transparency in MLLMs.
Suggested repo: omnitrace-py
"Know exactly which pixel or word influenced your model's output."
Estimated effort: 60h