Intel / article

Edgebench Measuring Real World Environment Learning And Discovering A New Scaling Law

Source: seed.bytedance.com
Author: ByteDance Seed
Date: Published Jul 7, 2026

Context

EdgeBench measures how an improves after receiving feedback over hours of work. ByteDance Seed built 134 tasks that support at least 12 hours of interaction, with agents revising their attempts in response to scores, errors and hints. Its research account reports that individual progress varies substantially, while average performance across tasks follows a closely fitted growth curve. The useful distinction is between starting competence and the ability to improve within a task: an agent’s first answer can miss capabilities that emerge through sustained experimentation.

Terms in this piece · Glossary

AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.
grounding — Tying a model's answers to checkable sources — retrieved documents, live data, tool results — instead of letting it answer from memory alone.
agent harness — The scaffolding around a model that turns it into a working agent — the loop, the tools it can call, and the rules for when to stop.

Read the source seed.bytedance.com