Job description
Neo Research • Singapore / London
London, UK, Singapore
- In this role, you'll run safety evaluations on frontier AI models focusing on loss-of-control and harmful manipulation risks.
- Execute evaluations like CyBench and SHADE-Arena against target models.
- Conduct open-ended research to elicit dangerous behaviours including scheming, sandbagging, and self-replication.
- Draft safety reports documenting evaluation methodology, elicitation techniques, and findings.
- Ensure evaluation rigour through careful scaffolding, tool access, and sampling design.
Neo Research is a research and evaluations lab based in Singapore, working to ensure that frontier AI can be safely deployed at scale.