ServiceNow releases AutoSynthData for generating agent training data
Understand the story
On October 2, 2026, the Hugging Face Blog published a first-hand report that ServiceNow's CoreAI team launched AutoSynthData, which automatically generates training data for post-training of enterprise agents. The method leverages a target model's failure cases in an environment as well as success trajectories from stronger teacher models to automatically generate executable, realistic, and challenging training tasks. Each task consists of a system spec, a user prompt, and a verifier, and must meet three requirements: feasibility, realism, and difficulty. The team demonstrated the pipeline using a public dataset from EnterpriseOps Gym, dynamically adjusting the curriculum as models improve. No further reports or additional information are available at this time.
AI overview of reporting · Updated 16 days ago
Reporting timeline
Explore different aspects of the story through its reporting.
- Hugging Face BlogAutoSynthData: Generating Training Data for Enterprise Agents
ServiceNow CoreAI launched AutoSynthData, which uses a target model's failures in an environment and successful trajectories from stronger teachers to automatically generate executable, realistic, and hard training tasks for enterprise agent post-training. Each task consists of a system specification, user prompt, and verifier, and must meet feasibility, realism, and difficulty requirements. The team demonstrated the pipeline with a public dataset from EnterpriseOps Gym and dynamically adjusted the curriculum as models improved.