OccuBench: Evaluating AI Agents on Real-World Professional Tasks via Language Environment Simulation

von AI Research Today