Reality Model Team

Reality Evaluation Scientist

Answer one question: is the model actually understanding reality? This role should resemble an evaluation researcher at a frontier lab far more than a QA engineer.

What you will do

• Create entirely new benchmarks, for example:
• Object permanence.
• Causality.
• Counterfactual reasoning.
• Multi-step prediction.
• Occlusion recovery.
• Failure detection.
• Long-horizon consistency.
• Physical plausibility.

Apply

Send us your CV and a link to your LinkedIn or anything you have built, shipped, or written. Name the role in the subject line.