A vision-language model pretrained to predict human hand keypoints from egocentric video, then fine-tuned with a learned analogical map to robot states, improves CALVIN success rates, especially with only 10% robot data.
Video prediction models as rewards for reinforcement learning
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.CV 1years
2025 1verdicts
CONDITIONAL 1roles
background 1polarities
unclear 1representative citing papers
citing papers explorer
-
AR-VRM: Imitating Human Motions for Visual Robot Manipulation with Analogical Reasoning
A vision-language model pretrained to predict human hand keypoints from egocentric video, then fine-tuned with a learned analogical map to robot states, improves CALVIN success rates, especially with only 10% robot data.