RARM

Reference-Anchored Reward Model for robotic policy learning.

🤗 Download model — checkpoint_epoch_45_generic.pt (233 MB)

Code and usage instructions · Project page

Place the downloaded model in the code repository's checkpoints/ directory. The current inference code also requires the separate DINOv3 ViT-B/16 base checkpoint, dinov3_vitb16_pretrain_lvd1689m-73cec8be.pth, in the same directory. See the code repository for setup and usage instructions.