apex-agents / eval.yaml
abridges's picture
Duplicate from mercor/apex-agents
04b8612
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