# Environments as Scaffold: Enriching Feedback to Bootstrap Self-Evolving Agents in Long-Horizon Tasks > Source: > Published: 2026-09-09 04:00:04+00:00 Large Language Models demonstrate remarkable proficiency in static reasoning, yet training them as autonomous agents through Reinforcement Learning (RL) for long-horizon tasks is often hindered by severe reward sparsity. While conventional agent-side warming up via supervised fine-tuning (SFT) can a