Dataset · mercor
APEX-Agents: AI agent benchmark
mercor/apex-agents
480 tasks from bankers, consultants, and lawyers to test AI agents in cross-application scenarios.
What's inside
The APEX-Agents benchmark by Mercor in English, with under 1k tasks (480 total). Tasks were created by investment banking analysts, consultants, and corporate lawyers, requiring work with documents, spreadsheets, PDFs, email, chat, and calendars in realistic environments. Includes 33 worlds: 10 for banking, 11 for consulting, and 12 for law.
What you can do with it
- Evaluate an AI agent on long-horizon cross-application tasks.
- Test work with documents, spreadsheets, PDFs, and email.
- Compare agents in banking, consulting, and legal scenarios.
- Languages
- English
- Size
- up to 1K rows
- Downloads this month
- 203.7K
- Likes
- 225
- Access
- on request on Hugging Face
Models trained on it
How to load it
from datasets import load_dataset
ds = load_dataset("mercor/apex-agents")