WorkGenesis creates realistic work scenarios to train language agents
WorkGenesis: Building the Worlds That Teach Agents to Work
Artificial Intelligence
Summary
Teaching language AI agents to do real jobs needs realistic practice tasks, but making these tasks by experts is slow and expensive. The authors introduce WorkGenesis, a system that builds realistic job tasks from real public documents guided by occupational knowledge. It also checks and improves the tasks by comparing what a model produces with expected results until the tasks pass quality checks. This approach helped train a language agent that outperforms similar systems on several benchmarks.
What this means in practice
- •For ai product teams: Create large-scale realistic training datasets for language agents that perform work tasks to improve their reliability and scope.
- •For software developers: Build tools that automatically generate and validate work scenarios to train or test AI systems on professional tasks.
Authors
Xinyu Zhu, Fenyi Liu, Yuzhu Cai, Shuo Tang, Rui Ye, Linfeng Zhang, Siheng Chen
Abstract
The ability of Large Language Model (LLM) agents to complete daily and professional work is receiving increasing attention. Training such agents requires realistic work scenarios. Expert-authored occupational work is costly and slow to produce, while unconstrained synthesis often yields tasks with weak factual grounding or internally inconsistent requirements. To bridge this gap, we introduce WorkGenesis, a framework that constructs executable occupational work from real-world artifacts through two core technical innovations: (1) Evidence-Based Work Construction, which grounds each unit of work in real-world evidence by retrieving public files guided by O*NET occupational knowledge and synthesizing the surrounding context, companion materials, work request, and itemwise rubric around them; and (2) Execution-Guided Consistency Verification, which renders a reference deliverable inside the constructed work, attributes every unsatisfied rubric item to the agent, the task, or the rubric, and uses task and rubric defects as feedback to iteratively repair the work until it passes the audit. Experimental results demonstrate that Fx-Work-35B, trained with simple supervised fine-tuning (SFT) on only 20K units of work synthesized by WorkGenesis, achieves the highest scores among all comparable-scale baselines on the five reported metrics across GDPvalAA-v2, APEX-Agents-AA, and JobBench (31.00 versus 24.79 average score), and even surpasses frontier models such as the 1.6T DeepSeek-V4-Pro-Preview. These results show that WorkGenesis provides scalable training data for working agents.