# AutoSynthData

ServiceNow CoreAI introduces AutoSynthData, which uses a target model's failures and a stronger teacher's successes to locate capability gaps and then generates and validates new executable tasks (system specification, user prompt, verifier); on the Hybrid domain of EnterpriseOps Gym, with Gemma-4-26B-A4B-it as target and Qwen3.8-27B as teacher, it generated 2,000 samples in about 18 hours, and after fine-tuning the best checkpoint (epoch 5) raised mean Pass@1 by 7.2 percentage points (a 35% relative improvement), lifted verifier success from 63.01% to 68.55%, and closed 59% of the original Pass@1 gap between Gemma and the reference model; on ITSM, with DeepSeek-V4.1-Flash as teacher, it generated 1,994 samples in 66 hours and raised mean Pass@1 from 18.77% to 27.18%.

ServiceNow CoreAI introduces AutoSynthData, which uses a target model's failures and a stronger teacher's successes to locate capability gaps and then generates and validates new executable tasks (system specification, user prompt, verifier); on the Hybrid domain of EnterpriseOps Gym, with Gemma-4-26B-A4B-it as target and Qwen3.8-27B as teacher, it generated 2,000 samples in about 18 hours, and after fine-tuning the best checkpoint (epoch 5) raised mean Pass@1 by 7.2 percentage points (a 35% relative improvement), lifted verifier success from 63.01% to 68.55%, and closed 59% of the original Pass@1 gap between Gemma and the reference model; on ITSM, with DeepSeek-V4.1-Flash as teacher, it generated 1,994 samples in 66 hours and raised mean Pass@1 from 18.77% to 27.18%.