AutoSynthData

ServiceNow CoreAI's framework that turns an enterprise agent's capability gaps into targeted synthetic training data.

ServiceNow · Data, training, and fine-tuning platforms (datasets, labeling, RAG/vector DBs, fine-tuning-as-a-service) · framework

Published in Feed

Overview

AutoSynthData is a framework from ServiceNow CoreAI, published October 2, 2026 on Hugging Face, that converts an enterprise agent's observed failures into new training tasks. It compares a target model's failures against a stronger teacher model's successes to decide what the model should learn next, then generates and validates new tasks exercising those capabilities, shifting the curriculum toward whatever the model still finds difficult as it improves. On EnterpriseOps Gym's Hybrid and ITSM environments, fine-tuning Gemma-4-26B-A4B-it on AutoSynthData-generated data improved mean Pass@1 by 7.2 percentage points (35% relative) and raised verifier success from 63.01% to 68.55%. ServiceNow released the EnterpriseOps Gym dataset alongside the framework to support further research into synthetic data for enterprise agents.

This is a sourced directory entry. Features and capabilities are reported in the linked sources; this entry does not document independent Aymorphic testing.

Key features

  • Targets training data generation at a model's demonstrated capability gaps, not generic synthetic data
  • Uses teacher-model successes vs. target-model failures to pick what to generate next
  • Curriculum shifts automatically toward the model's remaining weak spots as it improves
  • Validated on EnterpriseOps Gym's Hybrid and ITSM environments
  • Ships with the open EnterpriseOps Gym dataset release

Use cases

  • Fine-tuning enterprise agents on organization-specific workflows, tool integrations, and policy constraints
  • Closing measured capability gaps in IT service management (ITSM) and hybrid enterprise agent tasks
  • Generating targeted synthetic data instead of broad, generic synthetic datasets

Pricing

Open framework and dataset (EnterpriseOps Gym) published on Hugging Face; no separate licensing fee disclosed.

Pricing reflects the source record and may change. Confirm current terms with the provider.

Dates and sources

Released
2 Oct 2026
Record updated
3 Oct 2026
Record last checked
3 Oct 2026

Explore the Feed

Other entries filed under data, training, and fine-tuning platforms (datasets, labeling, rag/vector dbs, fine-tuning-as-a-service); inclusion does not imply equivalent features.

  • DigitalOcean Agent Data Services

    DigitalOcean's managed data layer for agents: ingestion flows, a catalog with vectors, and SQL or vector query, in private preview.

  • pplx-embed-v2-late

    Perplexity's late-interaction multimodal embedding models (0.6B and 9B) with 128-dim per-token vectors, open on Hugging Face.

  • EmbeddingGemma 2

    Open-weight 740M-parameter multimodal embedding model built on Gemma 4 that encodes text, images, audio, and video for on-device search.

Browse the Feed