AutoEnvScaling: Automating the Data Flywheel with Terminal Agents

Key points

• Terminal agents solve tasks in software engineering and scientific research by executing commands in a terminal.1
• Training them via reinforcement learning (RL) requires diverse and verifiable environments, but these environments are still mostly designed by human researchers, so their supply grows only as fast as researchers can work.1
• We show that training Qwen3.5-35B-A3B on AutoEnvScaling environments from GPT-5.6-sol improves its Terminal-Bench 2.1 score by 4.3 points over the matched Tmax baseline.1
• We show that Qwen3.6-35B-A3B, acting in both roles, gains 12.2 points on Terminal-Bench 2.1 and raises its Terminal-Bench 4.0 score from 1.6 to 3.2.1

Sources (1)

• [1]AutoEnvScaling: Automating the Data Flywheel with Terminal Agents
  Hugging Face Daily Papers · Oct 9, 12:00 AM

  We introduce AutoEnvScaling, which treats environment design as a terminal task and uses agents to automate the data flywheel.

  Terminal agents solve tasks in software engineering and scientific research by executing commands in a terminal.

Extractive summary: sentences quoted from the sources.