Structured evaluation sets that test models in realistic worker–agent interactions.

Examples include:

• Building a worker profile (multi-job history, skills, education).

• Negotiating shift schedules & manager approvals.

• Capturing financial motivations and consent.

Each eval comes with task success rates, bias detection metrics, and reward scores, giving hyperscalers a benchmark to stress-test AI fairness & empathy in Haryanvi Hindi.