APEX–Agents is a professional task evaluation benchmark for agents (Agent) released by Mercor in 2026, with the related paper available at APEX–Agents, designed to assess the ability of AI Agents to perform long-horizon, cross-application professional service tasks.

The dataset includes 480 tasks covering three professional domains: investment banking, management consulting, and corporate law. Tasks require agents to complete complex operations in realistic work environments involving documents, spreadsheets, PDFs, emails, chats, and calendars, with scoring based on standardized evaluation rules.

Dataset Composition

The dataset primarily comprises the following three professional domains:

• Investment Banking: 160 tasks and 10 world environments;
• Management Consulting: 160 tasks and 11 world environments;
• Law: 160 tasks and 12 world environments.

Each task primarily includes the following components:

• Prompt: Task instructions provided to the Agent;
• Rubric: Standardized scoring rules for evaluating task completion;
• Gold Output: Reference outputs created by experts;
• Metadata: Metadata such as task type, workflow, output format, and estimated completion time;
• World Context: The work environment, files, and other resources corresponding to the task.

Citation

@misc{vidgen2026apexagents,
  title        = {APEX--Agents},
  author       = {Vidgen, Bertie and Mann, Austin and Fennelly, Abby and Wright Stanly, John and Rothman, Lucas and Burstein, Marco and Benchek, Julien and Ostrofsky, David and Ravichandran, Anirudh and Sur, Debnil and Venugopal, Neel and Hsia, Alannah and Robinson, Isaac and Huang, Calix and Varones, Olivia and Khan, Daniyal and Haines, Michael and Richards, Zach and Mahapatra, Chirag and Foody, Brendan and Nitski, Osvald},
  year         = {2026},
  howpublished = {arXiv},
  url          = {https://arxiv.org/pdf/2601.14242}
}