Radar ·
SmolDataEnvs collects 5,000 verifiable RL tasks
Launched on September 25, 2026, the open-source collection includes tasks, environments, evaluations, and training to optimize small models for programming and data science.
SmolDataEnvs was announced on September 25, 2026, as an open-source collection of 5,000 verifiable tasks in reinforcement learning (RL) environments. The release includes environments, evaluations, and training, with a focus on optimizing small models for programming and data science.
Engineers can use the included tasks and evaluations to train and assess small models. To check the scope and findings, consult the original collection and inspect its available tasks and evaluations. If you use AI to study or apply the material, do not submit your organization’s confidential code, data, or information without authorization.
Original source: SmolDataEnvs: 5.000 tarefas verificáveis de RL