AgentLongBench: A Controllable Long Benchmark For Long-Contexts Agents via Environment Rollouts

Fang, Shicheng, Wang, Yuxin, Liu, Xiaoran · arXiv · 2026