Researcher, Alignment Science

Sourced from flexmatrix.html-5.me

This role is on OpenAI's Alignment Science team, focusing on scalable methods to make models follow user intent, stay honest about their limitations, and avoid harmful behaviors. The researcher will design and run alignment experiments, train and evaluate models using reinforcement learning, and develop evaluations for failure modes like hallucination and reward hacking. Successful methods will be integrated into frontier model development and published where appropriate.

• Design and implement alignment experiments focused on intent following, honesty, calibration, and robustness.
• Train and evaluate models using reinforcement learning and other empirical ML methods.
• Develop evaluations for failure modes such as hallucination, instruction-following failures, reward hacking, covert actions, and scheming.
• Study methods that encourage models to verify their behavior and report shortcomings honestly, including confession-style training objectives.
• Build monitoring and inference-time interventions that ensure compliant behavior or surface model issues to users or downstream systems.
• Investigate how alignment methods scale with model capability, compute, data, context length, action length, and adversarial pressure.
• Integrate successful techniques into model training and deployment workflows.