| name: APEX-Agents | |
| description: > | |
| APEX-Agents is a benchmark from Mercor for evaluating whether AI agents can | |
| execute long-horizon, cross-application professional services tasks. Tasks | |
| were created by investment banking analysts, management consultants, and | |
| corporate lawyers, and require agents to navigate realistic work environments | |
| with files and tools (e.g., docs, spreadsheets, PDFs, email, chat, calendar). | |
| evaluation_framework: archipelago | |
| tasks: | |
| - id: apex-agents | |
| config: default | |
| split: train | |