A frozen MLLM with LoRA-on-language and a discrete token waypoint interface navigates four unseen real-world environments after just 8.7 hours of training data.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.RO 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
GemNav: Discrete-Token Visual Robot Navigation using a Multimodal Large Language Model
A frozen MLLM with LoRA-on-language and a discrete token waypoint interface navigates four unseen real-world environments after just 8.7 hours of training data.