Reproduce “Thermo-VLA: A Thermal-Aware Vision-Language-Action Model with Gated Residual Modulation for Robotic Manipulation” 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
Jingxuan Chen, Junshi Luo, Chencheng Dong, Xiaojian Nie, Tongge Huang, Qiang Huang
Thermo-VLA makes thermal cues useful for robotic manipulation by separating cold and hot attributes and using the instruction to add relevant features to an RGB representation before generating actions. Across four thermal-dependent tasks, it reports a mean success rate of 0.86, versus 0.68 for direct-input π₀.₅ (RGBT). This is benchmark evidence; temperature reassignment remains difficult.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.