Robocurve
Robocurve is an independent evaluator that measures and publicly reports frontier robotics capabilities and safety. The Public Benefit Corporation builds open-source tools like the Inspect Robots framework, which runs any AI model on any robot embodiment across standardized real-world tasks. Its RoboHarm benchmark assesses refusal rates on harmful instructions, with recent results showing frontier policies completing dangerous tasks at alarming rates.
- Robotics
- Artificial Intelligence
- Data & Analytics
- Developer Tools
- Software Only
- AI Agents
Funding
Raised to date
$500KRaised to date based on public sources. This may differ from the amount the company actually raised and is based only on what is publicly available on the internet.
Founders
Founder details are not available yet.
Product
Problem
Frontier robotics capabilities are advancing rapidly, with labs racing to develop general-purpose robots, yet the public has no continuous, standardized way to measure or compare these systems. The internet is filled with cherry-picked demo videos, making it difficult to assess where the actual frontier of robotics capability lies. This lack of independent evaluation makes it hard for society, policymakers, and researchers to understand the pace and implications of physical AI progress.
Solution
Robocurve operates as an independent evaluator for robotics, measuring capabilities in the real world without any agenda from frontier labs. The company builds open-source tools and independent benchmarks, including the Inspect Robots evaluation framework, which allows researchers to run any AI model on any robot embodiment across standardized tasks. Robocurve reports its findings to the public rigorously and neutrally, covering general-purpose robots, vision-language-action models, and large language models controlling physical systems. The organization is legally bound as a Public Benefit Corporation to serve the public good and supports academic research through funding and hardware provision.
Target Audience
Primary audiences are academic researchers, government policymakers, civil society organizations, and robotics developers at frontier labs seeking independent, standardized assessments of physical AI capabilities. The company also engages a broader public interested in understanding the societal implications of general-purpose robot deployment.
Features
- Open-source Inspect Robots framework released under MIT license, free to use, modify, and distribute
- Real-world-first evaluation with simulation support for VLA, WAM, and LLM-based robot control
- First-class integrations with robotics ecosystems including ROS, Isaac Lab, Cap-X, and XPolicyLab
- Full trace logs and live Rerun visualization for transparent, reproducible benchmark results
- Independent research agenda with no direction from frontier labs on methodology or public results
- Academic support program offering $500k in total funding and free YAM arms for open-source benchmark development
- Standardized evaluation of tasks such as Tower of Hanoi, block stacking, Jenga block pulling, and the RoboHarm safety benchmark assessing refusal rates on harmful instructions