Research & insights
A clearer picture of AI performance.
Explore the methods, evidence, and open questions behind population-aware evaluation and assurance.
Explore research themesOur research priorities
Methodology
Population-aware metrics
Move beyond aggregate scores to meaningful subgroup outcomes.
Evaluation
LLM-as-judge reliability
Measure disagreement and calibrate automated assessments.
Applications
Speech & agentic AI
Study real-world conversational and multistep task success.
Evidence
Transparent reporting
Make methods, limitations, and confidence easy to inspect.