ByteDance· Seed / DoubaoResearch paper

EdgeBench: Measuring Real-World Environment Learning and Discovering a New Scaling Law

ByteDance released EdgeBench, an ultra-long-horizon benchmark for measuring how agents learn from real-world environments, comprising 134 tasks across six domains. The research finds that agent performance in environment learning follows a log-sigmoid curve with high precision, and that learning speed doubles roughly every three months. 51 tasks and the evaluation framework have been open-sourced.

From the source

We recently released EdgeBench, an ultra-long-horizon benchmark built to measure learning from real-world environments . The benchmark comprises 134 realistic and diverse tasks spanning six major capability domains, each allowing agents to operate continuously for at least 12 hours.

seed.bytedance.com

Announced

06 Jul 2026 16:00 UTC