
ZJU open-sources SuperNav, an agent harness that navigates without fine-tuning the MLLM
Zhejiang University’s SuperNav paper (arXiv:2610.12126, 8 October) and GitHub release equip a pretrained multimodal model with navigation tools and skills. Authors report 78% single-object success versus 34% for the strongest baseline they tested, and ship Learned Executor weights on Hugging Face.



































