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

- Company: ByteDance (bytedance.com)
- Announced: 2026-07-06T16:00:00+00:00
- Category: research-paper
- Subject: Seed / Doubao
- Source: https://seed.bytedance.com/en/blog/edgebench-measuring-real-world-environment-learning-and-discovering-a-new-scaling-law
- Record: https://forck.live/items/4662-edgebench-measuring-real-world-environment-learning-and-discovering-a-new

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.

## Evidence

Verbatim from https://seed.bytedance.com/en/blog/edgebench-measuring-real-world-environment-learning-and-discovering-a-new-scaling-law:

> 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.

---

Record: https://forck.live/items/4662-edgebench-measuring-real-world-environment-learning-and-discovering-a-new
Catalogue: https://forck.live/llms.txt
Feed: https://forck.live/feed.md