Reward Model Card for robometer
Robometer is a zero-shot general-purpose robotic reward model built on a fine-tuned Qwen3-VL backbone with progress, preference, and success heads. Given a video and a task description it outputs a per-frame progress signal in [0, 1] and a per-frame success probability — suitable for offline reward labelling and for low-frequency reward signals during RL fine-tuning of robot policies.
This reward model has been trained and pushed to the Hub using LeRobot. See the full documentation at LeRobot Docs.
How to Get Started with the Reward Model
Train from scratch
lerobot-train \
--dataset.repo_id=${HF_USER}/<dataset> \
--reward_model.type=robometer \
--output_dir=outputs/train/<desired_reward_model_repo_id> \
--job_name=lerobot_reward_training \
--reward_model.device=cuda \
--reward_model.repo_id=${HF_USER}/<desired_reward_model_repo_id> \
--wandb.enable=true
Writes checkpoints to outputs/train/<desired_reward_model_repo_id>/checkpoints/.
Load the reward model in Python
from lerobot.rewards import make_reward_model
reward_model = make_reward_model(pretrained_path="<hf_user>/<reward_model_repo_id>")
reward = reward_model.compute_reward(batch)
Model Details
- License: apache-2.0
- Downloads last month
- 12