Tev: Scaling Long-Horizon Agentic Environments with Dense Progress Rewards
Long-horizon agents have drawn growing attention for their ability to work autonomously -- from hours to days -- on economically valuable tasks, a capability essential to automated discovery and recursive self-improvement (RSI). Yet open-source models still struggle on long-horizon tasks, largely for lack of both training data and a working recipe, and scaling both is hard: lon
Reading in Plain English mode — toggle below to switch
Long-horizon agents have drawn growing attention for their ability to work autonomously -- from hours to days -- on economically valuable tasks, a capability essential to automated discovery and recursive self-improvement (RSI). Yet open-source models still struggle on long-horizon tasks, largely for lack of both training data and a working recipe, and scaling both is hard: lon
This paper addresses a critical bottleneck in developing advanced AI agents: the lack of high-quality, long-horizon training data with dense rewards. By introducing Tev-Gym, it provides a structured approach to creating verifiable environments, which is essential for scaling agent capabilities and moving towards more autonomous and economically valuable AI applications.