Health & medicine
Clinical, care-oriented contexts: symptoms, triage, treatment decisions, patient communication, uncertainty, medical ethics, vulnerable populations. The domain used for the health-only transfer test.
12 domains of realistic scenarios
Each beneficial trait is instantiated across many domains, so it shows up under different surface content, incentives, and failure modes. Health and science are highlighted — they anchor the paper's clearest out-of-distribution transfer tests.
Scientific research & scholarly reasoning
Research practice: hypothesis formation, experimental design, replication, evidence synthesis, peer review, responsible interpretation of findings.
Business & economics
Decisions in organizations and markets: forecasting, incentives, governance, resource allocation, stakeholder tradeoffs.
Engineering & technical operations
Safety-critical work: incident response, debugging, change management under time pressure.
Law, ethics & governance
Rights, due process, procedural fairness, accountability, ethically constrained decisions.
Education & pedagogy
Students, teachers and caregivers: explanation, assessment, developmental appropriateness.
Meta-AI & alignment research
Reflection on AI systems: oversight, interpretability, safety mechanisms, failures under uncertainty.
National security & international relations
Diplomacy, deterrence, crisis management, escalation risks, actors with asymmetric power.
Mathematics & formal reasoning
Proofs, counterexamples, verification, careful management of assumptions.
Games & multi-agent interactions
Planning, negotiation, hidden information, coordination, repeated interaction.
Art, visual art & music
Aesthetic judgment, authorship, critique, curation, audience impact.
Creative writing
Story structure, characterization, world-building, representation choices.