RewardScope 🎛️

A web tool to compare popular zero-shot reward functions on your own robot manipulation sensor videos.

Reward functions you can run on your videos:

• Brute Force — at each frame, sends the video up to that point to the VLM and asks for a progress score between 0.0 and 1.0
• Generative Value Learning (GVL) (Nov 2024) — from Vision Language Models are In-Context Value Learners, Google DeepMind
• Robo-Dopamine (Dec 2025) — from Robo-Dopamine: General Process Reward Modeling for High-Precision Robotic Manipulation by @FlagOpen, CVPR 2026
• Robometer (Mar 2026) — from Robometer: Scaling General-Purpose Robotic Reward Models via Trajectory Comparisons by @ygtkorkmaz , @aliang8 , USC
• RoboReward (Jan 2026) — from RoboReward: General-Purpose Vision-Language Reward Model for Robotics by @teetone, @ajwagenmaker, @kpertsch, Stanford & Berkeley
• TOPReward (Feb 2026) — from TOPReward: Token Probabilities as Hidden Zero-Shot Rewards for Robotics by @jcoleharrison , @chinsengi , UW

...and easy to add more!

How to run on your own videos

1. Create an MP4 video of robot manipulation (example). For efficiency, please downsize the file to 480p, as image pixels are passed as tokens.

2. Install prerequisites

   virtualenv venv
   . venv/bin/activate
   pip install -r requirements.txt
   pip install torch torchvision transformers accelerate qwen-vl-utils
   

3. Run the reward algorithms on your video:

   A. Run bruteforce_vlm, gvl, robodopamine, roboreward, and/or topreward

   Run the script to calculate reward functions on your video:

   python run_rewards.py --method bruteforce_vlm,gvl,robodopamine,roboreward,topreward --video <myvideo.mp4> --instruction <instructions e.g. `create a tower of 5 cubes`>
   

   Notes:

   • If you are running gvl or bruteforce_vlm, you must include an OpenAI API key:

   --openai-api-key <your key>
   

   • If you are running topreward, roboreward, and/or robodopamine, you'll need at least 16GB unified/GPU memory.

   B. Run robometer

   Compute the reward using the custom script here.

4. View the results in your browser:

   ./run_viewer.sh
   

TODOs

• add speed comparison (relative time/cpu to run each reward function)