Reproduce “FACT: A Multimodal Force-Action-Track Foundation Model for Human-to-Robot Dexterity” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Zhi Cao, Yining Hong, Jiaolong Yang, Yejin Choi, Li Fei-Fei, Jiajun Wu, Huang Huang
FACT plans pipette adjustment from 420 microliters toward 400 microliters by predicting visual motion and refining actions while predicting future fingertip force. In ten fine-tuning trials, it engaged the dial five times, with 6.4 microliters mean absolute error on engaged trials; T-Rex engaged twice, with 14 microliters error. Results are limited to this task-specific setup.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.