GenesisBench
How language intelligence can be used to improve physical intelligence.
How it works
GenesisBench evaluates how LLM agents can train the policies that run on real robot hardware.
Heuristic policies in simulation
No training and no H100s yet — agents iterate on code logic alone, optimizing heuristic policies for performance in simulation.
This stage ships nine tasks — each runs the same improvement loop in its own simulated environment.
Improve a programmatic Ant locomotion policy under a fixed autonomous research budget.
simulation_heuristics_ant_v1 ↗The loop nodes illustrate gait and planning edits for the Ant v1 task; all nine task packages live on GitHub.
Environments
Two environments: simulation for training and evaluation, the real world for operation.
Results
Leaderboards
Five-trial mean ± sample standard deviation for each environment's native raw score. Fail-closed raw-null trials use the starter-equivalent anchor. Dotted markers show the starter; dashed markers show the article reference.
Loading task leaderboards…
Final normalized score
Pooled IQM across all 45 trial-task scores; ± is the sample standard deviation of five per-trial IQMs. Positive display index = IQM + 100; the dotted marker is the starter-level aggregate.
Loading final leaderboard…
Research background
GenesisBench builds on research at the intersection of agentic coding, post-training, and embodied AI.
ENPIRE: Agentic Robot Policy Self-Improvement in the Real WorldNVIDIA GEAR · arXiv:2606.19980 Xiao, Xie, Zhang, Lin, et al. — LLM agents improving robot policies through closed-loop real-world iteration.All references11
- ASPIRE: Agentic Skills Discovery for RoboticsNVIDIA GEAR Lu, Wu, Kou, et al. — agents that discover and refine robot control programs in a self-improving loop.
- ENPIRE: Agentic Robot Policy Self-Improvement in the Real WorldNVIDIA GEAR · arXiv:2606.19980 Xiao, Xie, Zhang, Lin, et al. — LLM agents improving robot policies through closed-loop real-world iteration.
- SimFoundry: Modular and Automated Scene Generation for Policy Learning and EvaluationNVIDIA GEAR Ranawaka, Wong, Pai, et al. — automated scene generation for training and evaluating policies.
- Learning Beyond GradientsEssay Jiayi Weng — heuristic learning: coding agents iterating programmatic policies instead of gradient descent.
- Autoresearch for RoboticsGitHub Autonomous robotics research with simulation feedback.
- AutoresearchGitHub Andrej Karpathy — AI agents running training research automatically.
- How Brittle Are the VLA Models in Robotics?Article Yanan Chen — on the brittleness of today's vision-language-action models.
- RoboArena: Distributed Real-World Evaluation of Generalist Robot PoliciesProject A distributed framework for evaluating robot policies in the real world.
- Qwen RobotSuiteQwen The Qwen team's robotics suite.
- HumanoidMimicGen: Data Generation for Loco-Manipulation via Whole-Body PlanningNVIDIA Lin, Mandlekar, Garrett, et al. — data generation for humanoid loco-manipulation.
- InSight: Self-Guided Skill Acquisition via Steerable VLAsProject Wang, Osterberg, Tian, et al. (Stanford) — self-guided skill acquisition via steerable VLAs.