推荐阅读！
AutoEnvScaling: Automating the Data Flywheel with Terminal Agents
AutoEnvScaling automates terminal agent training with self-built environments
AISummary
AutoEnvScaling is a system in which a proposer agent builds complete Harbor tasks and sandbox environments for terminal agents. According to the post, one model acts as both builder and solver in self-play reinforcement learning, without humans in the loop.
Post on XView on X
Dongxi NLPVerified on X
@dongxi_nlp
Recursive self-improvement needs agents that can build their own training environments. But researchers still part of the loop, making it hard to scale. We introduce AutoEnvScaling, which turns Data Flywheel into a terminal task: a proposer agent builds complete Harbor tasks and sandbox. It builds just like human researchers. This enables self-play RL for terminal agents: one model improves as both builder and solver, without human-in-the-loop.
Source: Dongxi NLP · x.comPublished · added here