Microsoft's OfficeTown: generative office simulations yield 406 agent benchmark tasks
microsoft · hf · 2026-10-19
Microsoft introduces OfficeTown, a generative simulation framework creating persistent professional organizations whose work evolves over simulated time—assignments carry over, files accumulate (497 files over 181 simulated days in one org), and roles split between principals directing and assistants executing in authorized workspaces with private vs. shared files.
The framework generates work-grounded training data and evaluation benchmarks: 406 benchmarking tasks, 2,202 SFT trajectories, and 828 OPD tasks, with deliverables like editable DOCX memos, invoices, and dispatch jobs for evolving agent intelligence.
More from coding & agent
- Qwen Code v0.25.3 Ships Managed-Agent Fixes and Web-Shell Tweaks — qwen-code-review-bot · 2026-10-19
- Amazon Paper: QA Scores Miss Agent Memory Failures by 16.8 Points, New FEB-30 Benchmark Fixes It — rohanpaul_ai · 2026-10-19
- How One Founder Used Grok Bot to Submit Backlinks to 4000+ Sites in a Single Day — sujingshen · 2026-10-19
- Netflix shares how it built a knowledge base for production AI agents — AxSaucedo · 2026-10-19
- Reverse Engineer Anything crosses 300k GitHub stars as a movement — FinanceYF5 · 2026-10-19
- Letting Claude Code fill out job applications without hitting Submit — Serious-Werewolf7613 · 2026-10-19